looks cool.

any1 gave it a shot yet? If u can get it for merge ill def test it out

Just downloaded it... imported some hands and viewed my play quickly... looks fantastic and the fact that you are offering it to p2p'ers for the extended free trial is fantastic. Watching the ALCS right now, will play tomorrow and experiment more.

not being able to import from subfolders is a problem

concerning the "totally revolutunary!" and "quantum leap": the math of calculating the optimal betting/folding/raising % for hero is well known and already used in other products -- for instance in poker-academy. nevertheless it can be usefull for players & i'm absolutely sure people will pay their good money for it. (i mean, they even pay $ for a table-layout with their favourite toon-characters! and this product is called "the holy grail!" -- so its GOTTA be awesoooome!!!!)
however, what the more interesting part to include would be, is taking opponents' behaviour into account.

to what degree does your product based his results on villain's behaviour? or it just assuming & giving its suggestions based uppon the assumption, that villain is playing an optimal strategy? is it even basing it on players' tendencies of e.g. "bluffing wet boards in multiway pots more often" of showing specific lines, like "limp/raising for player 'br4ini4c19422' means KK or AA 88.2% of the time, so with only 11xpotsize behind, you should fold your 99 here 100% of the time"? until then, call it "the somewhat blessed grail".

Thank you for your post!

True Poker Academy offers advice, but I think that's where the similarities end. Poker Academy allows one to plug a program (like ours) into it and play against their most sophisticated AI. We did this for over 400,000 hands and literally won by 17 big blinds / 100 hands (i.e., about double the rate that the worlds best poker players beat their competition). Said differently, if our AI were a human it would make a nice living as a low stakes pro. If PA's AI were human it would lose money at .05/.10 online.

Our AI adjusts to almost all factors including those below:
- preflop raise%
- VP$IP
- Steal %
- Steal defense %
- C bet %
- Stack sizes
- Vulnerability of board
- Outs
- Scare cards
- and call, fold, check, bet, raise % on each street

Our AI does not current adjust to the most recent behavior of an opponent. So let's say you raise 10% of the time (for the last 1,000 hands). Then you start raising 50% of the time for the last 20 hands. We treat those 1,000 and those 20 hands the same. Of course we will plug this leak, but to date we have not.

I'm not sure I understand. You can browse to any folder or subfolder to import. Do you mean giving people the option to import from "all subfolders"?

If you select one folder, then it will import from subfolders.

I think we need to update our import video for this, right?

this makes me thinking of an indiana jones table.^^

however, the replayer/client has some really poor graphics. the view is unergonomically, has way to less contrast, positioning of elements is confusing... and so on. as it is the main element of your software you should focus on it.

about the skill level of your AI... well i have played it for 1k hands, chosed the full ring hardcore settings and i have to say it wasn´t as tough as i have expected. i saw a lot of questionable play by your AI especially i think it plays way to soft postflop compared to real live, but this is only an impression as i refuse to play this for additional time as long as it looks so bad.

btw. i found some bugs and would report them but i didn´t found a category for it in your forums.

basically i like the idea of the software. it would be cool to play against an AI which is able to adapt. you guys should think about an import of pt3/hem databases. also you should implement a database for your product instead of generating textfiles.

looks nice. does it take into account opponents' hud stats?

The AI does adapt to opponents (both preflop and postflop play). We will add the ability to import from PT3 and HEM. Our product does have a db (it's where all the data is stored). It's SQLite.

Yes, the AI adjusts to almost all factors including those below:
- preflop raise%
- VP$IP
- Steal %
- Steal defense %
- C bet %
- Stack sizes
- Vulnerability of board
- Outs
- Scare cards
- and call, fold, check, bet, raise % on each street

Our AI does not current adjust to the most recent behavior of an opponent. So let's say you raise 10% of the time (for the last 1,000 hands). Then you start raising 50% of the time for the last 20 hands. We treat those 1,000 and those 20 hands the same. Of course we will plug this leak, but to date we have not.

why doesn´t it store the imported hands as well as the played hands and why is it creating dumps (txt-files) ?

Tinkered a little with the analyze feature and its very impressive TBH. Not sure if I'm happy though because it might kill the micro and low stakes games if and when this gets around.

Hi, I'm one of the developer of this software.

Not sure I understand what you mean. Both imported and played hands are stored in the DB.

The txt files you see are just log files helping us to debug the software. I agree we should probably avoid to generate them for releases.

what i mean is when i close the software and open it again i miss my hands i have played/imported. however, i recognized in the meantime that i have to refresh the list and they will be shown. it would be cool if you could preload the hands.

also needed...
sorting by session and stakes/game
moving of processed histories to a different folder

Played 100-200 hands with it, looks promising.

yes. hem breaks up exported hands into multiple folders, so its a real pain to import them all one at a time if you have a lot of hands.

Yep. Our import currently works as you would like (i.e., choose the macro folder and then it'll import hands from all subfolders); however, we have not updated our tutorial video yet.

for me it only imported if you pointed it at the specific subfolder.

Are your AI models static or dynamic?

I also think that if you have a product like this, the developers should qualify their status as an elite poker player.

The AI is dynamic, adjusting to opponents historical play (preflop and postflop), stack sizes, bet sizes, position, and numerous other factors. We simulate the hand out many hundreds of times to calculate an expected value of each possible action.

I don't understand what it means to "qualify status as an elite poker player", but maybe this addresses your point...

The developers themselves are not elite poker players; however, we've received help from jballer88 and ACHILLES157. While they are not famous poker players, they are two of the all time highest winning $200 buyin players. jballer88 plays 6max and ACHILLES157 plays full ring. You can check them out on PokerTableRatings if you like.

That's more or less what I was asking. It's sort of a rare set of skills that have to come together in order for this to work well, and one of them has to be not only a poker player, but a very very good winning poker player.

Does it coach for HU as well?

Unfortunately, today it's not for HU. We'll add HU, but after we add tournaments. There are just so many more tournament (full ring or 6max) players than HU -- so we need to add that next.

I was never a professional poker player, but it was my primary hobby for five years before starting the project 4-5 years ago. I'm a winning player, but I've never committed to the game full time (other than w/ respect to this project). In my opinion to play at a pro level, one really needs to play full time and I have not had the time or inclination to do that.

You can read some background here.

The funny thing about professional poker players and poker artificial intelligence is that one will fail if they try to develop an "expert system" (i.e., in this situation do this). Such systems obviously lend themselves to using experts, but for poker there are simply too many situations to use such an approach. It works okay for FL and short stack -- that's why most poker AI is of that type.

However, to do it dynamically (ironically) you need less expert advice. Having said that jballer88, ACHILLES157 and another guy who's one of the world's best online cash players (but wishes to keep his name out of it) have been instrumental in the development of the AI.

Very interesting.

Best of luck with it, I'll definitely give it a look.