ALGORITHMIC TRADING

MODULE

NATIONAL STOCK EXCHANGE OF INDIA LIMITED

Test Details

Sr. No. Name of Module Fees (Rs.) Test No. of Maxi- Pass Cer- Nega-
Duration Ques- mum Marks tificate tive
(in min- tions Marks  (%) Validity Marking
utes) (in yrs)

  FOUNDATION  
1 Algorithmic Trading Module 6840 * 120 100 100 60 3 YES
2 Financial Markets: A Beginners’ Module ^ 1710 * 120 60 100 50 5 NO
3 Mutual Funds : A Beginners' Module 1710 * 120 60 100 50 5 NO
4 Currency Derivatives: A Beginner’s Module 1710 * 120 60 100 50 5 NO
5 Equity Derivatives: A Beginner’s Module 1710 * 120 60 100 50 5 NO
6 Interest Rate Derivatives: A Beginner’s Module 1710 * 120 60 100 50 5 NO
7 Commercial Banking in India: A Beginner’s Module 1710 * 120 60 100 50 5 NO
8 FIMMDA-NSE Debt Market (Basic) Module 1710 * 120 60 100 60 5 YES
9 Securities Market (Basic) Module 1710 * 120 60 100 60 5 YES
10 Clearing Settlement and Risk Management Module 1710* 60 75 100 60 3 NO
  INTERMEDIATE  
1 Capital Market (Dealers) Module ^ 1710 * 105 60 100 50 5 YES
2 Derivatives Market (Dealers) Module ^ [Please refer to footnote no. (i) ] 1710 * 120 60 100 60 3 YES
3 Investment Analysis and Portfolio Management Module 1710 * 120 60 100 60 5 YES
4 Fundamental Analysis Module 1710 * 120 60 100 60 5 YES
5 Options Trading Strategies Module 1710 * 120 60 100 60 5 YES
6 Operations Risk Management Module 1710 * 120 75 100 60 5 YES
7 Banking Sector Module 1710 * 120 60 100 60 5 YES
8 Treasury Management Module 1710 * 120 60 100 60 5 YES
9 Insurance Module 1710 * 120 60 100 60 5 YES
10 Macroeconomics for Financial Markets Module 1710 * 120 60 100 60 5 YES
11 NSDL–Depository Operations Module 1710 * 75 60 100 60 # 5 YES
12 Commodities Market Module 2052 * 120 60 100 50 3 YES
13 Surveillance in Stock Exchanges Module 1710 * 120 50 100 60 5 YES
14 Corporate Governance Module 1710 * 90 100 100 60 5 YES
15 Compliance Officers (Brokers) Module 1710 * 120 60 100 60 5 YES
16 Compliance Officers (Corporates) Module 1710 * 120 60 100 60 5 YES
Information Security Auditors Module (Part-1) 2565 * 120 90 100 60 YES
17 2
Information Security Auditors Module (Part-2) 2565 * 120 90 100 60 YES
18 Technical Analysis Module 1710 * 120 60 100 60 5 YES
19 Mergers and Acquisitions Module 1710 * 120 60 100 60 5 YES
20 Back Office Operations Module 1710 * 120 60 100 60 5 YES
21 Wealth Management Module 1710 * 120 60 100 60 5 YES
22 Project Finance Module 1710 * 120 60 100 60 5 YES
23 Venture Capital and Private Equity Module 1710 * 120 70 100 60 5 YES
24 Financial Services Foundation Module ### 1140 * 120 45 100 50 NA NO
25 NSE Certified Quality Analyst $ 1710 * 120 60 100 50 NA YES
  ADVANCED  
1 Financial Markets (Advanced) Module 1710 * 120 60 100 60 5 YES
2 Securities Markets (Advanced) Module 1710 * 120 60 100 60 5 YES
Derivatives (Advanced) Module YES
3 1710 * 120 55 100 60 5
[Please refer to footnote no. (i) ] YES
4 Mutual Funds (Advanced) Module 1710 * 120 60 100 60 5 YES
5 Options Trading (Advanced) Module 1710 * 120 35 100 60 5 YES
2500 per
6 FPSB India Exam 1 to 4** 120 75 140 60 NA N/A
exam *
7 Examination 5/Advanced Financial Planning ** 6000 * 240 30 100 50 NA N/A
8 Equity Research Module ## 1710 * 120 49 60 60 2 YES
9 Issue Management Module ## 1710 * 120 55 70 60 2 YES
10 Market Risk Module ## 1710 * 120 40 65 60 2 YES
11 Financial Modeling Module ### 1140 * 120 30 100 50 NA NO
12 Certified Credit Research Analyst (CCRA) – Level 1 1710* 120 80 120 60 NA NO
13 Certified Credit Research Analyst (CCRA) – Level 2 2565* 180 85 120 60 NA NO
                 
1 NSE’s Capital Market Aptitude Test (NCMAT) 1710* 120 100 100 60 NA NO
* in the ‘Fees’ column - indicates module fees inclusive of service tax
^ Candidates have the option to take the tests in English, Gujarati or Hindi languages.
# Candidates securing 80% or more marks in NSDL-Depository Operations Module ONLY will be certified as ‘Trainers’.
** Following are the modules of Financial Planning Standards Board India (Certified Financial Planner Certification)
- FPSB India Exam 1 to 4 i.e. (i) Risk Analysis & Insurance Planning (ii) Retirement Planning & Employee Benefits (iii) Investment
Planning and (iv) Tax Planning & Estate Planning
- Examination 5/Advanced Financial Planning
## Modules of Finitiatives Learning India Pvt. Ltd. (FLIP)
### Module of IMS Proschool
$ Module of SSA Business Solutions (P) Ltd.
~ Module of Association of International Wealth Management of India
The curriculum for each of the modules (except Modules of Financial Planning Standards Board India, Finitiatives Learning India Pvt.
Ltd. and IMS Proschool) is available on our website: [Link] > Education > Certifications.
Note: (i) SEBI vide notification no. LAD-NRO/GN/2012-13/30/5474 dated January 11, 2013 has specified the NISM-Series-VIII-
Equity Derivatives Certification Examination as the requisite standard for associated persons functioning as approved users and sales
personnel of the trading member of an equity derivatives exchange or equity derivative segment of a recognized stock exchange.
Contents

Chapter 1 Introduction to Algorithmic Trading..................................................6

1.1 Evolution of Algorithmic Trading.............................................................6

1.2 What is Algorithmic Trading...................................................................7

1.3 Different Trading Methodologies.............................................................8

1.4 Trends in Algorithmic Trading: Global and India.......................................8

1.5 Benefits of Algorithmic Trading..............................................................9

Chapter 2 Order Types.................................................................................... 12

2.1 Time Conditions................................................................................ 12

2.2 Price Conditions................................................................................. 12

2.3 Other Conditions............................................................................... 13

2.4 Market Orders................................................................................... 14

2.5 Limit Orders...................................................................................... 16

2.6 Immediate-or-Cancel Orders............................................................... 17

2.7 Market with Price Protection................................................................ 18

2.8 Stop Orders...................................................................................... 18

2.9 Disclosed Orders............................................................................... 20

2.10 Execution Trading Strategies............................................................... 20

2.10.1 TWAP Strategy..................................................................... 21

2.10.2 VWAP Strategy..................................................................... 22

2.10.3 Percentage of Volume (POV) Strategy...................................... 22

Chapter 3 Trading Strategies.......................................................................... 24

3.1 Calendar Spread................................................................................ 24

3.2 Cash Future Arbitrage Strategy............................................................ 25

3.3 Index Arbitrage................................................................................. 25

3.4 Pair Trading...................................................................................... 26

3.5 News Based Trading Strategies............................................................ 28

3.6 Conversion- Reversal......................................................................... 28

3
Chapter 4 Algorithmic Trading: System Architecture....................................... 32

4.1 Market Data...................................................................................... 34

4.2 CEP Engine....................................................................................... 36

4.3 Order Routing / Order Manager........................................................... 37

4.4 New Trends-MultiTap.......................................................................... 39

4.5 Colocation........................................................................................ 39

4.6 Trading Setup.................................................................................... 39

4.7 Smart Order Routing (SOR)................................................................ 41

4.8 Connectivity Options.......................................................................... 42

Chapter 5 Risk management in Algorithmic Trading........................................ 45

5.1 Different Stages involved in Risk Management....................................... 45

5.2 Risk Management Specific to High Frequency & Algorithmic Trading.......... 51

Chapter 6 Audit and Compliance process........................................................ 57

6.1 International Organization of Securities Commissions............................. 57

6.2 Auditing Process and Requirements

(As defined by NSE for member-broker)............................................... 59

6.3 SEBI Recommendations on Audit......................................................... 59

6.4 Exchange Audits................................................................................ 61

6.5 Technology and System Audit.............................................................. 63

6.6 Compliance Requirements................................................................... 68

4
 Distribution of weights of the
Algorithmic Trading Module Curriculum

Chapter No. Title Weights (%)

1 Introduction to Algorithmic Trading 12
2 Order Types 15
3 Trading Strategies 19
4 System Architecture 19
5 Risk Management in Algorithmic Trading 19
6 Audit and Compliance Processes 16

Note: C
 andidates are advised to refer to NSE’s website: [Link], click on ‘Education’
link and then go to ‘Updates & Announcements’ link, regarding revisions / updations in
NCFM modules or launch of new modules, if any.

The content has been developed by QuantInsti, which is a pioneering institute for Algorithmic
& Quantitative Trading education in Asia.

Copyright © 2013 by National Stock Exchange of India Ltd. (NSE)

Exchange Plaza, Bandra Kurla Complex,
Bandra (East), Mumbai 400 051 INDIA

All content included in this book, such as text, graphics, logos, images, data compilation etc.
are the property of NSE. This book or any part thereof should not be copied, reproduced,
duplicated, sold, resold or exploited for any commercial purposes. Furthermore, the book in
its entirety or any part cannot be stored in a retrieval system or transmitted in any form or by
any means, electronic, mechanical, photocopying, recording or otherwise.

5
Chapter 1 Introduction to Algorithmic Trading

With the advent of the modern computer world exchange based trading has undergone a sea
change. Since historical times, people have used advantage of faster information to derive
profits in trading – be it Rothschild who posted his men at war-sitesso that he could be the
first to learn the outcome of the England-France war and derive abnormal profits or Reuters
who used the pigeons to provide information to his clients faster that were much faster than
the post train, giving faster access to stock news from the Paris stock exchange. Information
today travels at a rate faster than any other time in recorded human history, and therefore
those who utilize this faster pace of transfer of information could benefit in their trading
operations.

As a natural response to this increased flow of information, computers have come to play a
large part in the trading on exchanges - beginning with back office to streamline operations,
to finally sending an order to the exchange based on predefined rules. The usage of computers
for sending an order to the exchange based on predefined rules is broadly called algorithmic
trading.

1.1 Evolution of Algorithmic Trading

The evolution of algorithmic trading can be traced back to both independent sell side and the
buy side endeavours in trying to leverage the falling prices of computational power to create
scalable infrastructural advantages.

As exchanges moved in the direction of electronic trading across the world, with NASDAQ
establishing the electronic board in 1971 and other exchanges following suit, computers came
to play a large part in the market place. This reduced the time required to process the post
trade information and also found way in lives of analysts trying to forecast the markets.

Fischer Black (popularly known for the Black-Scholes option pricing model) asked in 1971
whether trading could be automated and if the market maker (specialist) could be replaced
by a set of instructions using a computer. Within three decades, market forces had given a
response to the question and the answer was “Yes”.

Early to this changing scenario were the buy side and proprietary trading participants who
approached the trading systems and algorithms from the perspective of opportunistic advantage
and developed strategies that could identify profit opportunities and execute profitable trades
before manual traders could react. These activities were demonized when the markets fell
drastically on one single trading session in 1987 (more than 20% drop in prices in a day, and
much of the blame was heaped on program trading which were execution instructions to shift
the portfolio from stocks to bonds and led to a cascading effect in the markets).

