Review of Poker Academy Pro 2

Kef X-Schecter : November 6th, 2007

Poker Academy Pro 2 is a Texas hold’em program that reasonably simulates the experience of playing in a real online poker room. It simulates limit hold’em and no-limit hold’em, although it is currently missing pot-limit play. It supports both ring games and tournaments. It can run on either Windows or Mac OS.

Note that there is a much cheaper offering, Poker Academy. I expect a lot of what can be said about this program can be said about it as well, but consult this page for a list of features not in Poker Academy.

NOTE: This review applies only to the Windows version. I have not tried the Macintosh version.

The AI
——
The big advantage is you can play against tough competition without hurting your bankroll (aside from the initial price tag of $129). For limit poker, Poker Academy uses Poki [http://www.cs.ualberta.ca/~games/poker/], which is likely the toughest limit hold’em bot that exists today. It also includes Vexbot and Sparbot, two limit hold’em bots specialized for heads-up play. Vexbot doesn’t have a great grasp of basic strategy, but it is extremely adaptive to your play, which means after you play a few hundred hands with it, it will probably give you a beating. Sparbot is the opposite: it doesn’t try to read you at all, but it plays an almost theoretically perfect game of poker, which means about the best you can hope for is to break even. (While it does not play a truly theoretically correct game, it will certainly be more correct than any human can manage.) Neither Vexbot nor Sparbot can play in any situation other than heads-up, however.

For no-limit play, the archetypal player is Xenbot. As one might expect, Xenbot is not as good at no-limit play as Poki is at limit play. That said, it can certainly give average players a run for their money.

There are a few more bots to play against than these, for variety. These ones mentioned above are just the strongest. For limit play, there is Jagbot, a simple basic-strategy player with a complete inability to read others, and Simbot, which mentally simulates the outcome of the different possibilities in order to decide a move, somewhat like a chess program. For no-limit play, there is Jambot, which uses “The System” from Sklansky’s Tournament Poker for Advanced Players, and Oddbot, which deliberately makes weird plays from time to time. Most of the bots have some kind of configuration options, so it is possible to have a table full of Poki bots all playing differently.

The advisor
———–
Poker Academy Pro’s advisor is adequate. One nice feature is that it gives you a suggestion for varying your play. For instance, you may encounter a situation where you should either raise or fold. If the advisor thinks you should fold 3/4 of the time and fold 1/4 of the time, it will show you a little pie chart, 3/4 green and 1/4 red. If you wish to follow this suggestion, you can simply glance at your wristwatch. Look at the second hand. If the second hand is less than 45, fold, and if it’s above 45, raise.

Unfortunately, in a no-limit game, the advisor will not suggest how much to bet or raise. You’re left to guess on your own. Another inadequacy in both limit and no-limit is that the advisor does not expose its reasoning to you, so it is difficult to figure out why it suggests what it does. If it tells you to check, it will not tell you if you are checking with the intention of raising, calling, or folding if it is bet back to you. If it tells you to bet, it does not tell you if it’s for value or as a bluff or semi-bluff. It also won’t tell you if you’re checking because the board is showing a three-straight, or because you’re out of position without the lead. Because you are not exposed to its reasoning, you can only learn by trial and error, and by example. This is still far better than learning without the help of an advisor, though, because trial and error is a very bad way to learn poker if all the information you have is your results.

Fortunately, there are other tutorial displays to help. Above the advisor is a hand evaluator, providing an estimate of the strength of your hand, the pot odds, and potential for improvement. There is a more full-featured hand evaluator window that shows how it comes up with an estimation of hand strength. It shows you that the strength of your hand is calculated by how likely you are to have the best hand at the moment, through an analysis of your opponents’ likely holdings, and you can adjust the bias of the calculator to adjust for looser or tighter competition.

You also can stack the deck or freeze the button in order to repeat a given situation. If you have trouble playing AK in early position when you miss the flop, or you’re not quite sure how to play overpairs, this is the perfect feature for you.

Tournament play
—————
The program’s tournament play features are probably fine for simulating big multi-table tournaments. It is not great for simulating sit & go tournaments. I created a template to allow the program to play Prima Poker-style sit & go tournaments in order to practice them. At Prima, the blinds double every ten hands. This exponential increase of the blinds means that the tournament will often be over quickly and players will be putting moves on each other before too long, and this almost always happens with real players at Prima. However, the bots are far too conservative! The ten-handed table will still be almost full when the blinds are getting very high. The bots will not take strategic advantage of the bubble, nor will they follow the proper sit & go strategy of ensuring one makes it to third place rather than shooting for first. When I saw the weaknesses of its tournament play, I explored the idea of writing a bot specifically for sit & go tournaments…

Writing bots
————
Poker Academy Pro uses an API called Meerkat which you can use to write your own bots in the Java programming language. It is perfectly adequate for ring game play, but Meerkat does not provide enough information to the bots to play correctly in tournaments. Meerkat provides no facilities whatever for the bots to know how quickly the blinds accelerate or what the payout structure is. This must be one reason why the bots do not play sit & gos very well. It would be possible to design a bot to handle three-handed and heads-up situations differently, so that it will play tight until it’s in the money and suddenly loosen up, but such a bot would not be able to cope with any other payout structure.

