ALGORITHMIC PATTERN TRADING SYS-

TEM
Multi-Asset · Alternative Data · Swing Trading · Indian Markets

WHAT THIS SYSTEM DOES

This is a full end-to-end algorithmic trading research and signal system built
on your sister’s core insight: everything is connected, and everything has
patterns.
It pulls data from multiple sources — stock prices, volumes, weather, macro
indicators, forex, commodities — merges them into a single master dataset, and
uses rigorous statistical methods to find patterns that would be invisible to a
human analyst looking at one chart at a time.

SYSTEM ARCHITECTURE
���������������������������������������������������������������
� [Link] (orchestrator) �
���������������������������������������������������������������
�
�������������������
� data_engine.py � ← Fetches all data, builds master dataset
� � Price/Volume (yfinance)
� � Weather (Open-Meteo, free)
� � Macro (yfinance proxies)
�������������������
�
�������������������
� pattern_engine � ← The mathematical core
� .py � Cross-correlation at lags 0–15 days
� � Granger Causality testing
� � Conditional analysis (your sister's method)
� � PCA — what factors drive markets?
� � Cointegration — pairs trading foundation
� � Regime detection — Bull/Bear/Sideways
�������������������
�
�������������������
� signal_engine � ← Technical indicators + composite scoring
� .py � RSI, MACD, EMA, Bollinger, ATR, ADX,
� � Stochastic, OBV, Volume analysis

1
 � � Buy/Sell score 0–100
�������������������
�
�������������������
� risk_engine.py � ← Position sizing, portfolio risk
� � ATR-based stop losses
� � Kelly Criterion
� � VaR (parametric + historical)
� � Max drawdown monitoring
� � Correlation-adjusted sizing
�������������������
�
�������������������
� report_engine � ← Beautiful terminal output + HTML reports
� .py �
�������������������

[Link] ← All settings in one place (edit this first)

QUICK START
1. Install dependencies
pip install -r [Link]

2. Configure the system

Open [Link] and set: - PORTFOLIO_CAPITAL — your actual capital in INR
- MAX_RISK_PER_TRADE — default 1.5% (conservative for swing trading) - Add
instruments you want to track

3. Run the system

# Full daily run (recommended — takes 3-5 minutes first time)
python [Link]

# Fast signal scan only (uses cached data — takes ~30 seconds)
python [Link] --signals-only

# Deep dive on one instrument

python [Link] --instrument RELIANCE
python [Link] --instrument NIFTY50
python [Link] --instrument TATAMOTORS

# Force re-download all data

2
python [Link] --refresh

# Pattern discovery only

python [Link] --patterns-only

DATA SOURCES (ALL FREE)

Source Data How

Yahoo Finance (yfinance) Price, Volume, API, free
OHLCV
Open-Meteo Mumbai weather API, free, no key
(historical + current) needed
Yahoo Finance Macro proxies (VIX, Via yfinance
DXY, Crude, Gold)
Calendar Day-of-week, month, Computed
quarter effects

Optional (better weather data): - Sign up at [Link] (free

tier) - Add your API key to [Link] under WEATHER_API_KEY

WHAT THE PATTERN ENGINE FINDS

1. Lagged Correlations
“Does crude oil price today predict Nifty 3 days from now?” The system tests
every variable against every lag from -5 to +15 days.

2. Granger Causality
More rigorous than correlation. Tests whether past values of variable X contain
information about future Y, beyond what Y’s own history tells you.

3. Conditional Patterns (Your Sister’s Method)

Splits the data by condition and measures average returns: - Rainy days in
Mumbai → market returns next day: +X% vs baseline - VIX above 75th per-
centile → market tends to… - USD strengthening vs INR → Nifty tends to…
- Monday vs Friday effects - Month start / month end effects All tested with
t-tests for statistical significance.

3
4. PCA — Underlying Factors
Reduces all instruments to their underlying drivers. Answers: “What are the
3–5 real forces moving everything simultaneously?”

5. Cointegrated Pairs
Finds pairs of instruments in a long-run statistical relationship. Foundation for
spread/pairs trading strategies.

6. Market Regime Detection

Classifies each day as BULL / SIDEWAYS / BEAR using rolling returns +
volatility + KMeans clustering.

RISK MANAGEMENT
Every signal comes with:

Parameter How it’s calculated

Stop Loss Entry − (ATR × 2.0)
Target Entry + (ATR × 2.0 × 2.5)
Position Size Capital × 1.5% risk ÷ stop distance
R:R Ratio Fixed 1:2.5 minimum
Max positions 8 concurrent (configurable)
Max portfolio risk 6% of capital (configurable)

Kelly Criterion is also implemented for optimal long-run position sizing based
on historical win rate and average win/loss.
Value at Risk (VaR) computed both parametrically (assumes normality) and
historically (uses actual return distribution), plus CVaR (Expected Shortfall).

FILE STRUCTURE
trading_system/
��� [Link] ← Run this
��� [Link] ← Edit this first
��� data_engine.py ← Data fetching and caching
��� pattern_engine.py ← Statistical pattern discovery
��� signal_engine.py ← Technical indicators and signals
��� risk_engine.py ← Risk management and position sizing
��� report_engine.py ← Output formatting

4
��� [Link] ← pip install -r this
��� data/ ← Cached CSV data (auto-created)
��� reports/ ← HTML and CSV reports (auto-created)
��� logs/ ← Trade log (auto-created)

INSTRUMENTS COVERED
Indices: Nifty50, Sensex, BankNifty, Nifty IT, Nifty Midcap
Equities: Reliance, TCS, HDFC Bank, Infosys, ITC, SBI, Tata Motors, Bajaj
Finance, Wipro, Maruti
Commodities: Gold, Crude Oil (WTI), Silver, Natural Gas
Forex: USD/INR, EUR/INR, GBP/INR, JPY/INR
Global: S&P500, NASDAQ, DOW, VIX, FTSE, Nikkei, Hang Seng, Brent
Crude
Crypto: Bitcoin, Ethereum (for correlation analysis)

IMPORTANT DISCLAIMERS
1. This is a research tool, not financial advice. All signals are for
educational and research purposes only.
2. Backtesting is not the same as live trading. Discovered patterns
may not persist in the future. Past correlations can and do break down.
3. The weather-market pattern your sister found should be treated
as a hypothesis to be tested rigorously, not a confirmed edge. The system
tests its statistical significance, but be skeptical of any p-value until you
have seen it hold out-of-sample.
4. Never risk money you cannot afford to lose. Paper trade first.
5. This system does not connect to any broker. It is a research and
analysis system only. All trade execution remains manual.

YOUR MATHEMATICAL BACKGROUND — WHERE

IT APPLIES

5
Your knowledge Where it’s used in this system
Stochastic Calculus Foundation for understanding price dynamics
(GBM), VaR models
Probability & Statistics Hypothesis testing on patterns, VaR, Kelly
criterion
Linear Algebra PCA decomposition, correlation matrices,
covariance
Calculus Optimization in signal scoring, gradient of risk
functions
Economics Interpretation of macro signals, regime analysis
Geopolitics Context for why certain correlations exist
(crude→INR→Nifty)

Your PhD background means you can extend this system with: - Ornstein-
Uhlenbeck process for mean-reversion signals - GARCH models for volatility
forecasting - Hidden Markov Models for regime detection - Bayesian updating
of pattern probabilities as new data arrives