Champions Declared In AI Poker Tournament 54
the_newsbeagle writes "The annual computer poker competition has just wrapped up, in which artificial intelligences battled each other over the (virtual) Texas Hold 'Em table. A researcher who worked on one of the top programs, the University of Alberta's "Hyperborean" program, has blogged about this year's competition and entrants for IEEE Spectrum. His first post explains the rules of the game and why it's tougher for a computer to win at poker than at chess; his second post describes Hyperborean's strategies, and the third gives the results and takes stock of Hyperborean's performance."
Mindgames (Score:2, Insightful)
According to the human players, poker is largely about mind games.
In AI poker, the competitors should be able to send files to each other, or somehow exchange non-game information.
Re: (Score:3, Interesting)
According to the human players, poker is largely about mind games.
In AI poker, the competitors should be able to send files to each other, or somehow exchange non-game information.
The object is to hide you "tells" and bluffs. Maybe the programs should try and hack each other while they play. The program with the weakest security exposes his cards.
Re: (Score:1)
I seem to remember reading a good roshambo AI competition where things like that happened.
Re:Mindgames (Score:5, Interesting)
Hiding tells and bluffs are nearly impossible for even skilled players. The real trick is to show a "tell" or "bluff" when you aren't. But that means you must know yourself as well as anything else.
I play poker, I have a huge number of "tells". Take video of yourself playing poker and learn your "tells" and figure out a way to mimic those tells at will. Natural Tells are really really hard to fake, and take much practice to imitate. When you can imitate your own tells at will (accurately), then you'll have also mastered your tells. I can't fully hide my tells, but I can fairly accurately represent them when I want.
The greatest skill of poker is remembering everything you have ever seen. Knowing how players play (tendencies) over time also helps. Knowing their personality also can expose when their tells are real verses "practiced" ones.
Lastly knowing the Odds is hugely beneficial. As is Gaming Theory.
I am the greatest (Score:5, Funny)
By these criteria, I'm the greatest poker player alive. I have no "tells" of any kind. When I look at my cards, I haven't the faintest idea whether the hand is good or bad, so I couldn't tell you even if I tried. With my 10 luck stat and a silver dollar, I am unbeatable.
Re:Mindgames (Score:4, Interesting)
There is no "real trick" to poker, although there are certainly tricks that will work against specific types of opponents. To be a good poker player you have to acquire a lot of very different skills and knowledge some of which you may already have a talent for and most of which you must learn. You need to know about pot odds, expected value, stack sizes, position, outs, hand ranges, odds against making a drawing hand, and a bunch of other very basic ideas just to get started. If you have a talent for math and probability you will be that much farther ahead, but you don't have to be a maths wiz to be a poker pro. I would recommend Dan Harrington's "Harrington on Cash Games" for a beginner. The strategies for tournament and cash poker (i'm assuming we're talking about No Limit Hold'em) are very different. If you want to play tournaments you'd want to get Harrington's "Harrington on Hold'em" series of books.
Body language and tells become important when you are playing against live opponents, but you still have to be able to figure out what the observed tells mean, and whether they are genuine. You can read Navarro's "What Every Body is Saying" for a good review of body language stuff. After that you can start working on controlling your emotions at the table so you don't "go on tilt" and lose all your chips.
The math you need to play poker isn't difficult you just have to practice doing it in your head, and learn shortcuts so you're not staring off into outer space trying to solve complicated math problems at the poker table. Being able to easily memorize things would be helpful, but you don't have to be a savant.
The math isn't going to do you a lot of good if you can't employ logical reasoning skills. The most valuable piece of information in the game of poker is the knowledge of the cards your opponent holds. The better you are at determining what cards your opponent holds the better decisions you can make. To accomplish this you need two things:
1. information
2. logical reasoning skills
You take the information you have (the way your opponent has played the hand, any physical tells you might think you've picked up on, the way you think your opponent generally plays, etc.) and you apply logical reasoning to sort through the information and determine why your opponent is behaving the way he is and what cards he might have. Usually your opponent will be on what poker player's call a "level". A level just describes how a particular player thinks about and plays poker. Once you know what level your opponent is thinking on you will have a good idea of how he plays and you can better predict what range of hands he might hold, and exploit him. Likewise, if an opponent figures out how you think you should be prepared to adapt.
Re: (Score:3)
While all of Harrington's books are excellent, they aren't the best place for a real beginner to start. They assume the reader has pretty good knowledge of the game to begin with.
