the auto email says 6 months. so is it 3 months or 6?
also why does outlook express prompt me to install japanese characters when i try to open the email. i clicked cancel and it opens it fine

I'm sorry. It is supposed to be 3 months. We must have sent you an incorrect email. Thank you for sharing that with us!

NP.

I am importing my hands and its taking ages. cpu at 100%. after 3 hrs only 3k hands done. at this rate it will take more than 100hrs!

also there is no progress bar showing hands/min import speed and expected time to complete.

could you pls add them.

also in live pls its takes into acc villains stats like vpip pfr? what is the min no of hands it needs to take this into acc.

what happens if this min is not met. i am assuming it then takes a default optimal line?

How does it handle light 3betting?
If I have 75dd in the Btn vs a CO raise, will it always recommend fold or give a balanced percentage of folding/3b bluff?

2 issues and 1 question

a) the import from subfolders does not work. If I point @ the HEM archive folder for October which has day by day subdirectories it picks up ZERO hands. To import all hands I had to move each exported txt file into a single folder and go from there (or import each subdirectory one at a time)

b) As mentioned the import speed is far too slow. It took several hours yesterday just to pick up 18k hands

Question (in 2 parts)

If I replay a hand and look at probabilites, I see % of possible made hands - 1 pair+ or nothing

1) Does that list of probabilities (which always seems to total 100) include all the action so far rather than just his orginal estimated range?

e.g I have just replayed a hand and my opponent made a PFR; CB and has now fired a 2B. It says there is a probability of 1 pair 54% and nothing 22%. Does that probability take account of all his previous actions and include his turn bet? If so how do you "know" how often THIS opponent would fire again with air/draws; there are some where I would be pretty sure this % was close to zero!

2) Why do the probabilities not include draws? On the flop a guy who raises an oesd with a flush draw does NOT have nothing; he is a favourite against most hands

TY

Trevor

Seems like very impressive work. A bit worrisome though -- what would prevent someone from using this as a bot?

(1) Yes, import takes a long time. This is because we analyze every decision you make (each one takes 1-2 seconds). We will improve speed somewhat, but it will never be immediate as imports that do no analyze are. The best way to do it, is simply to leave it on all night until historical hands are imported. Then (unless you are 10 tabling ), it should import as you play no problem.

(2) Excellent idea about the progress bars during import. We'll try to do something like that.

(3) Yes, it incorporates opponent stats. If it has 0 hands then it assumes you play average. If it has 1 hand, then it assumes you play very very close to average, but slightly including what you did on that one hand. Each hand that's played tells the program more about you and you evolve from average to what you actually are over time (in the mind of the program).

If you 3 bet our AI, it will raise, call, and fold with a set group of hands based on stack size. If for example you raise 40% more than average, it will reraise more often and call more often (i.e., with weaker hands).

This is what we are really want to know, or perhaps some already do. Something like this could have really far-reaching implications.
Perhaps you should come up with a counter-measure to using it as bot.

a) Subfolders: you are right. We are fixing it. Thank you.
b) Speed: We'll improve the speed, but please understand that we are analyzing every decision made in every hand. To analyze a decision well, it takes the PC 1-2 seconds. So the best thing to do is simply let it run all night until historical hands are imported. Then it'll keep up with hands as you play just fine.

Maybe we'll add an option to have the hands analyzed on a remote server with crazy processing power, but this isn't high on our to do list at the moment.

1) Yes, so if you raise preflop it will have a point of view about how you'll play each of 169 hands based on your position, stats, other actions, etc. Now the flop comes and you are first to act. At this point it's basing the probabilities on how his preflop range interacts with the board. So you bet and he raises you. Now you'll see the probabilities change considerably reflecting the fact that he's folding his poor hands most of the time and raising you with his good hands most of the time. Here the program has a point of view about how he plays all 1326 possible hands and based on that updates the probabilities.
2) Good point. Right now the draws are incorproated into the existing categories -- some draws are 1 pair (e.g., bottom pair with 9 flush outs) and most are non-made hands. We should separate them. I'll put that on the list. Thanks!

PS They should always add to 100%, don't you think (other than rounding errors)?

Yes, this is of concern to me as well. Our original plan was to keep the AI on a remote server and do all of the processing there. We were very committed to this for the reason you mention. However, the poker sites said that they would prohibit our software if we sent hand histories (even disguised hand histories) to a remote server for processing. I am guessing they were concerned about us reselling hand histories or something like that which is not allowed. If we wanted hand histories it would be a lot easier to buy them or get them other ways, imo. We made our best (and obvious) arguments, but we were not successful in persuading them.

Accordingly, the only way to do this was to put our AI on the customers PC.

We have gone to very great lengths to keep someone from using our AI as a bot. However, we don't want to share what we did to prohibit this.

We are making great strides in further enhancing the AI to levels that even we thought we could not achieve. However, it is currently our plan to only allow that AI on a remote server (despite the fact that this AI could not be used to provide immediately feedback).