6
On the sell side the firms realized that they could create frameworks and offer it to their buy
side clients to facilitate their trading strategies. The first generation of the sell side strategies
were simple order slicing which primarily split large order sizes to small individual orders and
managed the order flow. The later generations evolved into complex systems that seek to
minimize not only the total cost of transaction but also optimize the liquidity constraints.

The growth of algorithmic trading is very closely linked to the development of the electronic
trading market. The important milestones in the global electronic trading industry are
tabulated below

Year Event
1969 Instinet allowed electronic block trading
1971 NASDAQ electronic board started
1972 Cantor establishes first electronic market place for US Government securities
1976 NYSE DOT routes small orders.
1978 ITS links NYSE and other US Stock Exchanges
1980 Instinet enables DMA to US exchanges
1982 Tokyo Stock Exchange introduces Computer assisted Order Routing and
Execution System
1986 Paris Bourse introduced an electronic trading system
1988 MTS platform creates secondary market for Italian government bonds.
1992 CME launches first version of GLOBEX electronic futures platform
1993 EBS adds competition for spot FX
1997 SEC changes order handling rules
1998 Eurex launches first fully electronic exchange for futures
2001 Liquidnet ATS created allowing “dark pool” buy side crossing for equities.
2007 US regulation NMS, European MiFID (Markets in Financial Instruments Directive)
regulation come into force

1.2 What is Algorithmic Trading

An algorithm is a set of instructions for accomplishing a task. Trading using algorithms has
been in the markets for a long time and has been known simply as rule based trading. Trading
using algorithms did not require the computer to send out the eventual order to the exchange
– this is the key step which differentiates rule based trading from algorithmic trading.

Algorithmic trading as defined by Indian Regulator SEBI, vide circular no. MRD/DP/09/2012
dated March 30, 2012 defines algorithmic trading as –Any order that is generated using
automated execution logic shall be known as algorithmic trading.

In the United States, CFTC [Commodity Futures Trading Commission] technical advisory
group on Automated and High Frequency Trading defines algorithmic trading as [use of]
algorithms for decision making, order initiation, generation, routing, or execution, for each
individual transaction without human direction; Orders sent to an electronic marketplace
must be generated by computer based decision making “the algorithm” as opposed to being
sent by a human (e.g. “point and click”).

7
For facilitating the human-machine interaction of any particular algorithmic trading strategy
there could be a set of values that are filled in by the trader, these values are called parameters
and are used by the algorithm to evaluate the mathematical expression which results in
estimation of order details - the price, size and side of the order.

1.3 Different Trading Methodologies

Following is the classification of different kind of algorithmic trading strategies based on the
typical holding period:

Strategy Description Typical Holding Period

Automated liquidity provision <1 minute

Quantitative algorithms for optimal pricing and execution of
market-making positions

Arbitrage Strategies No specific time period

Index Arbitrage but generally <1 minute

Put-Call-Parity during volatile markets

Machine Readable News Based <1 minute

Market microstructure trading <10 minutes
Identifying trading party order flow through reverse
engineering of observed quotes
Event trading Short-term trading on macro events <1 hour
Deviations arbitrage Statistical arbitrage of deviations <1 day
from equilibrium: triangle trades,
basis trades, and the like
Execution Strategies Order Size dependent
Volume Weighted Average Price
Time Weighted Average Price
Percentage Of Volume
Portfolio Trading No specific time period
Smart Order Routing No specific time period

1.4 Trends in Algorithmic Trading: Global and India

Algorithmic Trading Trends in Developed Markets

American markets and European markets generally have a higher proportion of algorithmic
trades than other markets, and estimates since 2008 have ranged as high as 73% proportion
in some markets.

8
Various studies indicate more than 50% of the US equity trading volume was through
algorithmic trading with High Frequency firms being the highest contributor to the volumes.

Equity Derivative markets are fairly easy to integrate into algorithmic trading along with
option trading and multiple legs can be executed simultaneously.

In the Foreign exchange markets

In the foreign exchange markets, use of algorithmic trading strategies increased to 16%
in 2011 from 12% in 2010 among the market's biggest and most active traders - those
generating more than $50 billion in annual FX trading volume.

Algorithmic TradingTrends in India

India has gone through a recent but swift algorithmic trading revolution since the regulator
SEBI permitted Direct Market Access (DMA), i.e. a facility that allows brokers to offer clients
direct access to the exchange trading system through the broker’s infrastructure without
manual intervention by the broker, in 2008. Order messages have surged from 30 million per
day five years ago to more than 300 million on the National Stock Exchange of India (NSE)
alone in 2012. Algorithmic trading based volume contributes a significant percentage to the
overall exchange volumes.

Milestones in India

Year Event
Apr, 2008 SEBI allows Direct Market Access (DMA).
Aug, 2010 Smart Order Routing Introduced.
Mar, 2012 SEBI issues broad guidelines for algorithmic trading
Jan, 2013 Forward Market Commission (FMC) issues guidelines for algorithmic trading
in commodities.

1.5 Benefits of Algorithmic Trading

The rapid growth of trading activities involving algorithms has led to the debate on the
advantages and disadvantages of algorithmic trading in the print media as well as academic
literature. Algorithmic trading has led to an improvement of the basic quality of the market,
and potentially decreased volatility over short time-frames along with increased liquidity. The
advantages can be classified broadly from the perspective of a retail investor and also from
the perspective of the market participant who is trading using the algorithms.

Stability

Algorithmic trading brings stability across various asset classes. Unlike manual trading, which
is limited in approaching activities in a sequential manner; algorithms can process information

9
in parallel and take actions. Due to this any price sensitive information is immediately reflected
in all the asset classes simultaneously and therefore a relatively more stable market is created.

Liquidity

Algorithms can react quickly to any new information flow and therefore they can update
their quotes much faster than humans; this leads to a narrower bid/ask spread because
the possibility of trades happening on stale information reduces and consequently losses
due to adverse selection are lower. This significantly reduces the impact cost for the market
participants.

Increased Market Efficiency

As new information gets reflected in the price almost immediately it increases the efficiency of
the market and aids price discovery in a more orderly method. This leads to a more efficient
market and therefore an increase in welfare of all market participants.

Uptime

These systems are able to operate all across the market hours and decisions can be made
immediately and don’t suffer from fatigue.

Risk Management

Risk management is automated using computers and it helps in eliminating the possible
operational human error

Speed

Lead to a faster decision making process devoid of emotions

Other advantages include reduction in cost per trade and consistent trade trail meets financial
control and regulatory requirements.

10
Self assessment test:

• For algorithmic trading, firms aiming for higher profits have invested a lot in having
faster access to:

– Insider information

– Information

– Coolants

– None of the above

• The first trading destination that allowed electronic block trading is:

– NASDAQ

– NYSE

– Instinet

– LSE

• Which of the following is least likely to be an algorithmic trading order?

– VWAP order

– TWAP order

– Point and Click order

– All of the above

• On NSE, Algorithmic Trading is only allowed for Equity Futures and Options.

a. True

b. False

• Dark pools for trading equities are available for Indian stocks

a. True

b. False

11
Chapter 2 Order Types

Orders are instructions sent to exchange by participants regarding their trading intentions.
Orders allow market participants to communicate their requirements, in terms of Time, Price
and Conditionality. The understanding of basic order types and their impacts on order book is
important in order to develop and optimize algorithmic trading strategies.

Exchanges support various order types in order to cater to needs of traders and investors. The
order types can be broadly divided into the following categories –

• Time Conditions

• Price Conditions

• Other (including conditionality) Conditions

2.1 Time Conditions

• GTC - A Good Till Cancelled (GTC) order is an order that remains in the system until it
is cancelled by the Trading Member. It will therefore be able to span trading days if it
does not get matched. The maximum number of days a GTC order can remain in the
system is notified by the Exchange from time to time.

• GTD - A Good Till Days/Date (GTD) order allows the Trading Member to specify the
days/date up to which the order should stay in the system. At the end of this period the
order will get flushed from the system. Each day/date counted is a calendar day and
inclusive of holidays. The days/date counted is inclusive of the day/date on which the
order is placed. The maximum number of days a GTD order can remain in the system
is notified by the Exchange from time to time.

• Day Order: A day order is valid for the day on which it is entered. If the order is not
executed during the day, the system cancels the order automatically at the end of the
Day.

• Immediate or Cancel (IOC) Order: An IOC order allows the user to buy or sell as soon
as the order is released into the system, as long as the price conditions are matched.
If the price conditions are not matched, the order is cancelled from the system. In case
a partial match is possible for the order, the remaining unmatched order is cancelled
immediately.

2.2 Price Conditions

• Market Price: Market orders have no price specified at the time order is released into
the system. For a Buy Market Order, the system matches the Sell Limit Orders in the
Order Book and for the Sell Market Order; system matches the Buy Limit Orders.

12
• Limit Price: An order to buy a specified quantity of a security at or below a specified
price, or an order to sell it at or above the specified price. This ensures that the
participant never pays a worse price than the limit price set.

• Stop loss: This order type allows the participant to release an order into the system
after the market price of the security reaches or crosses a threshold price.

2.3 Other Conditions

Disclosed Quantity (DQ) is an order with a disclosed quantity allows the market participant to
disclose only a percentage of the overall single order quantity on the exchange. For example,
an order of 1000 with a disclosed quantity condition of 20% will mean that 200 are displayed
to the market at a time.

Price-Time Priority

The principle of price/time priority refers to both orders and quotes. When an order is entered
into the order book, it is assigned a timestamp. This timestamp is used to prioritize orders
in the order book with the same price - the order entered earliest at a given price limit gets
executed first.

When a new order (or quote) is entered, Exchange first checks the limits of all orders contained
in the central order book. If the incoming order is immediately executable, meaning it is
capable of being matched against an existing order or orders; one or more transactions are
generated.

To be immediately executable, the order must be either of the following:

• A market order, where opposite limit orders already exist in the order book

• An order to buy at a price at or above the lowest ask in the order book

• An order to sell at a price at or below the highest bid in the order book.

Orders may not necessarily be executed at a single price, but may generate several partial
transactions at different prices. When a large order executes against the total available
quantity at a given price level, the next best price level becomes best. This process continues
as long as the incoming order remains executable. If not executed upon entry, an order is held
as a limit order in the central order book.

Also, it is possible for a single order to generate multiple executions at different points in time.
For example, an order may generate a partial execution upon entry, while the remaining open
order remains in the order book. The open portion may get executed a minute later, an hour
later, till the end of the current trading day or may not get executed at all.

13
Market orders have the highest priority for matching. Since the purpose of the market order
is to be executed as quickly as possible at the best possible price, they are entered without
execution restrictions. If several market orders are booked in the order book, Exchange takes
into account the timestamp of the orders to establish matching priority. The earliest market
order received by the exchange receives the highest priority.

In the case of limit orders, orders with the best possible prices (highest price limit for buy
orders, lowest price limit for sell orders) always take precedence in the matching process
over other orders with worse prices. Again, if the limit orders have the same price limit, the
criterion used for establishing matching priority is the order timestamp. This is the concept of
Price Time Priority.

2.4 Market Orders

Market Order is an order type that trades the given quantity at the best possible price at that
specific instant in the order book. The main risk of the market order is in the uncertainty of
the execution price, and on the other hand there is high certainty of execution. These orders
are not visible in the Order Book, but are executed as soon as they reach the system.

In order to understand the impact of Market Order, consider the Order Book -

Bid Size Bid Price Ask Price Ask Size

600 82.75 82.90 23
152 82.65 83.00 300
212 82.60 83.15 143
53 82.55 83.20 512
200 82.50 83.25 45

If a Sell Market Order of size 500 hits the order book at this instant, then, it will cross with
Limit Orders of Bid Price 82.75. The Sell Market order would report the price of execution as
82.75 and the order book would look as below.

