It's been about 24 hours since I released Strip Texas Holdem Poker. The game has, of course, been a dirty little snitch about how you all play it, so let's look at some statistics!

http://nanimoose.furryhome.com/holdem/stats.php
(those are live statistics, so the numbers there may not mesh with what I've observed now as time marches on)
(also you can click on the stat bars to turn data filters on and off)

First of all, it's been played about 1500 times in its first day, and the majority of games are actually played to completion and not abandoned. That's exceeded my expectations and makes me so happy.

Of the games that are completed, the win rate is about 12%. If it were truly random who of the six players wins, the win rate would be more like 17% (1/6). This means that my AI is posing some challenge, but is certainly not impossible to beat, which is exactly what it's intended to do.

For character choice, I'm surprised to see that newcomer Blackheart is the favorite (20.39%), edging out Pai Gow's favorite Raz (19.77%), and other newcomer Buck is in a distant third place (15.43%).

Speaking of statistics, do you wonder how the AI decides weather certain pocket cards are worth going in on or not? It turned out to be really hard to find odds calculations for the various card combinations for a variable number of opponents, so I made my own:

http://nanimoose.furryhome.com/stuff/PokerVsStats.gif

In this table, the two pocket cards are on the two axis, so there's a diagonal line for the pocket-pair cards. Anything above/right of that line is off-suit, and anything below/left is suited. Suits aren't differently valued in the game, so it doesn't really matter what the suits are, just if they match or not.

In that animated GIF you can see how the odds change as players are added, and the rarer matches become more probable. The most obvious example is Flushes; With 2 players, there's hardly an advantage to playing suited cards... but when you get up to 6 players, the advantage is staggering!

You can also see the tool I made to get these probability statistics:
http://nanimoose.furryhome.com/holdem/prob.html
This just brute-forces random games as fast as it can. The statistics start out pretty noisy, but after a while the law of averages kicks in. The longer you let it run, the more accurate the results are.