Posted Wed Oct 26, 2005 3:48 pm GMT by tame_deuces
A long post....maybe it is useless...But I needed something to do.
There have been some people asking questions about play money games on the boards lately, and as usual in such cases players have told people to start playing poker for real money instead, as the general agreement is that play money yields little or no experience. The response about various poker games have also gotten similar responses. The general consensus is that if you want to learn poker you need to play for cash, live or online. So personally I have just for fun wondered if some of these new poker bots are all bad, and I especially wish to compare them towards play money and low stakes cash games.
Since I had some study-related trips coming up which included a seeminly endless tirade of long bus rides, I loaded up my IPoD with my recently purchased complete Deep Purple discography, brought extra batteries for my laptop and purchased a poker program (which will remain nameless) which used the famous Poki-bots as adversaries. Ten years in the making, poki is regarded as the height of artificial intelligence poker programs.
The program came equipped with four kinds of bots. An adjustable scripted bot system for limit which plays hands according to a preset ruleset and does not adjust its game based on your play. The adjustable poki-bot which plays limit according to preset rules, but will also vary its play and try to adjust to your playing style. A scripted adjustable no limit bot and a variation of the adjustable poki-bot intended for no limit play.
1.) Goal. To play about 2000 hands of limit against the poki bots in full ring. Play about 1000 full ring game hands against the no-limit bots and about 500 hands of head's up no limit play. And after this try to judge if anything about poker can be learned from playing bots.
2.) Problems. My own limited limit experience is probably the main problem. Limit is the main course of poki, and if I am an under par limit player that will certainly render the results rather dubious. I'm comfortable with my no limit play, but I'm not exactly top of the class there either. The second problem is that the sample size is very small to base any judgement on. And last but not least, will 40 hours of bus rides be enough to decide if 'Stormbringer' is truly the greatest rock album of all time?
I'll just start right up with where I feel most at home.
No Limit Head's up play.
First of all, the scripted bots are rather useless in head's up play, they play far too generic to be of any real resistance to anyone with some head's up experience under his belts. They are easily read and easily beaten. I won't actually hold this against the program, the script bots are mostly there for giving you the ability to simulate the play various systems recommend, and for that they seem to work nicely.
The no-limit bot which tries to vary it's play also got problems head's up. It seems to have immense problem with handling unbridled aggression. On the plus side it actually tries to bluff. Adjusting the bot so it becomes looser partially solves the problem, but this bot is still nothing compared to a somewhat experienced human player.
Grade: 1/6
And then to the subject which I was most curious about. Poker AIs have tremendous troubles with no limit play, because whereas there is a general consensus on what is 'optimal' fixed limit play the amount of variables at play in no limit, especially the human ones, makes it very hard for a computer to play it any good. And no, this is not something I hold against limit or something I say to make no limit seems harder/better/cooler.
No Limit full ring and 6-max
Here the scripted bots are actually quite fun, since you can simulate systems with them you can experiment a little and fool around trying out various plays. You have the same options on the learning bots, but as said before they try to bluff and vary their play based on your actions. Compared to play money with human opponents the bot play is generally of a much higher quality. And the program will teach you the value of starting hands and position. But there is still a significant problem with the bots. Namely that they make errors....but they are not typical errors you find in human players at low stakes poker. So that means if you figure out a winning strategy, you are probably learning stuff that will not work in real life.
The bots also have a tremendous problem with shorthanded play, and the game becomes rather easy as the amount of players go down. Also the game had an advisory function, and I'll just say at once that those advice were very limit oriented and can point a no limit player towards making horrible mistakes.
Still it was pretty fun, and the play was of better quality than play money opponents, but still I will not recommend it to a beginning poker player. Too easy to pick up very, very bad habits. But as I mentioned, scripting bots to play according to systems you find in various books etc. and then just let them play amongst themselves for some thousand hands is pretty fun.
Grade: 3/6 (but not useful for learning no limit poker)
And then to the star of the show for poki.
