Something you may already know (or need to know)

[theredpill]

Hey. I'm a newbie so if you already know this then disregard. I'm a math guy so I like to dig into the mathematics of poker. I decided to check out some of the free odds calculators on various web sites. Well, I'm not impressed. I say let the fish use those odds calculators all they want. You probably should even encourage it. I decided to make my own odds calculator using C++. Don't worry it is too slow to use online. I believe it is much more exact than the other stuff being thrown out there. I won't mention which ones I tested in case a fish drops by. (hopefully, I'm not a fish anymore or am getting there) Sometimes, my calculations were different than the free programs by 10 % . There is one popular one out there that is total bull. My calculations were almost never in line with theirs. My calculations were sometimes off by 20 to 30 %. I was calculating the win % of a hand given a 3 card flop in my program.

Ok, here is an example:

You have A [img]/images/graemlins/spade.gif[/img] 9 [img]/images/graemlins/diamond.gif[/img] and the flop is 10 [img]/images/graemlins/diamond.gif[/img] 6 [img]/images/graemlins/diamond.gif[/img] K [img]/images/graemlins/club.gif[/img] . You are playing heads-up on the flop with a guy in early position. Of course, some of these programs don't take this into consideration. One program gives you a win % of 49.8 % and your opponent 47.4 % . My calculation for this hand gave a win % of 19% and opponent 75.8 % which I believe is more accurate. Another program gave a win % of 28 %.
What do you think ? I am new at poker so I can't give you any experience from a hand like this. I just don't think those odds calculators are all they are cracked up to be. If you know a fish that uses one, tell him he's on the right track.

It takes 20 minutes for my program to give an accurate answer and that's on a 2 ghz computer. I am still tweaking to see if I can get it down but I doubt I will get it down to a few seconds.

[jdl22]

We need more information.

</font><blockquote><font class="small">En respuesta a:</font><hr />
You have A 9 and the flop is 10 6 K . You are playing heads-up on the flop with a guy in early position. Of course, some of these programs don't take this into consideration. One program gives you a win % of 49.8 % and your opponent 47.4 % . My calculation for this hand gave a win % of 19% and opponent 75.8 % which I believe is more accurate.

[/ QUOTE ]

Depends on his hand. I'm assuming you forgot to put that in. How does your program calculate this? How does it take # of players in the hand and position into account?

Others simply play a gazillion hands of no fold'em hold'em and give the results. They are generally believed to be reliable for what they do but of little value because they don't take into account the way people play hands. If yours does I'm sure many here would like to know how.

[fnord_too]

20 minutes seems awfully long. I also have written code (in Java) for doing analysis. Though I don't attack the problem you are attacking with it, I can't immagine it would take even 1 minute if I did. If you PM me your basic approach to the problem (a little more detailed than "Monte Carlo" or "Deterministic") I would be happy to comment on it. Also, you may have some very sub-optimal code to compare hands. If I take two hands pre flop and run through every possible 5 card board, I think I take about 7 seconds. I would have to check this, it has been a while since I played with it. At any rate, I have some experience in this type of analysis, which may or may not be of benefit to you.

The other point I would make is that this seems to be a smarter hot and cold analysis, but still a hot and cold analysis. If you are not taking into account betting patterns and the play of the hand then you are leaving a lot of ground uncovered. (Maybe you are playing the hands out with betting and that's why it is taking 20 minutes to run the simulations.) I am guessing that you can vary pre-flop calling and raising standards of your opponents, and input the pre-flop action? For instance, in this case if the EP guy raised, I would say the A9 was a clear fold pre flop and just as clear post flop. If he limped, then did the A9 limp as well or raise? Is the A9 in a blind or in position? With an EP limper this hand is far from the nearly 4-1 dog you project, and most of the time I don't think the hand would go to a showdown in real play.

Just an FYI, there is an open source code base on Sourceforge I believe for this type of analysis. I haven't looked at it yet, but one of the Cardplayer columnists uses it and gives its location.