Bid Size Bid Price Ask Price Ask Size

100 82.75 82.90 23
152 82.65 83.00 300
212 82.60 83.15 143
53 82.55 83.20 512
200 82.50 83.25 45

If the Sell Market Order of size 700 hits the previous order book (with the bid size of 600 at
82.75), then it will cross with Limit Order of Bid Price 82.75, size 600 and then hit the lower

14
price limit orders, till the size is met. In this case, the Best Bid of 82.75 is completely executed
and then 100 shares at Price 82.65 are also executed. The order book at this instant would
look like

Bid Size Bid Price Ask Price Ask Size

52 82.65 82.90 23
212 82.60 83.00 300
53 82.55 83.15 143
200 82.50 83.20 512
65 82.45 83.25 45

Thus the execution price received for this market order of size 700 is 82.7357 which is worse
than the Best Bid level of 82.75. Since the liquidity at the Best Bid was exhausted, the order
reached lower levels and thus the Sell Price was worse. This is termed as Slippage on account
of market conditions and lack of liquidity in Order Book.

There can be scenarios when there is not sufficient liquidity in entire order book.

Bid Size Bid Price Ask Price Ask Size

600 82.75 82.90 23
152 82.65 83.00 300
212 82.60 83.15 143
    83.20 512
    83.25 45

If at this instant a Sell market order for 1000 shares arrives, then it will take out the entire
liquidity from the bid side of the order book. The remaining 36 shares will remain standing as
limit orders at Last Traded Price of 82.60.

Bid Size Bid Price Ask Price Ask Size

    82.60 36
    82.90 23
    83.00 300
    83.15 143
    83.20 512

Market Orders can have market impact and therefore usage of large size in market orders
is not preferred. Such orders can be split into several smaller orders. The benefit of market
orders is immediacy and certainty of execution but it can result in considerable slippage
(worse than best available market price) at times.

15
2.5 Limit Orders

Limit Order is an order type that trades a given quantity at a specified price or better. These
orders provide liquidity in the Order Book. Limit orders will take as much liquidity as possible
within the price specified. If there is no liquidity available within the price specified, then the
limit order will stand in the order book till either its executed or cancelled or end of trading
day. If the limit order is partially executed, then the remaining quantity will remain in the
order book.

Consider the Order Book –

Bid Size Bid Price Ask Price Ask Size

600 82.75 82.90 23
152 82.65 83.00 300
212 82.60 83.15 143
53 82.55 83.20 512
200 82.50 83.25 45

If a Buy Limit Order of Price 82.75, size 100 is added to the Order Book then Order Book is
changed to the following

Bid Size Bid Price Ask Price Ask Size

700 82.75 82.9 23
152 82.65 83 300
212 82.6 83.15 143
53 82.55 83.2 512
200 82.5 83.25 45

If a Buy Limit Order of Price 82.80, size 100 is added to the Order Book, then it changes to
the following

Bid Size Bid Price Ask Price Ask Size

100 82.80 82.90 23
600 82.75 83.00 300
152 82.65 83.15 143
212 82.60 83.20 512
53 82.55 83.25 45

If a Buy Limit Order of Price 82.90, size 23 is added to the order book (in the initial order
book with 82.75 as the best bid price), then the limit order gets executed immediately. This is
known as Marketable Limit Order. The resultant order book looks as follows:

16
 Bid Size Bid Price Ask Price Ask Size
600 82.75 83.00 300
152 82.65 83.15 143
212 82.60 83.20 512
53 82.55 83.25 45
200 82.50 83.45 100

In case, the Buy Limit Order size is greater than the size of the Asks that have a lower price
level than Buy Price, then the Buy Limit Order will create a new Bid will created in the Order
book. For example, Buy Limit Order of Price 83.00, size 500 is added to the initialOrder Book
(with 23 shares at the best bid of 82.90), the Order Book looks like the following –

Bid Size Bid Price Ask Price Ask Size

177 83.00 83.15 143
600 82.75 83.20 512
152 82.65 83.25 45
212 82.60 83.45 100
53 82.55 83.50 200

The main advantage of Limit Order is that there is no uncertainty in execution price; however
the risk is in uncertainty in execution. If the price moves away from Limit Order Price, then
the order might not get executed.

2.6 Immediate-or-Cancel Orders

An Immediate or Cancel (IOC) order allows a participant to buy or sell a security as soon as
the order is released into the market, failing which the order will be removed from the market.
Partial match is possible for the order, and the unmatched portion of the order is cancelled
immediately.

Consider the Order Book –

Bid Size Bid Price Ask Price Ask Size

600 82.75 82.90 23
152 82.65 83.00 300
212 82.60 83.15 143
53 82.55 83.20 512
200 82.50 83.25 45

If a Buy IOC Order of Price 82.75, size 100 is added to the Order Book then it gets cancelled
and Order Book does not change.

17
If a Buy IOC Order of Price 83.00, size 100 is added to the Order Book then it gets filled with
23 shares at 82.90 and 77 shares at 83.00 and Order Book changes to -

Bid Size Bid Price Ask Price Ask Size

600 82.75 83.00 223
152 82.65 83.15 143
212 82.60 83.20 512
53 82.55 83.25 45
200 82.50 83.45 100

If a Buy IOC Order of Price 82.90, size 100 is added to the initial Order Book then it gets filled
with 23 shares at 82.90 and 77 shares get cancelled and Order Book changes to -

Bid Size Bid Price Ask Price Ask Size

600 82.75 83.00 300
152 82.65 83.15 143
212 82.60 83.20 512
53 82.55 83.25 45
200 82.50 83.45 100

Immediate or Cancel Orders provide an advantage of getting a price not worse than indicated
at the time of sending the order. Moreover, it gets filled or cancelled immediately, giving the
user an advantage of maintaining lower control over this order type.

2.7 Market with Price Protection

Market Price protection functionality gives an option to a Participant to limit the risk of a market
order, within a pre-set percentage of the Last Traded Price (LTP). It offers the immediate
execution that Market Orders offer along with protection of an inbuilt price limit. It ensures
that execution is not achieved at off-market prices. With a reasonable protection limit, it
ensures a balanced mix of immediacy in execution and certainty in price. All NSE market
orders are implemented as market with price protection orders. The pre-set market price
protection percentage is by default set to 5% of the LTP. The participants can change the pre-
set market price protection percentage.

2.8 Stop Orders

An order to buy or sell a security is when its price surpasses a particular point, thus ensuring
a greater probability of achieving a predetermined entry or exit price, limiting the participant’s
loss or locking in his or her profit. Once the price surpasses the predefined entry/exit point,
the stop order converts a market order. These orders are not visible in the order book.

18
Stop Loss orders are stored in Stop-Loss Book till the trigger price specified in the order is
reached or surpassed. When the trigger price is reached or surpassed, the order is released in
the Regular lot system. The stop loss condition is met under the following circumstances:

1. Sell order - A sell order in the Stop Loss book gets triggered when the last traded price
in the normal market reaches or falls below the trigger price of the order.

2. Buy order - A buy order in the Stop Loss book gets triggered when the last traded price
in the normal market reaches or exceeds the trigger price of the order.

The market order generated by a Stop Loss can lead to significant slippage especially when
the markets are volatile. If the order book is as follows and a Buy SL order has been placed of
size 500 with Trigger as 83.00 and a Buy Market Order of size 50 hits the Order Book.

Bid Size Bid Price Ask Price Ask Size

600 82.75 82.90 23
152 82.65 83.00 300
212 82.60 83.15 143
53 82.55 83.20 512
200 82.50 83.25 45

Once the Market order is processed, the Limit Order at 82.90 is traded and Limit Order at
83.00 is partially traded. However Last Traded Price matches the Trigger Price for the SL order.
The Order Book at that instant, when the Buy SL Order is triggered is as follows –

Bid Size Bid Price Ask Price Ask Size

600 82.75 83.00 277
152 82.65 83.15 143
212 82.60 83.20 512
53 82.55 83.25 45
200 82.50 83.45 100

At this moment, Buy SL order of size 500 is processed and it hits the order book. The Order
book after execution of SL order is as follows –

Bid Size Bid Price Ask Price Ask Size

600 82.75 83.20 432
152 82.65 83.25 45
212 82.60 83.45 100
53 82.55 83.50 200
200 82.50 83.75 100

19
2.9 Disclosed Orders

The disclosed quantity of the original order is indistinguishable from any other Limit Order in
the Order Book. Every time this disclosed quantity is completely executed, a new order with
price same as before and size equal to disclosed quantity is placed in the Order Book. The
hidden part of the original order loses the time priority as it joins the Order Book afresh.

In the following Order Book, a Disclosed Order of original Size 1000 is placed, however only
10% of the Quantity is to be disclosed. Thus it contributes a size of 100 at Price 83.00 in the
Order Book.

Bid Size Bid Price Ask Price Ask Size    

600 82.75 82.90 100  
152 82.65 83.00 100 83.00 100
212 82.60 83.15 143  
53 82.55 83.20 512  
200 82.50 83.25 45    

If a Buy Market Order of size 200 hits the Order Book, It will take away the Price Levels of
82.90 and 83.00. Since the Disclosed Order is now completely executed, another order of 100
shares will be displayed and now the Order Book would be as follows –

Bid Size Bid Price Ask Price Ask Size    

600 82.75 83.00 100 83.00 100
152 82.65 83.15 143    
212 82.60 83.20 512    
53 82.55 83.25 45    
200 82.50 83.45 100    

2.10 Execution Trading Strategies

Execution Trading Strategies aim to reduce market impact, improve performance versus
benchmarks, and streamline the trading process. The classification of the strategies is mainly
into the following categories –

• Schedule-Driven

• Opportunistic

• Evaluative

On the less “structured” side, are the opportunistic strategies, in the sense that these strategies
do not have pre-defined execution schedules; instead, they utilize real-time information

20
to actively search for optimal times when trades can be executed. These strategies create
execution schedules as they go along. At the beginning of an order, the execution schedule
is not known. The schedule driven strategies basically split a large order into smaller orders
and try to minimize the market impact costs while trying to reduce effect of execution on the
market price.

At the other extreme, the more “structured” end – are algorithms that follow precisely
defined execution schedules; are the schedule-driven strategies. All VWAP- and TWAP-based
strategies, for example, can be categorized this way.

Between these two ends, lie the evaluative strategies. These strategies combine approaches
of both opportunistic and schedule-driven algorithms.

Trading Benchmarks

In order to assess the performance of an execution, participants need benchmarks. These

benchmarks vary according to execution methodology, asset classes and affect both investment
and execution decisions. They are also reflective of intraday trading conditions. The often
used benchmarks are TWAP (Time Weighted Average Price) and VWAP (Volume Weighted
Average Price).

TWAP

TWAP benchmark is used in scenarios where the execution is determined by time availability.
It is the average of all observed trades over a given period of time. All trades are equally
weighted irrespective of their sizes. So, even small trades at extreme prices affect the TWAP
in a significant way. A large order might be broken up into smaller orders and spread out
evenly over a specified time period in order to lessen its market impact – that is, to avoid
an adverse effect on the price of the security. A TWAP benchmark is preferred over Volume
Weighted Average Price (VWAP) benchmark when the security is illiquid and where volume
analysis is not of significance.

VWAP

VWAP is the most used and transparent benchmark. It gives an indication of how price has
moved over a period of time, along with trade executions. VWAP is the average of all trades
weighted by their volumes. In contrast to TWAP, small trade sizes at extreme prices have
smaller effect on the benchmark and it is representative of the bigger trades.

2.10.1 TWAP Strategy

TWAP is the average of all observed trades over a given period of time. All trades are equally
weighted irrespective of their sizes. This strategy attempts to match this benchmark by
breaking down the order into equally spaced time periods.

21
For example, a trade of size 5000 is broken into smaller trades of 500, to be executed in 15
minutes intervals. The trading pattern is linear with time and the trade is scheduled to finish
in 10 intervals. The trade execution has no relation to the market volume.

2.10.2 VWAP Strategy

The VWAP strategy aims to meet the VWAP of the given symbol. The VWAP is simply the
summation of the price multiplied by the volumes at each trade divided by the summation of
the volumes of each trade.

Where,

Vn: Volume in the ‘n’th trade

P­­n: Price of the ‘n’th trade

However, in the case of the strategy, the key is to be able to predict the volume happening
on the day correctly. The most popular prediction methodology is using historical volume
profiles.