But if ring games are your thing, and you’re a programmer, this is a very powerful feature. Make no mistake: if you don’t know how to program, writing a poker bot is probably not a good first project. But if you’re already experienced, either as a professional or a hobbyist, it can be a bit of fun. You can try to come up with a bot that can defeat Poki, or a no-limit bot that plays like Doyle Brunson — or like you! Unless you’ve studied this sort of thing for years, you probably won’t succeed… but you can certainly discover more about the game along the way.

Poker Academy Online
——————–
I had never used this feature before, so I tried it out specially for this review. You can play against other people online for play money. This is different from any ordinary play money site in that the players actually care about playing serious poker (as one would expect from somebody who shells out $129 for a poker program), and there is a rating system to estimate who the good players are. The rating system works like this: winning chips from a high-rated player will give you more points than winning chips from a low-rated player. Also, a large data sample will be worth more points than a small sample. So, to get a high rating, you need to consistently beat high-rated players. It makes sense, and you can get an idea of how good you are without damaging your bankroll.

If there’s a place where poker can be appreciated as a sport rather than mere recreation or a money-making grind, this is probably it. This is the chess club of poker.

Profiles
——–
Profiles provide both a strength and a weakness of the program. A “profile” is basically a screen name the program knows you by. All your hand data is stored in a database tied to the profile, including what the bot opponents know about you. If you have two profiles, “Dick” and “Jane”, the bots will see them as two different players and will react to them differently (if they see the two profiles as having different personalities). The weakness is… exactly the same thing! You see, if you use a feature like “stack the deck”, your opponents will still be observing your play, and they will draw incorrect conclusions because they’re only seeing you playing with certain kinds of hands or in a certain position. For example, if you are always dealt AA, you will be betting and raising far more often than you would if you were being dealt random cards, and never folding before the flop, so you will look far more loose and aggressive to the bots than you really are. The solution is to use two profiles: one for normal play against the bots, and another for toying around. The developers are aware of this problem and there may be an alternate solution eventually, but right now it will have to do. It’s good enough as long as you are aware of the problem.

I mentioned that all your hands go into a database. You can analyze all hands, online and offline, as you would in PokerTracker (and in fact there is a separate program you can use to import PAP2 hands into PokerTracker; see the PokerTracker forums). You can’t break the data down in quite as many ways, but it’s still the same idea. There is also a feature called the “Luckometer” that allows you to compare your run of cards to your performance, so you can get an estimate as to whether your current bad run is due to bad luck or a lack of skill. Although it can take a while to build up a large enough data sample, this is very helpful information to have.

By the way, sometimes you might want the bots to forget what they’ve learned about you. Maybe you’ve become a completely different kind of player, or you just want to start over with a clean slate. To do this, you have to find a certain folder on your hard drive. This will usually be “C:\Documents and Settings\[your name]\Application Data\PokerAcademyPro2″ (regardless of where you installed the program itself). WARNING! Don’t try to delete this folder or anything in it if you don’t know what you’re doing, since if you delete something unrelated to Poker Academy Pro, bad things can happen to your computer. If you just want to delete your opponents’ models of you, then you delete all the files in the logs\models subfolder. If you want to delete your hand history database and everything, you will need to delete the entire PokerAcademyPro2 folder. I believe uninstalling the program will not delete this data for you.

The user interface
——————
The user interface is simple, pretty, and not unpleasant to use. It looks very much like a typical online poker room. In my opinion, it even looks better than some of the sites I play at regularly. There are no avatars, but the chips, tables, and user interface elements are well drawn. There is a four-color deck option, and you can choose actions in advance just as you can in an online poker room. (Even better, the computer opponents can’t pick up tells when you use them!) Like Turbo Texas Hold’em, the program features a “zip to the end” feature to end the current hand and fast-forward to the end, although this feature is not nearly as fast as it is in Turbo Texas Hold’em, and it may still take a second for the hand to complete, but if that is the price for stronger opponents, that’s fine with me.

Summary
——-
The Good:
* No-limit hold’em!
* Strong AI, especially for limit play
* Poker Tracker-like analysis
* Luckometer
* Rated online play

The Bad:
* AI not good/realistic enough for sit & go tournaments
* No pot-limit play
* Advisor does not expose reasoning enough

AI: 10/10
Among the toughest in the world.

Features: 8/10
All the essentials are there. It would be nice to have some more, though. The online play is a very interesting feature.

Graphics: 9/10
Looks pretty nice! Easy on the eyes, and not boring. No avatars, but that’s OK.

Sound: 10/10
Typical noises, but they sound exactly as they should.

User interface: 9/10
Nothing is counterintuitive.

Overall (not an average): 9/10
A lot of effort went into this, and it shows. It’s a good and useful program, too.

VERDICT: Buy it!

No tags for this post.