Do Computer Backgammon Software Cheat?
1. What is backgammon?
Backgammon is one of the oldest and most popular board games
in the world. The oldest backgammon sets discovered date back to
5000 BC [1]. Aside from being a household game in many countries
around the world, there are international competitions and prizes
involved, which have led to some players pursuing professional
careers in the game. Like chess, backgammon has been a subject of
interest to many computer programmers, resulting in computer
software having been developed to take part in competitions, face,
and even beat human participants.
Traditional backgammon
consists of a board, 30 pieces (two different colors, 15 apiece) and 2
dice. The game play involves the two players rolling the 2 dice in
turns and moving their pieces accordingly, and the winner is the
player that manages to bring all of his pieces to his base and take
them out first [2]. The players have to use tactics to both move their
pieces more quickly and efficiently, and also anticipate and counter
the opponents strategies. Nevertheless, like all other games that
involve the rolling of dice, luck has an influence on the game.
2. Why are computers accused of cheating?
A player wins a match after winning a previously agreed number
of games (usually five games), and in regular human games, luck
has a very small influence in determining the overall winner.
Although it is important in the short run, in the long run the players
tactical abilities are a much more decisive factor than the numbers
on the dice. When a human player plays against a computer
software on a PC or a smartphone, it is the CPU that rolls dice for
both players in turns. The high number of doubles and also the high
overall dice total that the CPU player gets compared to what the
human player gets, and the frequency with which this occurs,
however, has led to many people accusing computer software of
cheating [3]. A double is when both of the rolled dice have the
same face values. It is significant because it allows the player to
move twice the total number on the rolled dice, or in other words
four times the number on one of the dice. For example when a
player rolls a double 5, he will have the ability to move 4*5 = 20
times. Using the general doubles formula that 4*(the number on one
dice) = (the number of allowed moves), we get the following table
for moves allowed for a double.
Moves\Di
ce
12
16
20
24
3. What are the possibilities of rolling doubles?
It was mentioned that one of the ways that backgammon
software are accused of cheating is by rolling many doubles in every
game. But what are the chances of rolling a double? The chances of
rolling any number from 1 to 6 are (1/6). The chances that the
number on the next dice is the same as the number on the first one
are (1/6). Therefore the chances of getting a double every time you
roll the two dice are (1/6)*(1/6) = (1/36) or 0.028.
4. What are the probabilities of getting each of the different
number combinations?
The different dice face value combinations which we can get
every time we roll the two dice are as follows: [(1,1),(1,2),(1,3),(1,4),
(1,5),(1,6),(2,1),(2,2),(2,3),(2,4),(2,5),(2,6),(3,1),(3,2),(3,3),(3,4),
(3,5),(3,6),(4,1),(4,2),(4,3),(4,4),(4,5),(4,6),(5,1),(5,2),(5,3),(5,4),
(5,5),(5,6),(6,1),(6,2),(6,3),(6,4),(6,5),(6,6)].
The possible number of moves we can get from these combinations
(taking the previously mentioned doubles rule into account), ranges
from 3 which we can get from (1,2) or (2,1), up until 24 which we
can get from a double 6, i.e. (6,6). The probability of us getting a 3
from a (1,2) combination is (1/6)*(1/6) = (1/36). However that is not
the only combination which gives us a 3, and we can also get it from
a (2,1) combination, which also has a probability of (1/36). Note that
the probability of getting any of the 36 above combinations is
(1/36). The probability of getting any number of moves is
(1/36)*(the number of combinations that give us that value).
Therefore the probability of getting a 3 is (1/36)*2 = (1/18).
Similarly, the probability of getting a 24 is (1/36)*1 = (1/36). The
following table shows possible moves and the possible combinations
that give us those numbers.
Move\Combinati
ons
3
4
5
6
7
8
9
10
11
12
(1,2),(2,1)
(1,3),(3,1),(1,1)
(1,4),(4,1),(2,3),(3,2)
(1,5),(5,1),(2,4),(4,2)
(1,6),(6,1),(2,5),(5,2),(3,4),(4,3)
(2,6),(6,2),(3,5),(5,3),(2,2)
(3,6),(6,3),(4,5),(5,4)
(4,6),(6,4)
(5,6),(6,5)
(3,3)
16
20
24
(4,4)
(5,5)
(6,6)
The probabilities of getting each of these numbers of moves using
the aforementioned formula (1/36)*(the number of combinations
that give us that value) are:
Combination Probabilities
0.18
0.16
0.14
0.12
Probability
0.1
0.08
0.06
0.04
0.02
0
3
10
11
12
16
20
24
5. Do they really cheat?
The following data is from the 65 times that I have played the
CPU player on my mobile phones backgammon software by AI
games, one of the leading softwares available according to the
developer [4]. It clearly shows that the CPU has gotten (258215)*100/(215) = 20% more doubles than I have.
Also, if we divide the data in the chart in section 4 into lower
numbers [3,8] and higher numbers [10,24] of allowed moves, we
can clearly see that there is a lower probability of getting higher
numbers and especially doubles combinations of 3, 4, 5 and 6. In
the end, we can see that despite it being unlikely, the CPU player
gets an extra double on average per game (258/65 = 4 compared to
215/65 = 3) as evidenced by data collected over a relatively large
number of games, helping increase its dice total for each game and
giving it an edge over its human opponent.
Although this may show that the software cheats however, it
must be mentioned that there are limitations to the above
arguments. First and foremost it should be noted that the above
equations are probability equations, and although something may
look unlikely, its occurrence is not impossible and is not necessarily
a sign of cheating. Second, although we have a fairly large database
of 65 games, it may not be big enough to give an accurate general
analysis. In order to conduct a more accurate comparison, it is
better to have a larger database of 100+ games, and then we can
judge the results with more certainty.
References
[1] [Link]. (2012). Backgammon History [Online].
Available: [Link]
[2] Backgammon Galore. (2012). Rules of Backgammon [Online].
Available: [Link]
[3] AI Factory. (2012). Backgammon Programs Cheat [Online].
Available:
[Link]
htm
[4] AI Factory. (2012). Quarterly/Biannual Periodical [Online].
Available: [Link]