2.10.3 Percentage of Volume (POV) Strategy

In POV strategy, market participant specifies the percentage of volume (of the overall exchange
volume during that duration) he wants to participate in. The objective of the strategy is to slice
down a big order into multiple smaller orders. The size of the smaller orders is proportional to
the volume in the market as the target percentage is specified by the market participant.

The participant specifies the time duration during which he wants to participate on volume
and also the target percentage. If the target percentage is high then the execution happens
faster but can have a higher market impact.

22
Self Assessment

• A buy IOC order will not impact the order book if

– the order price is less than the best bid in the market

– the order price is equal to the best ask in the market

– the order price is greater than the best ask in the market

– the order price is greater than the best bid in the market

• TWAP is preferred over VWAP when

– the instrument is illiquid

– the market hours opening hours are very short

– the order book is very condensed for that instrument

– trading derivatives instead of stocks

• In VWAP strategy, the idea is to

– Trade an instrument in equal volumes throughout the day

– Trade different instruments in the ratio of their volumes traded in the market

– Trade an instrument in such a way so as to match the Volume Weighted

price of that stock in the market for a given time period

– Trade an instrument every time, when beyond a particular amount of volume

gets traded in the exchange

23
Chapter 3 Trading Strategies

Algorithmic trading openedlot of opportunities in form of various trading strategy. In the

following chapter, we will be discussing various trading strategies and their application through
algorithmic trading.

3.1 Calendar Spread

Definition:

A calendar spread is terminology used for trade where simultaneous Buy and Sell is being
carried out on futures/options with different expiries. In a more simple term, the strategy has
two legs, with one leg being Buy/Sell of a particular instrument of one expiry and the other
leg being Sell/Buy of same instrument of another expiry. The idea behind the trade is to profit
from the differential spread existing at different point of time over the course of expiry.

Calculation of Calendar Spread:

Suppose Stock “AAA” current month is quoting at Bid: 100 and Ask: 100.25, while far month
is quoting at bid: 101.10 and Ask: 101.35. Then, the calendar spread available in the market
is buying spread is: 1.35 and sell spread is 0.85.

Execution of Calendar Spread:

Algorithmic trading plays important role while execution of the calendar spread. Consider the
above example and suppose, a trader wants to take the spread available in the market and
can execute the spread available in the market. However, trader can also choose to quote in
one leg (in either expiry) and take the price in other expiry as soon as order is hit in the expiry
where trader is quoting.

Calendar Spread in options:

Calendar spread in option is more intricate. The strategy can be executed with calls or puts
also. It involves (very similar to calendar spread in futures) buying one option and selling
another option of the same type and strike, but with different expiration.

Options in near-month expiry typically have more time decay than far months. The strategy
profits from this difference in decay rates. It is best executed when implied volatility is low and
when there is implied volatility "skew" between the months used for trade.

The strategy is profitable in a limited range around the strike executed. The maximum profit
is realized when the underlying expires at the strike price used.

Uses of Calendar Spread:

Calendar spread helps trader to benefit from the time decay i.e. decrease in spread over the

24
time of two expiries. Moreover, the calendar spread provides trader a low risk opportunity to
play in the market as the strategy is protected against the extreme movement in the market
since trades are taken in the both direction (BUY-SELL) on the same underlying.

Calendar Spread for the options can help trader to take advantage of time decay. Option
traders often witness the time decay eating away the profits as options approach expiry. A
long calendar spreads provide a low-risk way to take benefit of time decay inbuilt in different
expiration dates. The strategy profits within a range. It profits from an increase in implied
volatility and are therefore a low-cost way of taking advantage of low implied volatility options.

3.2 Cash Future Arbitrage Strategy

When we deal with arbitrage strategy, it refers to process of clearing the inefficiency created
in related instruments in course of trading. In Cash Future Arbitrage strategy, a trader
expects to profit from the difference in the prices for cash and future for the same underlying
instruments. Trade is setup, usually during the early days of expiry cycle, by (typically, but not
necessarily) selling the future and buying the underlying stock. Near the close of expiry, the
trader would square off the position by reversing the trade i.e. buying the future and selling
the shares. The extension of the strategy could be doing arbitrage between Nifty futures and
the underlying constituent shares.

Definition

The trade entails selling the futures (that are quoting at a premium to the cash prices), and
buys same quantity of the underlying stocks. The arbitrageurs seek to make riskless profit.
The strategy becomes viable when the premium so desired to be earned is more than the
opportunity (carrying) cost and transaction costs. Since such opportunities are momentarily
available in the market, it becomes imperative to use algorithmic execution for the same to
avoid execution related errors.

3.3 Index Arbitrage

This strategy is extension of normal Cash-Future arbitrage. The trade has two parts, with one
part involving index future while other being basket of stock representing cash (stock) index
value.

Definition:

Strategy designed to profit from the mispricing of the Index future and the member constituents
in the index. The philosophy behind the strategy is as follows. Since both Index and the
member constituents are traded instruments, there exist temporal price discrepancy due to
market inefficiency in the derived index value and the actual traded index value.

Suppose Nifty Index trades at 5900 and Nifty Future trades at ‘5900 + X’. If cost of carry is

25
say ‘C’ and transaction cost ‘t’(includes brokerage cost and the liquidity cost), then

• If X > C + t implies then its profitable to setup Index arbitrage by selling Index
future

Once the Index Arbitrage is setup, trader can look forward to be in trade till expiry, as Index
Future price will converge to Index level.

Execution of Index Arbitrage

Accurate execution forms the backbone for this strategy. Once the arbitrage conditions are
satisfied, the algorithm is supposed to execute both side of the order of arbitrage i.e. Index
on one side and Equity Basket of the index constituents on other side. The execution needs to
be accurate and precise. Traders can choose to send equity basket in one go or can choose to
quote on the illiquid instrument first.

Once the Arbitrage is setup, the arbitrageur has an option of either square off trade before
expiry (If the premium decreases and with cost of carry being covered), or rollover* future to
next month. If neither of the process is followed, trader needs to offload basket of stock once
the future contract expires on expiry day.

*Rollover is the process through which current position is shifted to next month. Suppose
a trader is short in current month and wants to rollover. Trade would be selling the next
month contract and buying the current month. Similarly, long position can be moved into next
month.

3.4 Pair Trading

Pair Trading has been one of the most popular algorithmic trading strategies across global
markets. Being long and short simultaneously, the strategy provide trader to harness the
relative movement between stocks without taking high unwarranted market risk.

Definition:

Pairs Trading, also known as Statistical Arbitrage Trading , is defined as trading one stock
(or basket of stock) against another stock (or basket of stocks) in such a fashion that long
position is taken on one Leg and Short on other. Strategy is also termed as convergence
trading strategy/ “contrarian strategy”designed to harness mean-reverting behaviour of price
ratio of the stocks.

Normally, strategy is executed either Rupee Neutral Basis or Market Neutral Basis. Rupee
Neutral Scheme entails where the value of the both the executed Leg are similar, while in
Market Neutral Basis, stocks are executed in proportion to their relative betas and it is more
suitable to protect the trade from the market movement-up or down.

26
Compared to arbitrage strategies, the pair trading is not a risk-free strategy. Key risks involved
in trading are as follows:

• Contrary to expectations, the prices of the two securities begin to diverge (drift apart),
i.e. the spread picks up trend rather than mean-reverting to the original mean.

• A surprise event in a security can trigger extreme movement in pair ratio.

Strict risk management of Stop Loss is required to handle adverse situations once the mean-
reverting behaviour is invalidated.

Methodology of Pair Trading:

Methodology used to determine various trading scenarios can be generated as follows. Trader
calculates Pair Ratio as:

Pair ratio = Price of Stock A / Price of Stock B

The trader looks forward to Sell Pair ratio (i.e. Sell Stock A and Buy Stock B) or Buy Pair ratio
(i.e. Buy Stock A and Sell Stock B) when its seems to be profitable to trade, once the spread
has diverged to extreme values from the mean value.

Fig 1: In the above graph, Blue line depicts Pair Ratio, Green Line Depicts typical upper
bound, Purple line as typical lower Bound and Red line as average. It is to be noted that upper
bound and lower bound are indicative parameters for extreme pair behaviour and no-way
guarantees that the pair ratio cannot exceed beyond these values.

27
A typical pair trading looks as shown in graph. Trader can choose to execute Sell Pair, once
the upper bound is reached / breached and Buy Pair, the lower bound is reached /breached.

3.5 News Based Trading Strategies

Traditionally, news based trading has been one of the most common form of trading followed
by the traders. Traders of this genre try to gain an understanding of the market in terms of
the reaction one can expect with the arrival of new information. These strategies, where the
trader expects to harness the temporary mispricing occurring in the market due to sudden
news, are also termed as Event Driven.

As communication becomes faster, traders utilizing such faster information could respond to
events at a very quick pace. Such events could be corporate quarterly results, new merger &
acquisition or economic figures released by governmental or statistical agencies.

Concept of Algorithmic News Based Trading

Various data providers provide machine readable digital news. The system is designed to
make sense out of the news. A trader can input the expectation before the arrival of important
financial result or policy decision by central bankers, etc and compare with actual results and
take the informed decision based on it. The success in a news based trading system depends
on the pre-defined trading rules chosen and the heuristic model applied in arriving at these
rules.

There are significant road-blocks in such system. It becomes increasingly difficult for the
system to decipher and estimate the impact of the news, if the event is highly unexpected.
Moreover, system need to be adept at pricing the news and arrive at the right price since just
a limit order cannot be put which will be hit. Unavailability of large enough historical data,
especially in digital format makes quantifying the news a daunting task while trying to create
a news based trading system. The strategies assume a specific range of parameters and
keywords. However, if data lies outside the pre-defined range, it can lead to erroneous signal
and thus erroneous trade. Hence, it calls for strong risk management in the event the news
data being outside the known historical range.

3.6 Conversion- Reversal

Similar to cash future arbitrage, the conversion and reverse conversion helps in keeping
the efficiency in the options markets. In ideal scenario, the returns from the conversion
–reversal strategy should be not more than the risk free interest rate prevailing in the market.
However, momentary mispricing greater than risk-free interest rate can occur. The trader
looks forward to harness this mispricing occurring between the Put-Call-Future prices. When
the trader executes this opportunity, the mispricing ceases to exist, leaving no room for delay

28
in execution. Floor Traders and the market makers have utilized this strategy profitably for a
long time since they operated with lower cost of capital and transaction cost.

Definition

Conversion trading strategy involves Buying Future, Selling Call and Buying same-strike Put
option (Expiry date for the options are same). Since the long created by the future buying is
offset by the synthetic short created by the Selling of Call and Buying of Put, the trade creates
the directional neutral trade. The difference in long future and short synthetic future is the
profit trader looks forward to make.

The reverse conversion (or simply referred to as reversal) is exactly opposite to that of
conversion. A future short is setup against the synthetic future long i.e., trader takes Short
future, Long Call and Short Put (options being of the same-expiry and same-month) position
to setup this arbitrage.

Calculation for the conversion Strategy

To have better grasp of the strategy, let us consider the following information for the stock
“AAA”. Assume that both cash as well as future price for AAA be quoting, bid at 99.0 & ask at
99.1; August month Call option for strike 100 trading at 3.8 (bid) and 3.9(ask); while August
month Put option trading at 4.35 (bid) and 4.40(ask). A trader looking to set up the trade
would be Buying future at 99.1; Selling Call option at 3.8 and Buying Put option at 4.40. To
evaluate the following trade, consider the table as follows:

Option Option Option Time

Option Expiry Intrinsic Value*
Trade Strike Price Value

August Short Call 100 3.80 (bid) 0 3.8

August Long Put 100 4.40 (ask) 0.9 3.5

     
Conversion Profit Diff of Time value 0.3

*Intrinsic Value has been calculated based on the Buying Price of 99.1

Profitability is being calculated based on the difference in the time value. So, whatever the
price of the Future at expiry, the trader stands to make Rs 0.3 out of this transaction

The conversion strategy has been executed for At-the-money options, the resulting delta would
for the short call option is approximately -0.5 and for the long Put option will be approximately
-0.5. The net delta resulting from the combination will be nearly close to -1, which is offset by
the +1 delta resulting from the long future trade. Thus, the strategy provides the trader with
the direction neutrality.

