Poker-AI.org

If have quitted botting more than a year ago so I decided to give you an insight into my AI.

At this page you can past a partypoker cash game hh an see how the bot would have played and what ranges he assumed for the opponents.

It is a GTO based approach. I always consider my hole range. I bet the raise and bet the best and the worst part of my range and assume my opponent is doing more or less the same.
The preflop ranges for hero are more or less hard coded. The range of the opponents are calculated from the average opponent on party poker NL10 2 years ago. Additionally they are adopted in the direction of the preflop stats seen so far of the player. To see this effect you have to choose usernames from active NL10 players from early 2011 :/.

I know there would be a lot of stuff to improve and the representation is not very shiny - nevertheless have a look and play around

http://www.pokerarithmetic.com/analyse_strategy/

Interesting

How, exactly? Do you have any rules for that?

_________________
Бетономешалка! Мешает бетон!

Also one thing is missing - suggested action for the hero, considerig his hole cards.

_________________
Бетономешалка! Мешает бетон!

You have done alot of work here and its a shame your not improving it. Your obviously a very talented programmer.

Sure. The Basic idea is to sort the hole range depending on its current strength on the particular board. Then I compare my range to the range of villain. If my is overall ahead I bet a bigger percentage of the range. The bluff part of the betting range is chosen from the weakest part of the range but preferring some hands that still have a little chance to improve to a strong hand (like a gutshot).
Some more rules handle different stack sizes and dry and wet boards. But the rules are on a very general level - nothing like: If TopPair - TopKicker - then....




Well, you just have to look at the displayed ranges what hero should do with his hand. I mostly have "mixed" strategies like: call20% raise 80%.


Thanks! But to be honest: in the end I was obviously not good enough to make some serious profit from it. Otherwise I would not have quitted.

What was your bb/100 and volume played?

_________________
Бетономешалка! Мешает бетон!

Overall my bots played about 500k hands with a something like -0.4bb/100. I expect the latest version of the strategy to beat NL10 with about 2bb but I do not have a proof since I have only about 30k hands played with it.

My bot setup was not the best. I lost a decent amount due to wrong interaction with the client. I guess my setup would not work with the current party poker client.

Yes but its the little things that make the biggest difference. One thing I did notice was that it was just floating gutshots. It should be floating more then that..like two overs, backdoor str8 and flush, pair and backdoor etc just to name a few. Just overcards are pretty significant as you usually have 6 outs to their pair.

Other problem areas are balance issues. Like the bot should be checking back many value hands X amount of the time to keep its bluff range polarized. It will also keep the villain from betting the turn when the flop goes check check etc. Another problem area is pot control on the turn which makes the bot transparent (and not to mention you lose any possible fold equity). A good villain will pounce on this putting the bot to tougher decisions.

Most bots problems are lack of aggression. I assume you ran the HH files thru PT to check for leaks. Usually they fold the turn too often, not enough barreling and definitely not enough bluffing. So in general the non showdown stats are usually the first place to look. From your stats it looks like your really close so do not give up!

Does really betting top N% and bottom M% are more profitable than just betting top N% or top (N+M)% ?
Ii matters only if opps are adjusting and is this a case on NL10 ?

_________________
Бетономешалка! Мешает бетон!

probably on NL10 a more explotive strategy would be more profitable. I designed the concept with the hope to beat a much higher limit one time.