So did the Poker Academy authors, but a really dedicated reverse-engineer could get the code eventually... Just as with software cracking; there is always going to be somebody who will circumvent your protections, but the big difference here is it has the potential to make big $s for them (as opposed to just "cracker kudos").

If your are really serious about blocking it's use, then the best thing would be contact the security departments of the sites and provide them with some "telltale" play-based information that could be used to detect it's use (you as the author should know of some of your own products weaknesses; in this regard to approximations/abstractions used, etc).

Juk

Excellent idea! We will do that!

I agree that it is impossible to make it fool proof on the user's PC, but we did our best. One issue is hacking and another issue is turning the hacked product into a bot. We did our best to prevent both -- but it could still happen. (this is why we tried so hard to avoid it).

Thank you!

Two questions:-

1) the check box that we can see in one of the videos to allow us to see only those hands where HGP disagrees with our play seems to have disappeared from my software

2) In the video we see hands analyzed where HGP indicates say "Fold 75%"
It also shows suggested % for call and Raise. In these cases call is -EV of course (as Fold is zero) but we seem to see Raise as a +EV value.

This seems to indicate that in a vacuum the raise is a more profitable play. If it is making that judgement then it seems to be making it without suggesting a raise size which seems odd because the EV will depend on how much we risk to get the fold.

Am I mis-reading the screen? Is the +EV for raise "true". Does HGP not suggest this play because of becoming unbalanced?

TY

Trevor

There are two sections in the analyze tab (i.e., two sub tabs). One is "Replay" and one is "Reports". In Replay, we just show you all the hands you filter w/ or w/o deviations. I think this is what you are looking at. There's no check box here. In reports, you can filter hands, and then look at specific leaks. In "reports" you can use the check box to specify if you just want to see deviations or if you want to see all hands.

Thank you for the question. I'm not sure of the specific example; however, here's what I think is happening -- which happens frequently.

Let's say for example, you flop a gut shot and someone c bets 2/3rds the pot. Here a raise bluff could be the most profitable thing to do (if you only consider a single hand in a vacuum as you say); however, you can't raise bluff all the time here.

So what frequently happens in reality (and with our program) is that a bluff is most profitable; however, you can't do it very often (or it becomes less profitable).

You are right about raise size. It is using it's normal raise sizing (i.e., a distribution of normal raise sizes) in coming up w/ that EV. So there's not one raise size determining that (but rather a distribution).

What's up with the tiny 2.5x 3bet sizes?



Also in this hand, it advocates that betting is better than checking against an aggrofish (in this 4bet pot). Clearly checking is better, since you will get value from the fish his air, yet it advocates checking only 7% of the time and betting 93%.

It can also use an auto-reload feature, that all regs use in regular games.

Thanks! Looks like we have an issue that's actually causing some min 3bets as well. We'll fix. See next release!

Interesting. Is this based on an approximated GTO, or does it play exploitatively?

1/2 6max game:

UTG raises to 6, CO calls 6, SB calls 5, a 'good tag' (25/20) in the BB squeezes... to... 15??! Seems a little small doesn't it? Same with postflop bets.


It also doesn't advocate 3betting the button? Or anything as a bluff for that matter?

Yes. It mixes up its raise sizing, but I agree this is too small. Thank you!

It'll reraise light when the cutoff steals, but doesn't do too much of it otherwise.

It should do some more. The 3 bet / 4 bet is a weaknesses in the program. We are planning a very big enhancement that will take some time and will make some dramatic enhancements to that part of the program. So we generally aren't doing small quick fixes to the preflop now.

So hang in there with us.

Interesting. I think you'd really enjoy this video here on continuation bets. We imported a ton of hands from two of the world's all time best $200 buy in players (jballer88 and ACHILLES157). I think they would c bet here a large portion of the time.

There's great value in this overall c bet strategy. To employ the strategy well, you need to c bet a large portion of the time. They c bet their strong and strongest hands more than our program.

However, I don't think I can argue that in this particular hand (w/o considering the overall c bet strategy), you might do better checking. The reason I like the check more than normal is that given the stack sizes if we bet half the pot or more, then he can't bluff us.

BTW, you might like to turn on the "heads up display" in the options area.

Yeah I know jballer88. He plays in my games regularly. This is a 4bet pot though, and the SPR is such that you want to check here, since we're never folding. If the aggrofish will very likely autobet his entire range when checked to, this should be a check. If you bet, his air folds. Also by checking you might get Jx or worse to stack off. And if villain checks back, he might pick up equity on the turn that he's now willing to stack off with incorrectly. Obviously in single raised pots our strategy should be different.

Regarding the hud, I gave these players preset preflop ranges so I don't need the hud since their play will statistically strive to match the preset numbers.

The best way I can answer this question is "a little of both". We'd like to be able to play more exploitative; however, doing so opens us up to more exploitation ourselves.

We are working on a more exploitive strategy -- and I think it's going to be awesome! However, we have a long to do list atm.