I'd start with Lee Jones' "Winning Low-Limit Hold'em" or something along those lines, then move on to Harrington once you're familiar with the basics of the game.
Re: (Score:2)
I'd say that the "harrington on holdem" series should be a must-read for any player of any style of poker. Dan (and his co-author) has the rare combination of an insight into the game combined with an ability to explain that insight in a simple manner.
I'm not that huge a fan of "harrington on cash games", since a good portion of the book (or at least the portion I was slogging through when I put it down) is based on the idea that you're playing online for real money at one of the major sites using a heads u
Re: (Score:2)
To be pedantic for a minute, regularly enough presenting a false tell is functionally the same as hiding your tells. A tell only works against you if it gives your opponent a consistent edge. Each time you can throw out a fake one, that's diluting how much information you're leaking with the real ones. I don't think there's really a line between "can't tell if it's real" and "hidden" here.
Re: (Score:2)
Re: (Score:2)
We already have Core Wars [wikipedia.org] if that's the sort of game you want to play.
Re: (Score:2)
And the biggest mindgame is telling you that poker is about watching your opponents' faces instead of concentrating on your strategy.
Re: (Score:1)
> In AI poker, the competitors should be able to send files to each other, or somehow exchange non-game information.
Why? Every AI would know that every other AI was simply playing mind games with them, so such information should be ignored.
Re:Mindgames (Score:5, Insightful)
According to the human players, poker is largely about mind games.
Really? The pros I have heard have spoken about probabilities, trying to determine your opponents' strategies (are they betting conservatively? do they bet on weaker hands more frequently than expected?), and measuring expected returns quickly. Tells and psychology seem to be a small part of their strategy, and unsurprisingly, professional poker players defeat the AI players despite the fact that computers have no psychology to play against.
Re:Mindgames (Score:5, Interesting)
There's certainly some value to understanding tells:
Does your opponent look back at their hole cards when a middling card hits the board? Do they reflexively look at their chips when the flop comes out? Do they make a bet and give a speech about why they're betting? Are they quiet when they weren't before? Are they sitting upright when they were slouched before? All of these mean things - and they all mean different things coming from different players of different experience and skill...
....but none of those things are as important as understanding bet sizing and hand ranges and putting that together with a player's history. The glint in your opponent's eyes isn't nearly as important as the fact that he re-raised out of the small blind when defending against a raise from the cutoff. What hands do this and why? How often does he do this? What does that say about his cards -- or what does it at least say about what cards he's representing. What does it mean when the early limper gets raised and then HE re-raises?
Sure, players lie, but bets speak WAY louder than tells.
Side note: A number of top game pros take beta blockers. Most of their edge comes from understanding the game, but they'll do everything they can do protect the remaining part of the puzzle that is tells -- especially the impossible to prevent ones that adrenaline rushes cause (and beta blockers prevent).
Re: (Score:1)
Can't tell if he's a computer... Or just bluffing with that typeface.
Re: (Score:2)
i'm going to go out on a limb and assume that what GGP means was that poker "mind games" were refering to the following:
1. know what your playstyle and betting patterns represent
2. know that your opponents are analyzing your play and betting for trends and information
3. purposely varying your playstyle and betting trends in order to disguise your hands
It works under the assumption that you are already doing all of that yourself, and trying to keep your opponents from doing the same to you.
do I raise 3x big
Re:Mindgames (Score:5, Interesting)
Depending on tells is what people who watched Quantum of Solace too many times do. Good poker players base their play on a) a solid game themselves and b) determining their opponents' play style. Does the guy usually check a raise? Does he bet the flop or pocket pairs or a pocket ace more than he should?
Most professional poker players these days play online most of the time, where tells are almost nonexistent.
Re: (Score:1)
According to the human players, poker is largely about mind games.
Really? The pros I have heard have spoken about probabilities, trying to determine your opponents' strategies (are they betting conservatively? do they bet on weaker hands more frequently than expected?), and measuring expected returns quickly. Tells and psychology seem to be a small part of their strategy, and unsurprisingly, professional poker players defeat the AI players despite the fact that computers have no psychology to play against.