Now, let us analyze at the profitability of the above trade at time of expiry

29
 P&L at Different Expiry Price
Instrument &Trade
105 100 95

Short Call, 100 Strike, sold at 3.8 -1.2 3.8 3.8

Long Put, 100 strike, bought at 4.40 -4.4 -4.4 0.6

Future Buying at 99.1 5.9 0.9 -4.1

Total Profit 0.3 0.3 0.3

Fig 2

Thus, it is noted that, in the above mentioned strategy, trader stands to gain 0.3 at expiry
irrespective of expiry price.

On the similar lines, we can create the reversion trade and calculate the profitability at different
scenarios of expiry.

30
Self Assessment:

• If you want to create a rupee neutral strategy where the 2 legs involved are USDINR
currency future[lot size 1000] and NIFTY future [lot size =50] for the same month. If
the trade level for Nifty is 6000 and for USDINR is 60. How many lots of USD INR would
you trade against 1 lot of Nifty.

– 10

– 3

– 0

– 5

• Pair Trade can never be beta neutral

– True

– False

• A trader trades a calender spread in currency futures on NSE such that [s]he is long
November [expiry on 27th November, 2013] and short December [expiry on 29th
December, 2013]. At 12:15 pm (when contract expires) on 27th November, 2013.

– Trader has zero position.

– Trader long December short January spread.

– Trader has a short December position.

– Trade has a long January Position.

31
Chapter 4 Algorithmic Trading: System
Architecture

Any conventional trading system would consist of the following modules in one form or
another:

1) Market data adapter

 Reads market data packets from the network.

 Manages the market data network session if required

 Parses the data packet and passes it on to the rest of the system if required

2) Historical data store

 A store of all data that is being received

 A store of all decisions taken on that data

3) Analytics module

 This module is used for analysis over data stored in the historical data store.

 Might involve some complex mathematics and data mining.

4) Application/Graphical Interface (GUI) for trader inputs

 Application available to trader for operating

 Analyze current data with visual patterns found in the data store.

 An interface where the decisions/trading signals can be generated by the

trader

 Also provides an interface for the inputs given by trader.

5) Order Manager

 Maintains the state of all orders.

 Routes the orders to the destination.

32
 Fig1 – Structure of a conventional trading system.

With the advent of DMA and automated trading, following changes in architecture took
place:

 Order management became more robust and order management logic moved
from the application to the server.

 Risk management done in automated fashion in real time by the Order

Management System (OMS) before the order is generated. Global position
monitoring introduced as a second level of monitoring.

 Automation of analysis of historical data and decision making led to the advent
of Complex Event Processing (CEP) engines. These CEP engines have their own
storage requirements for event history to identify future opportunities.

 Low latency systems for generating orders and handling market data were
introduced. To generate orders quickly, market events are now handled in the
server instead of the application.

 With more processing power available, more complex mathematical operations

are included and these are usually handled in dedicated calculation blocks (e.g.,
options Greek calculations).

 Risk Management checks introduced in application for checking trader inputs as

well.

 Application became merely a view for the trader and a medium for inputs and
monitoring positions/orders.

33
  Standardized protocols for communication evolved in order for the server to
communicate with the exchange.

 Data normalization introduced to convert market data from the exchange format
to platform specific format, since server now took on several additional tasks
and the information needed to be propagated between these blocks.

 Additional regulatory storage requirements introduced requiring storage of

trade information in addition to market data.

 An additional block introduced into the architecture to listen to analytics and

data from third party vendors as well.

 Simulator block introduced into the architecture to test strategies. This block
replays market data as well as acts as a destination for test orders.

Fig 2 – Architecture of Automated Trading System

4.1 Market Data

Events in the market are communicated to the trading server over the network in the form
of market data. This data forms the main input to the algorithmic trading system. Following
flow illustrates how the market data flows and reaches the strategy.

34
 Fig 3 – Market Data processing

Market data can be received in any of the following three ways depending on the connectivity
and segment.

MULTICAST SNAPSHOT Tick By tick (TBT) DATA BUCKET TBT DATA

DATA
Is not generated on every Is generated for every scrip in Is generated for every
event in the exchange. the segment. Much more data subscribed scrip in the
periodical snapshot flowing into the trading system. session.
Multicast data, hence no TCP data. Guarantees sequencing TCP data. Guarantees
sequencing guaranteed. of data. sequencing of data.

Order book depth is fixed by Order book constructed at Order book constructed
the exchange. the server end. Hence depth at the server end. Hence
determined by the trading depth determined by the
server. trading server.
UDP server needs to be No need of a UDP server. Trading No need of a UDP server.
running to receive UDP data server directly initiates TBT Trading server directly
connection. initiates TBT connection.
Once the UDP server sends The connection is between the The connection is
the multicast packet over trading server and the exchange between the trading
the local network, every only. No other server can listen in server and the exchange
machine on that network on this data. only. No other server can
would receive it. listen in on this data.

35
 MULTICAST SNAPSHOT Tick By tick (TBT) DATA BUCKET TBT DATA
DATA
No login required Authenticated with login and Authenticated with login
password. Login to online and and password. Login to
offline server required. both online and offline
1) login to online and offline server required.
server using username and 1) login to online and
password offline server using
2) each packet received with username, password
a sequence number for the and list of symbols
packet 2) each packet received
3) if sequence number is not with a sequence
incremental, then connect number and session
to offline server and request sequence number
for missing sequence number for the packet
range 3) if session sequence
4) once synchronization done, number not
start reading from online incremental, then
server connect to offline
server and request
for missing sequence
number range
4) once synchronization
done, start reading
from online server
Available for all segments. Available for all segments. Available for FO segment
only.

Recently, NSE has also introduced Tick by Tick (TBT) data over multicast for connections
within the NSE colocation facility. Multicast TBT is on UDP protocol and does not require the
member to login to either online or offline server.

4.2 CEP Engine

The CEP engine is the core of the architecture of an algorithmic trading system. It listens to
market data and determines what actions to take based on the settings and inputs given by
a trader.

The input to this engine is an event. This could be

 Market data, news

 Order acknowledgement/fills/rejects

 Exchange published information

 Change of settings/inputs by the trader.

36
The output after the processing is

 Order request to the order manager.

 Notification to the trader and the application.

Within the CEP engine, the flow of tasks is as follows

 Decode the event and store it in the event history.

 Check if the event falls in known patterns.

 Run the decision processing rules on this event. This would determine what to
do next in case a pattern is recognized.

 If an action needs to be carried out, this is communicated to the action

destination outside the CEP engine.

 This action is also stored in the action-store.

 The action is also conveyed to the trader and the application.

 Remember that the processing rules themselves could be generated on the fly
based on the event store and action store

 In case some complex mathematical calculations are involved, dedicated

resources like CPU and memory are allocated to the CEP.

Fig 4 – Flow of events in CEP engine

4.3 Order Routing / Order Manager

The action notifications from the CEP engine are received by the order manager (OM). The

37
order manager receives such notifications typically from multiple CEP engines. This module
is responsible for generating and managing orders to external destinations as well as for Risk
management checks.

A single order manager typically manages orders for multiple segments. Since all orders
would flow through this module, RMS checking is implemented at this stage before the order
is generated. Depending on the segment and the settings it could use the FIX API or native
API to send and receive orders.

In addition to above, the OM also maintains the state of the orders and manages the
invitations.

Tap Server and Invitations

In the case of Non Neat systems, all interactive communication to the exchange is handled
through a TAP server, which acts like an intermediary. Upon connecting to a TAP server, an
invitation message is received with one or more invitations. For every invitation received, the
OM can send one message to the exchange.

Since the OM handles multiple CEP engines, it also manages the contention for an invitation.

For example, if CEP1 and CEP2 send new order signals to the OM which has only one invitation,
the OM would only be able to forward one of the order signals while keeping the other on wait
until a fresh invitation is sent by the TAP server.

Fig 5 Flow of events in OM

38
4.4 New Trends – MultiTap

As mentioned in section 4.3 for every invitation received from the TAP server, the OM can
send one message to the exchange. This could result in a bottleneck situation where the
OM runs out of invitations. To circumvent this problem, NSE now provides a MultiTap server.
Since every TAP server is associated with a network ip (that is provided by the exchange),
multitap requires all the necessary ips to be configured on the same ethernet device using
virtual ips. The Multitap server can then be configured to behave like multiple tap server
processes (which earlier had to be run on different servers). This would mean the OM can
now form multiple connections (one for each CTCL id on each of the TAP server processes),
thus increasing the number of invitations and hence reducing the contention for invitations.
Another added advantage of the multitap server is that the message rate (which is particular
to the scenario being used to connect to the exchange) would also increase. The new message
rate is effectively the sum of individual message rates across all the individual TAP ips.

4.5 Colocation

As discussed in the earlier sections, the advent of automated trading systems and high
frequency trading systems led to a need for low latency trading architectures. One of the
ways to reduce latency in a very big way is to move physically closer to the exchange. That
means the signal generated by the OM for an order reaches the exchange faster. To cater to
this requirement, co-location facilities (rack space for the trading server within the exchange)
are offered.

The main advantages are

 Market data arrives at the server faster.

 Once the CEP engine communicates the decision to the OM, the message/order
reaches the exchange faster.

 This also means the acknowledgements/rejects reach the server faster. Hence
the next action can be taken much earlier inside the colocation than outside.

 In the event of a network error on a TBT or bucket TBT connection, sync up with
the offline server is much faster upon reconnection from inside the colocation.

 TBT connectivity + interactive connectivity to different segments are available

on the same physical line as against having a separate line for TBT data of each
segment outside.

4.6 Trading Setup

Typically trading setups can be classified into two broad types: Colocation setups and non-

39
colocation setups. In either case, the application typically runs in remote location (on the
trader’s desktop). However, as discussed above, the category in which the trading setup
would fall would depend on where the heart of the system i.e., the CEP engine and the OM
(collectively referred to as the Trading server) are running. For colocation setups, the trading
server is in colocation and similarly, for non-colocation setups, it is typically located in a data
centre.

Fig. 6: Architecture of typical trading setup inside exchange co-location

40
 Fig. 7: Architecture of typical trading setup in data centre

4.7 Smart Order Routing (SOR)

The Securities and Exchange Board of India (SEBI) has approved SOR facility since August
2010. Using this facility, the broker trading servers (CEPs) could choose execution destination
based on factors like price, etc that the broker seems fit.

The advent of SOR has led to the following changes

 The OM router now handles multiple locations.

 The market data from different destinations now is normalized into one standard
format for the use of CEP.

 Communication via standard protocols like FIX gaining popularity unless native
API offers significant latency advantages.

41
Following diagram showcases these changes

Fig 8: OM connects to multiple destinations.

4.8 Connectivity Options

There are two types of communication options available for trading members.

 VSAT (Very Small aperture terminal) – satellite based wireless communication

 Leased Line (wired terrestrial based communication)

The trading member needs to choose a scenario based on categories mentioned below

A category B category C category T category

Message rate 40 messages/ 100 messages/ 200 messages/ N.A
sec sec sec
Bandwidth 2Mbps 2Mbps 2Mbps 2Mbps
Market Data Available but N.A N,A Available but
multicast only Tick by tick
connection only
User ids 20 shared 50 shared 100 ids shared N.A
between CM between CM between all
and FO and 10 and FO and 10 segments
in CD. in CD.

42
 A category B category C category T category
Scenarios A1 – single B1 – Single C1 – Single T1- single leased
VSAT leased line Leased Line line

A2 – Single B2 – Dual leased C2- Dual leased T2 – Dual leased

leased line line to one or lines to one or lines to one or
two POPs two POPs. two POPs.
A3 – Single
leased line with
VSAT as backup

A4 – Dual leased
lines to one or
two POPs

A5 – Dual leased
lines and VSAT
as backup

Currently Indian exchanges enable Algorithmic Trading only on Leased Lines. Algorithmic
trading is not allowed on VSAT or Internet Based Trading (IBT)

In addition to the above scenarios:

In the case of NEAT terminal – Every approved user is assigned to a branch id. For every 5
NEAT users under a branch id, there is a requirement for on valid NCFM certification.