[AKQJ10]

[ QUOTE ]

Just an FYI, there is an open source code base on Sourceforge I believe for this type of analysis. I haven't looked at it yet, but one of the Cardplayer columnists uses it and gives its location.

[/ QUOTE ]

Is this it?: http://sourceforge.net/projects/pokersource/

Instead of digging around Cardplayer.com trying to find this reference, i figured it was easier to dig around the search results from SourceForge. There are plenty of poker-related projects but most of them seem to be front-ends to play over the Net.

[fnord_too]

I think that is it... I just dug around cardplayer to find it, and here is what I got.

http://pokersource.sourceforge.net/
And
www.seriouspoker.com

The latter being the columnist's site, where you can get all the code he writes for his analyses.

[droidboy]

[ QUOTE ]
You have A [img]/images/graemlins/spade.gif[/img] 9 [img]/images/graemlins/diamond.gif[/img] and the flop is 10 [img]/images/graemlins/diamond.gif[/img] 6 [img]/images/graemlins/diamond.gif[/img] K [img]/images/graemlins/club.gif[/img] . You are playing heads-up on the flop with a guy in early position. Of course, some of these programs don't take this into consideration. One program gives you a win % of 49.8 % and your opponent 47.4 % .

My calculation for this hand gave a win % of 19% and opponent 75.8 % which I believe is more accurate.

What do you think ? It takes 20 minutes for my program to give an accurate answer and that's on a 2 ghz computer.

[/ QUOTE ]

It really depends on exactly what you are calculating. The equity of As8d in this spot is 50.19% versus a random hand. That looks like what the "one program" is trying to calculate. Equity is very different from the chance of winning. It's important that anyone who uses software understands what is being calculated, and how to use that information. As you imply, just having the raw numbers doesn't necessarily give you the answer.

As for spending 20 minutes running your software before you get a convergent answer, that is a bit too long. What exactly is it that you are computing? For what it's worth, my software gives the A9o about 20% equity. This is the equity from a specific distribution of hands (66+,A7s+,KTs+,76s+, KJo+,ATo+). Again, that's not the chance of winning, but it does seem closer to what you are calculating.

- Andrew

www.pokerstove.com

[theredpill]

Oh, I see what you mean. Yes, I believe I am computing equity. I did get 19 % to your 20% so maybe our programs are similar. How fast is yours? I'm not done with my program and many things that I wrote were done without thinking of the best way. I will go back and fix some things to make it faster. I was basically computing what a A [img]/images/graemlins/spade.gif[/img] 9 [img]/images/graemlins/diamond.gif[/img] would do over the long run against one person in EP who is in on a 10 [img]/images/graemlins/diamond.gif[/img] 6 [img]/images/graemlins/diamond.gif[/img] K [img]/images/graemlins/club.gif[/img] . If you were able to follow a poker player around for 5 years and record the results everytime As 9d with 10d 6d kc on the flop and he always went to a showdown against a guy in EP. I believe he would win 19% of the time.

[droidboy]

[ QUOTE ]
Oh, I see what you mean. Yes, I believe I am computing equity. I did get 19 % to your 20% so maybe our programs are similar. How fast is yours?


[/ QUOTE ]

Oh, about 3000 times faster than yours [img]/images/graemlins/wink.gif[/img]. (see link below).

[ QUOTE ]

If you were able to follow a poker player around for 5 years and record the results everytime As 9d with 10d 6d kc on the flop and he always went to a showdown against a guy in EP. I believe he would win 19% of the time.

[/ QUOTE ]

Maybe this is true. But if you had that information, would you raise, fold or call when the early player bets into you on the flop? Is this "equity" number different if you raise or call? Which is better?

- Andrew

www.pokerstove.com

[theredpill]

Free ? hey that's great. Andrew, your program looks good but how do I use it? I put in player 1's hand and the 3 cards on the flop and hit EVAL but nothing is happening.

[droidboy]

You'll need to specify two hands. Click "Player 2", then click the "Two Card" tab, and select the hand which you think the early position player might hold.

- Andrew

www.pokerstove.com