Limit Full Ring
A bit disappointed by the no limit bots, the limit bots were actually quite shocking. A bit on the loose side since I'm used to playing NL 6-max, I soon found out I had to adjust my hand selection. Some bots almost seemed to have personalities, some tried to steal alot of blinds, some liked checkraising and sandbagging whereas others where more on the laggy side of things. I even found myself silently cursing at a particularily loose passive poki-bot which runner-runnered me 2 times and at other occasions gave other more tight-aggressive all the pot odds they could ever need for calling me down. (Then I realized I was sitting in a bus with my headset on, and probably cursing pretty loud so I shut up).
The program also had some nifty help tools for the player, showing your current pot odds, hand starting strengths etc. Not necessary for experienced poker players, but with this aid anyone should be able to learn the concepts in 10 minutes. On a negative side the game had an 'tutorial' which advocated some very weak-tight myths about how to play limit. The poki-bots seemed to play better poker than the authors so to speak.
The bots seemed like they were programmed with the important basic limit aspects in mind, and they seemed to respond and adjust to aggressive/passive play on my behalf. When I for testing purposes started folding alot preflop, one boot seemed to be raising into my blinds constantly after a short while. And I also got the impression that various plays for trying to extract 'that extra bet' could actually work against them.
I won't claim that these bots were very tough. It seemed like a tight preflop game and the right amount of aggression would atleast guarantee you break even or a half decent winrate. But still the quality of the play was a very positive experience, and I would happily recommend any player to take on these bots as opposed to playing for play money. I personally felt I picked up an important thing or two about hand selection in limit play. Also in limit play the possibility to adjust the bots really, really shines.
On the negative side, the bots had some problems with slow adjustment to a player's play. I suspect a human will pick up on things rather quickly if you raise from EP four times in a row and not give you credit for a muscle hand each time. Such easily guessable reactions won't be found in human players and can easily give a player bad habits.
4/6
The end verdict
The verdict is that nothing will compare to playing for real money for learning poker. I don't think fake money be it play money online or bits in a single player game can compare to putting your own cash on the line. Most importantly, as it is now I would not recommend using poker bots too learn no limit or pot limit in any way, and I wouldn't play for play money either. And don't think playing for play money or using poker bots too learn about limit is a good idea either, play for real money, read up and understand the game. But if you do this, from what I saw, good limit poker bots can be a useful add-on tool. But remember that it is a computer playing, for all its computing power and power to calculate it won't be as tricky as a human player can be. But I certainly think poki can beat the crap out of your typical micro stakes limit players though.
Oh, and ofcourse....'Stormbringer' is defintively the greatest rock album of all time.
Did you know that participating in a poker forum can help you improve your own game? Be it by sharing experiences or simply asking for help, participation in a forum helps you focus and keep 'on topic' which will help you improve your game. You can learn from other players feedback and from their experiences. Why the THP poker forums? We offer one of the best managed texas holdem poker forums available, and the community within is far more friendly than those typicaly found on other sites. We've made a 'lurkers edition' of the poker forum available here on Holdem Poker Online, but we encourage all visitors to register and join in on the conversations on TexasHoldem-Poker.com
Posted Thu Oct 27, 2005 3:07 am GMT by Muck
Cool review tame_deuces 
I was thinking of getting a poker game for my gf (her sister has a couple on her Xbox) to build her interest. She could play on play money tables but she just likes consoles more than PCs.
Posted Thu Oct 27, 2005 5:26 am GMT by tame_deuces
I'm impressed someone actually read it, or for that matter, parts of it. It was a long busride...that's all I can say.
If she just wants to play poker for fun and maybe learn a little without risk, I'd say that a game using poki should do well. Atleast for limit.
Posted Thu Oct 27, 2005 5:35 am GMT by Muck
I read it all but I’m interested in the subject
I don't have anything against FL but I’d prefer my gf to learn NL so that we could play casino tourneys together. I thought a game may get her interested enough in the game to fancy reading a couple of my books.
Every time I read something like this I get the burst of enthusiasm to work on my pokerbot, but after a couple of days designing and coding it starts to ebb away as the shear complexity of the problem squashes my brain under a mountain of “it depends”.
|
|