In the case of Non-NEAT CTCL terminal – each location is assigned a branch ID which is
contained in a 12 digit CTCL ID (apart from the three digits that explains the type of trading/
order). For every 5 CTCL ids in a branch id, 1 user should be NCFM certified.

43
Self Assessment

• Market data pattern detection and decision making happens in the

– CEP block

– OM block

– Application block

– Algorithmic trading platform

• Risk management is handled in the

– Order Manager and the application block both

– Order Manager

– Complex Event Processing block

– Complex Event Processing block and Order Manager

• The function of a simulator is to

– Simulate the market data emanating from an exchange

– Simulate an exchange where orders could be sent (instead of sending

to actual exchange)

– Simulate the performance of a strategy

– Simulate the computational complexity of the algorithm before actual execution

in real markets

44
Chapter 5 Risk management in Algorithmic
Trading

Risk management in general is the process of managing the uncertainty associated with
business operations – especially the undesirable ones. Since trading involves dealing with
a lot of uncertainties related to the behaviour of the markets and market participants, risk
management is a very critical task and department within trading operations. All instances
of failures within trading firms are primarily linked to poor execution of risk management
functions – mostly through under-estimation of extent of undesirable outcome; or the failure
to identify certain factors which might lead to undesirable outcomes; or improper adherence
towards risk management processes.

Operationally, automated trading is more complicated than non-automated trading –

technological complications get added to trading related complexities. To ensure the proper
functioning of the automated trading system, a lot of different technological components
must interact with each other in real time – often this becomes a complex web of complicated
linkages between different components of the automated trading system.

Since, from an operational perspective, automated trading is much more complicated as

compared to non-automated trading, the uncertainties involved with the proper execution of
automated trading are also higher.

In the sections that follow, we will look at risk management for trading operations at a
holistic level. We will then look at risk management processes which are specific to automated
trading.

5.1 Different Stages involved in Risk Management

The following are the different stages of the risk management process

1) Setting risk management structures & policies

2) Identifying sources of risk

3) Evaluating different risk components

4) Establishing risk appetite & setting risk limits

5) Designing systems with strict adherence to risk controls

Stage I: Setting risk management structures & policies

This is the first stage within risk management. This in turn involves the following activities:

i. Setting up a dedicated team of professional for the risk management department

45
ii. Separating the trading department from the risk management department – i.e. there
should be no conflict of interest. The incentives of the risk management department
should not be based on the trading profits as this will incentivize the risk management
department to allow more trading risks.

iii. Providing full autonomy and empowerment to the risk management department
to oversee trading activities. Given that even minor breaches of risk policies have
the potential to cause huge losses in a very short duration itself, therefore the risk
management team should be provided ample and direct empowerment to raise the red
flag and stop trading activities as and when they deem necessary

iv. There should be a proper process oriented system for introducing new products,
trading strategies and operations. Such a process oriented system will ensure that all
the checks and balances are performed and known pitfalls addressed while embarking
on anything new

Stage II: Identifying sources of risk

After creating the systems in place, the risk management department should try to create an
exhaustive list of all risks against which they should protect the firm. The risks can usually be
classified into the following categories:

i. Market Risks

(Risks arising from change in prices of securities in which the firm holds some
position)

ii. Credit / Counter-party Risks

(Risks arising from failure of counter-party to meet obligations and commitments)

iii. Financing Risks

(Risks associated with the necessity to finance monetary requirements for normal
operations)

iv. Regulatory Risks

(Risks associated with change in rules and regulations)

v. Liquidity Risks (Exogenous & endogenous)

(Risk of loss because of inability to perform a trade owing to absence of enough

liquidity in the exchange at favourable levels. Moreover, additional losses also happen
due to the necessity to pay the bid-ask spread while doing a trade. Furthermore, while
trading huge quantities in a particular direction, the trade itself causes the market
prices to move against the direction of the trade)

46
vi. Operational Risks - Systems, Mechanical, Criminal, Natural disaster, Terrorism, etc

(Risks associated with normal day to day operations – like proper working of the
technology systems, physical security, political risks, etc. It can also cover risks like
fraud, legal risks, etc)

Stage III: Evaluating risk components

The next stage involves designing methodologies to quantify current statuses so that the
current situation can be monitored with respect to risk concerns.

i. Market Risks :

Market Risk can be evaluated using either sensitivity analysis or by using VaR analysis
& Stress tests. In sensitivity analysis, the total exposure of the firm to different market
risk conditions (like vega exposure for change in volatility, exposure, etc) are evaluated.
For different scenario conditions, the profit/loss of the trading firm is then calculated.

In case of VaR analysis, the amount of loss with a threshold probability is set as a
threshold.

In case of Stress test, various extreme case scenarios are analyzed and the situation
of the firm checked in those extreme case situations.

1. Sensitivity Analysis

a. Total Greeks, Dividend, Currency, Commodity, Interest Rate , Sector-

wise, Geography-wise exposures

b. What-if scenario analysis

2. VaR analysis

3. Stress tests

ii. Credit / Counter-party Risks :

Most firms use a score-card approach to ensure that exposure to different counter-
parties is within a limit. These score-cards could use the ratings of independent third-
party rating agencies to determine limits to be assigned for each counter party

A common methodology is to use the IRB method proposed under Basel II which is
built on top of the Jarrow Turnbull model.

iii. Financing Risk

The risk has to be assessed on two fronts: (i) rho position of the firm (ii) exposure to
shift in yield curve.

47
iv. Regulatory risk

This can be difficult to quantify and evaluate. A best attempt can be done to anticipate
future change in regulations on the basis of news analysis and historical precedent.

Some of the common regulation based risks involve:

• Short Selling Ban

• Margin Increase

• New Taxes / Change in tax rates

• Ban on trading in certain class of securities

v. Liquidity Risks

The risks associated with loss of liquidity is measured using Liquidity adjusted VaR

L-VaR = VaR + Exogenous Liquidity Cost (worst expected divergence of bid-ask spread
from fair value)

vi. Operational Risks (Systems, Mechanical, Criminal, Natural, Terrorism, etc)

The current status of the system has to be checked to ensure that the system is not
operating under stretched conditions. This involves checking the following the system
under the following categories:

Systems:

1. Robustness of the entire system – the ability of the trading systems and
processes to handle stress situations

2. System Load handling capacity – the current load of the trading systems and
maximum possible load on the systems and processes

3. Maximum order flow before system detects failure – the minimum time required
to detect failures should be tested from time to time

Criminal:

1. Accessibility and security of key resources – the number of access points,

number of people with access controls, etc

2. Fraudulent activities prevention checks

Natural Disaster and Terrorism:

1. This would involve verifying vulnerability of operations to such external risks.

These can be quantified using a scorecard approach (similar to the approach
used by insurance and actuarial firms). Moreover disaster recovery also involves

48
 creating a backup of current operations in another location. This risk metric
should be checked from time to time on the basis of readiness to shift to
recovery location.

Stage IV: Setting risk limits

After having devised methodologies to evaluate the current status of the operations with
respect to risk metrics, the next stage involves setting controls which will protect the operations
against negative outcomes

i. Market Risks

Having defined ways to measure exposure to market risks on different metrics, limits
should be set on trading positions to ensure that the following exposures are within
limits:

• Total cash exposure

• Exposure to geography

• Exposure to sector

• Exposure to asset class

• Exposure to assignment / delivery risks (settlement risks)

• Settlement Type (future vs. cash)

• Exposure to interest rates

• Exposure to exchange rates

ii. Credit / Counter-party Risks

Setting total possible credit limit per counter party , per counter party category (say
all counter parties from a particular geography, or all counter parties with similar credit
rating) and overall credit at risk

iii. Financing Risks

Total rho exposure of the firm should be fixed at an upper limit

Moreover exposure per segment of the yield curve should also be fixed at an upper
limit.

iv. Regulatory Risks

Stock borrowing should be limited to a fixed value to offset risks of ban on short-
selling

Likewise, for rules increasing margin requirements, etc – limits on position sizes should
be set

49
v. Liquidity Risks (Exogenous & endogenous)

Liquidity adjusted VaR for the firm should not fall outside a predefined limit

vi. Operational Risks - Systems, Mechanical, Criminal, Natural, Terrorism, etc

The following limits should be set

• Systems:

 Max exposure per strategy

 Max orders in a fixed time frame (per second in case of automated

trading)

 Max orders in a day

 Max exposure per application

 P&L fluctuation per application

 Price Range check

 Max order size

 Max Value Traded

 Net Value of portfolio

• Criminal/Fraud/ Theft, etc:

 Access Control

 Transparency of operations

 Rotation of team members

 Audit (internal & external)

 Centralized P&L reconciliation

 Independent verification of price to pricing models

 Online Infiltration & Virus Protection

Stage V: Designing systems with strict adherence to risk controls

The entire risk management is still incomplete after determining all the sources of risk, defining
the ways to calculate these risks and then setting limits on those risk exposures. The entire
process will only be completed after the establishment of a system which provides the ability
to monitor the operations in real-time and control all operations from a centralized system.

50
Therefore the following will need to be established:

• Centralized system which monitors all trading activities in real-time

 Summarizes net positions exposures to different risk categories (different

market risks, credit risk, financing risk, etc)

 Displays overall system load & performance

• Centralized control of all trading operation

• Pre trade controls

5.2 Risk Management Specific to High Frequency & Algorithmic

Trading

Risk management for automated and high frequency trading is a more critical and complicated
process. The following characteristics of automated trading make risk management even
more vital as well as complex:

i) Orders flow out of the trading system on their own on the basis of pre-defined triggers
and parameters – ‘without human control’

ii) Because a tremendous number of orders can flow out of the trading system in very rapid
time in case of an error, therefore trading portfolio positions could reach dangerous
levels in no time – even before a human being can realize (and respond), tremendous
damage would have already been done

iii) Higher reliance on technology for this method of trading implies increased sys-ops
risk. An automated trading system is composed of a number of different components
linked together – any single link between two components not working in perfect order
will wreck havoc

iv) Traders now have to keep technological complexities in mind while trading the financial
markets – a proper understanding of the algorithmic black box is necessary for traders
and therefore exposure to technology for financial professionals involved in trading is
required.

To give an indication into the new types of errors which can happen only with automated
trading systems, imagine the situation when the market data price feed from the exchange
goes down – the algorithm will calculate and keep sending orders based on stale data. The
reasons why the market data price feed is not working could again be traced back to a variety
of possibilities – physical disconnection, software disconnection from the exchange, software
crashes, etc.

51
• Risks specific to automated trading can be classified into the following categories:

 Access

These are related to ensuring that the current system connectivity in place and
working properly

 Consistency

This involves ensuring that the processes are working in sync and in real
time (i.e. ensuring that the market data and the calculations are not delayed
significantly)

 Quality

This involves ensuring that the quality of the data is proper and the trading
system is not operating on stale and garbage data.

 Algorithm

Here we need to validate the robustness as well as the correctness of the

trading algorithm

 Technology

The network, operating system and hardware involved with the automated
trading system should be monitored and checked

 Scalability

Since scaling an automated trading system is very easy - a strategy tested on

one underlying only has to be executed for multiple underlying. However, the
system might not work well when scaled up.

