Hold'em roll-out simulators?

Greetings -

I was wondering if there's any good Hold'em calculators out there, either as freeware, open source, or for purchase. Best I managed to find was the calculator tool in ddpoker, although the demo is pretty limited and I'd prefer something fast enough to do exact simulations rather than partial.

I ended up writing a simple one for a windows command line that can do most of what I want, but I'm wondering if I'm just re-inventing the wheel. If not, maybe I'll put mine up on the web for folks to play with. At the moment it's able to do up to 10 hands with common and mucked cards specified, will give breakdowns of which types of hands you'll make using either 0, 1 or 2 cards from you hand and can do a complete roll-out with one random hand in less than 10 minutes on an Athlon64 3200+.

If you do a random hand it's kind of a beast for memory, since I pre-calculate and store the best 5 card hands for all 133M+ 7 card hands (it uses roughly 700MB of memory on my machine while running). If there's no random cards it only uses about 50MB and usually finishes in around 15s or less. I know the command line would probably drive most folks off, but I'm still curious about whether folks would find it useful.

[Nomad84]

Is this what you are looking for? Poker Stove

[euri10]

look at pokersource on sourceforge, one of the packages poker-eval does all you need

I have to admin that pokerstove is pretty amazing for how fast it rolls out. I did feel that my tool was somewhat unique, since it will break hands down by type as well. I put a version up at the following location:

http://home.comcast.net/~holdout/

It's command line only for win32, with a simple installer that will show you the readme. Send me some email at holdout@comcast.net if you have any comments, bugs to report or requests for features.

[pokergrader]

Storing all possible combinations of the 7 card hands is incredibly wasteful. In fact, because most computers will store some of your huge memory in virtual memory, it would be much faster just to efficiently process each 7 hand combination individually.

PokerStove is so fast not because it can process hands quickly, but because it is very smart about which boards it needs to process and which ones it doesn't, and lumps them all together.

I have not looked at or run your program, but if you really want to make it more viable for anybody to run, fix up the code that uses >500 MB of ram, as it is probably slowing the program down.

[ QUOTE ]
Storing all possible combinations of the 7 card hands is incredibly wasteful. In fact, because most computers will store some of your huge memory in virtual memory, it would be much faster just to efficiently process each 7 hand combination individually.

I have not looked at or run your program, but if you really want to make it more viable for anybody to run, fix up the code that uses >500 MB of ram, as it is probably slowing the program down.

[/ QUOTE ]

Thanks for your input, and I have to say again how impressively fast pokerstove is.

Truthfully, this program was not designed with a lot of forethought... it was more like a "what the heck" that just kept growing. When I designed the hand comparison code, I had it do 5 card hands for maximum flexibility... unfortunately this generates 21 combinations to evaluate for every 7 card group, which is expensive.

Since my machines at both work and home have 1GB+, I just threw memory at the problem instead which showed a huge increase in speed. Even my 768MB sempron doesn't swap after a couple of runs. Since a random hand against one other will generate almost every possible 7 card combination, loading the array from disk saved a lot of evaluation time (since the result is now an indexed reference). Without a random hand I don't do it since most combinations are not used.

I do have optimizations in there already to pare the random hands down to ones that are uniquely suited, but I don't do the same for the board evaluations.

One useful thing I ended up with is being able to bin the results by hand types and strengths, as well as separating 0, 1 and 2 down card hands for win/loss/tie. Using the random hand it's also possible to have the program iterate all the hands with better or equal pot equity, and at some point I could even sort these by relative strength or use some threshold criteria besides better/equal.

I also calculate hand strengths based on the number of 5 card combinations that are beaten, which I was hoping to be able to do some statistical analysis on... unfortunately hands do not follow a gaussian distribution where something like a variance is useful.

PokerStove appears to be about an order of magnitude faster which makes the 2+ random hand calculations a bit more practical. I just thought that the hand breakdown showed enough merit to throw it out there for folks.