I wish Silver Creek Entertainment would spend the time developing in a Backgammon game that would play WELL, rather than play LUCKY. Regardless of the commentary from Silver Creek, the game makes some poor opening moves and has runs of simply improbably luck.
For example, if the computer opens 6-4, it plays 8-2, 6-2, making a point on the 2 pip, which is too deep to defend. If it opens 5-3, it plays 8-3, 6-3 which is worse than 13-8, 13-10. These are simply examples, there are many moves when the computer opts to build (or land on) an anchor when leaving a blot with the intention of building is better.
Despite these errors, the machine wins 90% of the time. For instance, today I was ahead. I had a computer piece on the bar and both of us had built primes on the 1, 3, 4, 5 and 6 spaces. I had a lousy roll and had to put a blot on the 7 position, which opened me to a 2-5 (or 5-2) shot. I grumbled because I figured the computer would roll a 2-5 - which it did. A 5.5% shot.
Then I couldn't get in on the computer's 2 spot for the next 11 rolls. The chance of me getting in on any roll is 11/36 or 30.5%. The probability of me staying out for 11 rolls is 30.5% "raised to the 11th power", or 1 in 461,326.
If you average 50 rolls each in a game (this is a reasonably long game), you would expect to see this happen about once in 9,200 games (or on average pop up about the 4,600th game). I have only played a couple of hundred games with the computer so this is not very likely.
Get this - the chance of the computer rolling a "2" followed by me not being able to get in for 11 rolls is far worse, 1 in 8,387,745...
I have also documented a NUMBER of times when I can't get in when there is one spot open - for 8 consecutive rolls. By the same logic this only happens 1 in 13,160 times. Again, once every 263 games - but I've seen this happen a dozen times.
The machine cheats.
If Silver Creek would like me to provide a Hypothesis Test asking what are the chances of seeing this behavior and having honest dice, I can provide another post. It takes a little time - but the probabilities are much less than 1 in 9,200 games or 1 in 263 games.