• Risks specific to automated trading are handled pre-order (i.e. before the order has
flowed out of the trading system) and not post-trade. There are two main areas where
these risks are checked

 Within the trading application

 In the Order Management System (before generating the order)

RISK Methodology

Access Connectivity to an exchange Sending heart-beats to check if the

goes down connectivity is up
Exchange disconnects you Sending heart-beats to check if the
connectivity is up

52
 RISK Methodology

Network issue Hardware, Operating System , ping to the

router, ping time to the router
Consistency Market Data is stale Validating Time-stamp of market data with
current time
Trading strategies are Time-stamp of analytics compared to
running in real-time (huge current time
processing time might result
in strategy responding
much later after the event
in the market)
Order Manager adaptor is Time-stamp
responding in real time
Quality Market - data is garbled Market Data adaptor should typically
handle it. Moreover the trading application
should reject such data
Loss of liquidity during The trading application should stop the
high-volatility periods strategy when the market liquidity has
– during such periods dried below a particular level.
the market data used to
calculate strategy values as
well as strategy orders will
be grossly miscalculated
because the underlying
prices itself are not properly
available
Algorithmic Margin breached The Order Manager should not allow the
system to send orders which increase
the margin usage beyond a particular
threshold
Exposure limit set by Exposure limit should be set in the
exchange system
Risk limits exceeded Check for acknowledgements before
sending order

53
 RISK Methodology

Incorrect strategy setting P&L fluctuation check – if the P&L of the

leading to continual bad strategy drops a lot, then that could mean
trades that a lot of bad trades are happening
-do- Order release (throttle) rate – if a strategy
setting is wrong, then the strategy will
end up sending orders at wrong levels
which will then be traded against by
other market participants. Therefore one
metric is to measure the number of orders
flowing out of the system – in case this is
too high, then it could be the result of a
wrong strategy
-do- Fat finger settings check – the trading
application should not allow the trader to
input values which are way different from
expected range of values
-do- Max Value Traded. Wrong parameters will
lead to lots of wrong trades – this metric
will ensure that the number of wrong
trades is stopped after a traded value
reached a limit
Incorrect order generation Price range check – orders can only flow
out of the system if the order prices do not
vary too much from the market levels
Order throttle Exchange reject limit – the number of
orders that can be sent per second.
Technology Hard disk space Independent check

Virus /Trojan Firewall, Anti-virus

System Crash Operating System
Application crash Heart-beat to check application

Protocol Mismatch Third-party software compatibility check

54
Mandatory checks required by the exchanges and regulators for automated trading
systems

Basic RMS description:

Manual trading disabled: Manual orders are disabled for auto-trading systems

Trade Price Protection Limit: Algorithmic orders should not be released in breach of the
bad trade price as defined by the Exchange for the security in respective segments

Quantity Freeze Limit: For each instrument an order quantity size is defined by the exchange.
Algo order should not breach this limit.

Price Range Check: Order should not breach the circuit limit (daily price range) of an
instrument

FII restricted list: FIIs cannot trade in a select set of stocks (RBI directed)

Market Wide Protection Limit: Cannot trade derivatives to increase Open Interest beyond
a pre-defined threshold

Shares available for selling: Overnight long position that is available per share for selling

Automated Trading enabled: Automated trading to be enabled for a select list of instruments
only

Index change check: The system cannot send Buy/Sell orders if the index has gone up/
down beyond a certain percentage point.

Client Position Limit: Algo orders should not be released in breach of position limit as
defined by the trading member for the client.

Margin Limit: If a threshold of the available margin is reached, then the application should
not send orders to increase the position further

Exposure Limit Check: Orders should not be released in breach of exposure limit as defined
by the trading member for the client.

Order Value: Exchange prescribes Maximum Value any single order can have. System should
incorporate check so that any given single order should not exceed that value.

55
Self Assessment:

• What kind of risk is to be managed for Portfolio Gamma exposure (i.e. Change in
Delta)?

– Credit Risk

– Market Risk

– Financing Risk

– Regulatory Risk

• Which of the following is not a type of risk associated specifically with algorithmic
trading?

– Access

– Consistency

– Novelty

– Quality

• Which of these is not a mandatory check required by the exchange for running
automated trading systems

– Trade Price Protection Limit

– Price Range Check

– Market Wide Protection Limit

– Value at Risk check

• Loss of liquidity during periods of high volatility would be classified under which category
of algorithmic trading specific risk?

– Access

– Quality

– Consistency

– Market

56
Chapter 6 Audit and Compliance process

The advent of algorithm had posed new challenges for compliances, which in turns paved path
for the introduction of new regulation to prevent participants to indulge in malpractices. SEBI
has advised exchanges to conduct audit of member-brokers periodically to ascertain proper
conduct on their part.

6.1 International Organization of Securities Commissions

International Organization of Securities Commissions (IOSCO) was formed in 1983. It was

the expansion of inter-American regional association that was created in 1974. IOSCO is an
association of organizations that regulate the world’s securities and futures markets, i.e. an
association of different securities market regulators (like SEBI in India) across the globe. It
has members from more than 100 different countries that regulate more than 90% of the
world’s securities markets.

There are three different categories of Membership for IOSCO members:

1. Ordinary Members: They are the primary regulators of securities and/or futures
markets in a jurisdiction and each member has one vote. Securities Exchange Board
of India (SEBI) is an ordinary member in IOSCO.

2. Associate Members: In case there is more than one regulator in a jurisdiction, then
the regulator(s) other than the primary regulator are listed here. Though associate
members do not have any voting rights but they can be members of the Presidents’
Committee. Forward Market Commission (FMC) is an associate member of IOSCO.

3. Affiliate Members: This category includes the exchanges, industry associations,

etc. Affiliate members do not have voting rights and cannot be a part of Executive
committee or Presidents’ Committee. Most of the leading exchanges of India are all
affiliate members of IOSCO.

IOSCO’s Technical committee provided the principles for Direct Electronic Access (DEA) to
Markets in August 2010. IOSCO laid down the following principles for Pre-Conditions for DEA,
Information Flow and Adequate Systems and Controls:

1. Pre Conditions for DEA

a. Principle 1: Minimum Customer Standards

The market authorities should have rules in place that requires intermediaries
to have minimum customer standards like appropriate financial resources and
procedures

57
 b. Principle 2: Legally Binding Agreement

There should be a recorded, legally binding contract between the intermediary

and the DEA customer, the nature and detail of which should be appropriate
to the nature of the service provided. Each market should consider whether it
is appropriate to have a legally binding contract or other relationship between
itself and the DEA customer.

c. Principle 3: Intermediary’s Responsibility for Trades

An intermediary retains ultimate responsibility for all orders under its authority,
and for compliance of such orders with all regulatory requirements and market
rules.

2. Information Flow

a. Principle 4: Customer Identification

Intermediaries should disclose to market authorities upon request and in a

timely manner the identity of their DEA customers in order to facilitate market
surveillance. In those jurisdictions where sub-delegation is permitted, the
intermediary also has such responsibility to the market authorities with respect
to any sub-delegates.

b. Principle 5: Pre and Post-Trade Information

Markets should provide member firms with access to relevant pre- and post-
trade information (on a real time basis) to enable these firms to implement
appropriate monitoring and risk management controls.

3. Adequate Systems and Controls

a. Principle 6: Markets

A market should not permit DEA unless there are in place effective systems and
controls reasonably designed to enable the management of risk with regard to
fair and orderly trading including, in particular, automated pre-trade controls
that enable intermediaries to implement appropriate trading limits.

b. Principle 7: Intermediaries

Intermediaries (including, as appropriate, clearing firms) should use controls,

including automated pre-trade controls, which can limit or prevent a DEA
Customer from placing an order that exceeds a relevant intermediary’s existing
position or credit limits.

58
 c. Principle 8: Adequacy of Systems

Intermediaries (including clearing firms) and markets should have adequate

operational and technical capabilities to manage appropriately the risks posed
by DEA.

6.2 Auditing Process and Requirements

(As defined by NSE for member-broker)

General audit requirement are as follows:

• The Half yearly System Audit should be conducted only for Algorithmic trading facility
approved for the Trading Member through Exchange empanelled system auditors.

• Audit Report for Algorithmic facility is required to be submitted to the Exchange through
NSE ENIT

• It is mandatory for the auditors to provide ratings /remarks / recommendation on all

the points wherever not complied. Additionally, auditors are required to give ratings
‘Strong’, ‘Medium’ and ‘Weak’ on each broad areas of audit.

• Auditors are required to provide the list of all the algorithmic strategies approved for
the members and audited by them on their letter head.

• Members are required fill up the electronic summary sheet and upload the details of all
the algorithmic strategies in the template provided on ENIT system

6.3 SEBI Recommendations on Audit

SEBI has been advising and setting broad guidelines on Algorithmic Trading from time to
time. In the Circular [Link]/MRD/DP/09/2012 dated March 30, 2012 on Broad guidelines on
Algorithmic Trading. These guidelines are as follows:

Guidelines for Exchanges:

Stock exchanges to ensure the following while permitting algorithmic trading:

(i) The stock exchange shall have arrangements, procedures and system capability to
manage the load on their systems in such a manner so as to achieve consistent response
time to all stock brokers. The stock exchange shall continuously study the performance
of its systems and, if necessary, undertake system up gradation, including periodic up
gradation of its surveillance system, in order to keep pace with the speed of trade and
volume of data that may arise through algorithmic trading.

(ii) In order to ensure maintenance of orderly trading in the market, stock exchange shall
put in place effective economic disincentives with regard to high daily order-to-trade

59
 ratio of algorithmic trading orders of the stock broker. Further, the stock exchange
shall put in place monitoring systems to identify and initiate measures to impede any
possible instances of order flooding by algorithms.

(iii) The stock exchange may seek details of trading strategies implemented through
algorithmic trading for such purposes viz. inquiry, surveillance, investigation, etc.

(iv) Terminals of the stock broker that are disabled upon exhaustion of collaterals shall
be enabled manually by the stock exchange in accordance with its risk management
procedures.

(v) The stock exchange shall include a report on algorithmic trading on the stock exchange
in the Monthly Development Report (MDR) submitted to SEBI inter-alia incorporating
turnover details of algorithmic trading, algorithmic trading as percentage of total trading,
number of stock brokers / clients using algorithmic trading, action taken in respect of
dysfunctional algorithms, status of grievances, if any, received and processed, etc.

(vi) The stock exchange shall synchronize its system clock with the atomic clock before
the start of market such that its clock has precision of atleast one microsecond and
accuracy of atleast +/- one millisecond.

(vii) Stock exchange shall ensure that the stock broker shall provide the facility of algorithmic
trading only upon the prior permission of the stock exchange. Stock exchange shall
subject the systems of the stock broker to initial conformance tests to ensure that
the checks mentioned below are in place and that the stock broker’s system facilitate
orderly trading and integrity of the securities market. Further, the stock exchange shall
suitably schedule such conformance tests and thereafter, convey the outcome of the
test to the stock broker.

For stock brokers already providing algorithmic trading facility, the stock exchange
shall ensure that the specified risk controls are implemented by the stock broker.

(viii) The stock broker, desirous of placing orders generated using algorithms, shall submit
to the respective stock exchange an undertaking that -

• The stock broker has proper procedures, systems and technical capability to
carry out trading through the use of algorithms.

• The stock broker has procedures and arrangements to safeguard algorithms

from misuse or unauthorized access.

• The stock broker has real-time monitoring systems to identify algorithms that
may not behave as expected. Stock broker shall keep stock exchange informed
of such incidents immediately.

60
 • The stock broker shall maintain logs of all trading activities to facilitate audit
trail. The stock broker shall maintain record of control parameters, orders,
trades and data points emanating from trades executed through algorithm
trading.

• The stock broker shall inform the stock exchange on any modification or change
to the approved algorithms or systems used for algorithms.

In addition to above guidelines, in the Circular [Link]/MRD/DP/16/2013 dated May 31, 2013,
SEBI laid out additional guidelines pertaining to Audit.

1: The stock brokers/ trading members that provide the facility of algorithmic trading
shall subject their algorithmic trading system to a system audit every six months in
order to ensure that the requirements prescribed by SEBI / stock exchanges with
regard to algorithmic trading are effectively implemented

1.1: Such system audit of algorithmic trading system shall be undertaken by a system
auditor who possesses any of the following certifications:

a) CISA (Certified Information System Auditors) from ISACA;

b) DISA (Post Qualification Certification in Information Systems Audit) from

Institute of Chartered Accountants of India (ICAI);

c) CISM (Certified Information Securities Manager) from ISACA;

d) CISSP (Certified Information Systems Security Professional) from International

Information Systems Security Certification Consortium, commonly known as
(ISC)

1.2: Deficiencies or issues identified during the process of system audit of trading algorithm
/ software shall be reported by the stock broker / trading member to the stock exchange
immediately on completion of the system audit.

1.3: In case of serious deficiencies / issues or failure of the stock broker / trading member
to take satisfactory corrective action, the stock exchange shall not allow the stock
broker/ trading member to use the trading software till deficiencies / issues with the
trading software are rectified and a satisfactory system audit report is submitted to
the stock exchange. Stock exchanges may also consider imposing suitable penalties
in case of failure of the stock broker/ trading member to take satisfactory corrective
action to its system within the time-period specified by the stock exchanges.

6.4 Exchange Audits

The Exchange audit process shall broadly cover the following aspects:

• Approved features and system parameters implemented in the trading system

61
• Adequacy of input, processing and output controls should be tested

• Adequacy of the application security should be audited

• Event logging and system monitoring

• Robust Password management standards

• Network management and controls

• Backup systems and procedures

• Business continuity and disaster recovery plan

• Proper Documentation for system processes

• Security features such as access control, network firewalls and virus protection should
be actively managed

The stock exchange shall ensure that all algorithmic orders are necessarily routed through
broker servers located in India and the stock exchange has appropriate risk controls mechanism
to address the risk emanating from algorithmic orders and trades. The minimum order-level
risk controls shall include the following:

a. Price check: The price quoted by the order shall not violate the price bands defined
by the exchange for the security. For securities that do not have price bands, dummy
filters shall be brought into effective use to serve as an early warning system to
detect sudden surge in prices.

b. Quantity Limit check: The quantity quoted in the order shall not violate the maximum
permissible quantity per order as defined by the exchange for the security.

In the interest of orderly trading and market integrity, the stock exchange shall put in place
a system to identify dysfunctional algorithms (i.e. algorithms leading to loop or runaway
situation) and take suitable measures, including advising the member, to shut down such
algorithms and remove any outstanding orders in the system that have emanated from such
dysfunctional algorithms. Further, in exigency, the stock exchange should be in a position to
shut down the broker’s terminal.

The stock broker, desirous of placing orders generated using algorithms, shall satisfy the stock
exchange with regard to the implementation of the following minimum levels of risk controls
at its end -

(i) Price check– Algorithmic trading orders shall not be released in breach of the price
bands defined by the exchange for the security.

(ii) Quantity check– Algorithmic trading orders shall not be released in breach of the
quantity limit as defined by the exchange for the security.

62
(iii) Order Value check- Algorithmic trading orders shall not be released in breach of the
‘value per order’ as defined by the stock exchanges.

(iv) Cumulative Open Order Value check– The individual client level cumulative open order
value check, may be prescribed by the broker for the clients. Cumulative Open Order
Value for a client is the total value of its unexecuted orders released from the stock
broker system.

(v) Automated Execution check– An algorithm shall account for all executed, un-executed
and unconfirmed orders, placed by it before releasing further order(s). Further, the
algorithmic trading system shall have pre-defined parameters for an automatic stoppage
in the event of algorithmic execution leading to a loop or a runaway situation.

(vi) All algorithmic orders are tagged with a unique identifier provided by the stock exchange
in order to establish audit trail.

6.5 Technology and System Audit

System Audit includes the following main areas:

• Risk Management

• Security Policy and Implementation

• Capacity Management

• Disaster Recovery and Backups

• Vulnerability

The following gives a detailed account of the areas of system audit:

Location Confirmation

• Complete address of broker’s server routing Algorithm orders to the Exchange trading
system

Risk Management

The installed system is capable of assessing the risk as soon as the algorithm orders are
generated and informs the user of rejection of the order (if any) within a reasonable period.

Risk Management Tools:

• Should allow for risk management of the orders placed and online risk monitoring of
the orders being placed.

• The system has functionality for mandatorily routing of orders generated by algorithm
through the automated risk management system and only those orders that are within

63
 the parameters specified in the risk management systems are allowed to be released
to exchange trading system.

• The risk management system has following minimum levels of risk controls functionality
and only algorithm orders that are within the parameters specified by the risk
management systems are allowed to be placed.

 Individual Order Level

 Quantity Limits

 Price Range checks

 Trade price protection checks

 Order Value Checks(Order should not exceed the limit specified by the
Exchange)

 Market price protection

 Spread order Quantity and Value Limit

 Client Level

 Cumulative Open Order Value Check

 Automated Execution check

 Net position v/s available margins

 RBI violation checks for FII restricted stocks

 Market-wide Position Limits (MWPL) violation checks

 Position limit checks

 Trading limit checks

 Exposure limit checks at individual client level and at overall level for all
clients

 Branch value limit for each branch ID

 Security wise limit for each user ID

 Identifying dysfunctional algorithms

• Does system has functionality to specify values as unlimited for any risk controls listed
above?

• Does the member have additional risk controls / policies to ensure smooth functioning
of the algorithm?(if yes, please provide details)

64
Execution of Orders / Order Logic

The installed system provides a system based control facility over the order input process

• Order Numbering Methodology

• The system has an internal unique order numbering system.

• All orders generated by Algorithm system are offered to the market for matching and
system does not have any order matching function resulting into cross trades.

• Whether algorithm orders are having unique flag/ tag as specified by the Exchange. All
orders generated from algorithmic system are tagged with a unique identifier – 13th
digit of NNF field is populated with 0.

Database Security

The system has sufficient controls over the access to and integrity of the database

• The access to the database is allowed only to authorized users / applications

• The database is hosted on a secure platform

• The database stores all the details of user ids activated for along with user names and
passwords securely

• Frequency of antivirus patch updates

System Authentication

The system Authentication mechanism is as per the guidelines of the NSE

• The password policy / standard are documented

• The system allows access to only authorized users

• The system has a password mechanism which restricts access to authenticated users

• The system requests for identification and new password before login into the system

• The system has appropriate authority levels to ensure that the limits can be setup only
by persons authorized by the risk / compliance manager

The installed system’s Password features include

• The Password is masked at the time of entry

• System mandated changing of password when the user logs in for the first time

• Automatic disablement of the user on entering erroneous password on three consecutive

occasions

• Automatic expiry of password on expiry of 14 calendar days

65
• System controls to ensure that the password is alphanumeric (preferably with one
special character), instead of just being alphabets or just numerical

• System controls to ensure that the changed password cannot be the same as of the
last password

• System controls to ensure that the Login id of the user and password should not be the
same

• System controls to ensure that the Password should be of minimum six characters and
not more than twelve characters

• System controls to ensure that the Password is encrypted at members end so that
employees of the member cannot view the same at any point of time

• System control ensures Re-initialization of access on entering fresh passwords

System Backup

The Installed systems backup capability is adequate as per the requirements of the exchange
for overcoming loss of product integrity.

• Are backups of the following system generated files maintained as per the exchange
guidelines?

• At the server/gateway level

 Database

 Audit Trails

 Reports

• At the user level

 Logs

 History

 Reports

 Audit Trails

 Alert logs

• Does the audit trail capture the record of control parameters, orders, trades and data
points emanating from trades executed through algorithm trading?

• Are backup procedures documented?

• Are backup logs maintained?

• Have the backups been verified and tested?

66
• Are the backup media stored safely in line with the risk involved?

• Are there any recovery procedures and have the same been tested?

Information Security

To ensure information security for the Organization in general and the system in particular
policy and procedures as per the NSE requirements must be established, implemented and
maintained.

• Whether installed systems & procedures are adequate to handle algorithm orders/
trades?

• Maintenance of User details: Whether details of users activated for algorithm facilities
is maintained along with user name, unique identification of user, authorization levels.

• Does the organization’s documented policy and procedures include the following
policies and if so are they in line with the NSE requirements?

 Information Security Policy

 Password Policy

 Network Security Policy

 Application Software Policy

 Backup Policy (data should be available for minimum 5 years)

 Audit Trail Policy (data should be available for minimum 5 years)

• Whether all the documents are classified as per CIA(Confidentiality, Integrity and
Availability)

• Does the organization follow any other policy or procedures or documented practices
that are relevant?

Firewall

• Is a firewall implemented?

• Are all servers placed in a DMZ and segregated from other zones by using a firewall?

• Is there segregation between application and database servers?

• Are user and server zones segregated?

• Is specific port/service access granted on firewall by following a proper approval

process?

Physical Security

• Physical Access Control

67
• Server Room/Network Room Security (Environmental Controls)

• Server Room/Network Room Security (UPS)

• Server Room/Network Room Security (HVAC)

System Records

The system and system records with respect to Risk Controls are maintained as prescribed
by the Exchange

• The limits are setup after assessing the risks of the corresponding user ID and branch
ID

• The limits are setup after taking into account the member’s capital adequacy
requirements

• All the limits are reviewed regularly and the limits in the system are up to date

• All the branch or user have got limits defined and that no user or branch in the system
is having unlimited limits on the above stated parameters

• Daily record of these limits is preserved and shall be produced before the Exchange as
and when the information is called for

• Compliance officer of the member has certified the above in the quarterly compliance
certificate submitted to the Exchange

6.6 Compliance Requirements

In the circular, NSE/CMTR/21793 dated September 28, 2012, exchange laid out compliances
requirements:

• The Algorithmic orders are routed through broker servers located in India.

• The system has the capabilities to monitor algorithms real-time to identify those
algorithms that may not behave as expected and bring it to the notice of the Exchange
immediately.

• The system maintains logs of all trading activities including record of control parameters,
orders, trades and data points emanating from trades executed through algorithm
trading to facilitate audit trail.

• The system releases further order(s) only after accounting for all executed, unexecuted
and unconfirmed orders placed earlier. Further, system shall have pre-defined
parameters for an automatic stoppage in the event of Algorithmic execution leading to
a loop or a runaway situation.

68
• Any modifications / Change to the approved decision support tool / Algorithm to be
effected only on prior approval of Exchange.

• All Algorithmic orders emanating from the system to be tagged with the unique identifier
as specified by Exchange from time to time

System compliance requirement for Algorithmic Trading Facility on annual Basis:

Trading members who have obtained approval from Exchange for CTCL trading software
(including all the applications that is CTCL, IBT, DMA, STWT and SOR) are required to submit
to the Exchange the System Audit Report for the year ended March 31, every year, after
getting the CTCL trading facility audited from certified auditor, independent of the empanelled
vendors of the Exchange and/or Partners/Directors of the trading members. Members are
required to submit the system audit report to the Exchange through NSE ENIT electronically
on or before April 30.

System audit requirement for Algorithmic Trading Facility on half yearly Basis:

As part of half yearly compliance as mandated by SEBI, trading members who have obtained
approval from Exchange for Algorithmic trading software are required to submit the System
Audit Report for the half year ended March 31 (i.e. for the period from October 01 to March 31)
and September 30 (i.e. for the period April 01 to September 30), after getting the Algorithmic
trading facility audited from Exchange empanelled system auditors (CISA certified).

69
Self Assessment

• Audit is compulsory for

– Exchange members who are providing brokerage services

– Institutional brokers

– All members of the exchange

– Only those exchange members who are trading using algorithms

• The full form of CTCL is:

– Centralized Trading and Clearing Limited

– Computer To Computer Link

– Computerized Trading Corporation Limited

– Centralized Time Correction Link

• Orders to be included in Audit trail are:

– Executed orders only

– Unexecuted orders only

– All orders

– All new orders only

• It is not possible to include market orders for cumulative open order value check.

– True

– False

• As per the prescribed back up policy, data should be available for at least:

– 1 year

– 3 years

– 5 years

– 8 years

70
• Following type of trades are exempted from the audit compliance process:

– Exchange member’s proprietary trades on the exchange

– Trades done by the client of exchange member

– Trades done by the institutional client of the exchange member

– None of the above

• A broker can route his trade from the servers located anywhere in the world but his
head office has to be in India.

– True

– False

71