"Trying to determine your opponents' strategies" is a portion of that mind game. The fact that professional human beings continually smoke current AIs is an strong indicator that psychological ploys are at work. If the game was all about probabilities then the computers would destroy human beings every time. As you say, the computers have no psychology to play against, they are generally predictable, and do not have the algorithms to judge character, intent, motivation (aka the psychological mind games) tha
Re: (Score:2)
(computers) are generally predictable, and do not have the algorithms to judge character, intent, motivation (aka the psychological mind games) that would let them in turn predict the human's actions.
Well, I actually took a course on poker AI, and computers can learn opponents' strategies to at least some degree. There has been some successful work on adaptive strategies that adjust the computer's probability of raising/calling/folding in a given situation based on how an opponent behaved in similar situations.
Typically, a computer poker (Texas Hold'em) player will be organized like this (but other approaches have been studied; this is just what I remember from the Alberta group's research, which
let's play global thermonuclear war (Score:1)
let's play global thermonuclear war
Re: (Score:1)
No let's play global thermonuclear war
I want a half-CPU and half-EGO based competiton... (Score:2)
Re: (Score:2)
Re:I want a half-CPU and half-EGO based competiton (Score:4, Interesting)
They were, by large, profitable at low limits playing high volume against mediocre players.
If you'd like to play head's up poker against a computer, it's available at many casinos:
http://www.slotmachinesforum.com/showthread.php?5792-Texas-Hold-em-Heads-Up-Poker-(IGT
How long were the games? (Score:2)
I couldn't find it anywhere in the article or links, but how long do these games take to play? Is there a time limit per player decision?
I can imagine scenarios where the whole game takes mere microseconds to complete, or (because I remember the old "start compile and go get some coffee" days) games that take hours.
Anyone find out that information?
Re: (Score:1)
It's in the first article. Seven seconds per action.
Limit should be solvable (Score:5, Interesting)
No limit is difficult to solve vs players who change their gears a lot. Players that don't change gears, you just gotta see their play style and play counter to them. However if I was just going to write a program who played NL, I'd start by just playing my own cards and being conservative. I believe the conservative NL player can still win online.
Poker is a very complex game, but appears forgiving on the outlook because even bad players win occasionally. The trick is to be profitable. And even when you're profitable, you want to keep becoming more skillful because it helps your short term and long term profits. Anyone who hasn't tried this game, I recommend you play in freerolls(never use your own money when you have less than 1500 hours of experience), and work your way up. It is like a long and difficult RPG.
I'm quite good at Texas Holdem myself and I'm profitable over the long run. I've played about 4,000 hours though. If you go with the motto: Never risk any of my own money, and my Holdem Bankroll is separate from my liferoll, you can play. But if you don't respect your money and treat it like gambling, you could destroy your own life.
Re: (Score:3)
Make a few mistakes, and you're back farming all of your equipment again
Re: (Score:3)
Re: (Score:2)
"Solved" means you can program a computer to be provably unbeatable. Any decent programmer should be able to write a chess or poker or go program that will beat most human players most of the time. That's not solved.
Poker, of any form, is impossible to solve because it has an element of luck. I suppose you could theoretically "soft solve" it by writing a program that is statistically unbeatable though.
Re: (Score:1)
Not quite - 'solved' here means the program is provably optimal, not unbeatable. It's trivial to conceive of a scenario (say a specific mid or late game chess board) where even if the optimal move is taken every turn the program still loses. Similarly, a poker program which always makes the optimal choice could still lose by chance, it simply has the highest possible asymptotic probability of winning. If such a program existed, poker would in fact be a solved game.
Re: (Score:2)
http://en.wikipedia.org/wiki/Solved_game [wikipedia.org]
No. You're correct, I should have been more specific and said "unbeatable given a choice of starting positions" but for a game to be solved, at the very least, you have to be able to say, from the beginning of the game, whether two perfect players are going to win, lose or draw. You cannot do that with poker, so it cannot be solved.
Re: (Score:3)
Conversative no-limit is only profitable given the existence of players who lose money against that strategy. I do agree that many on-line players play no-limit in a way that does well again.
The fact that you're profitable is based on the same thing: being more skilled in some way than the other players who you happen to be playing against. After a few thousand hours of play myself, I concluded that by far the most valuable skill for online play is being able to read whether there are players I can make
Re: (Score:2)
So coming back to the age-old question (Score:2)
Is there non-anecdotal evidence for poker being (predominately) a game of luck or a game skill?
Re: (Score:2)
Who are you playing against? With people at very different skill levels, for non-trivial length games, poker is undoubtedly dominated by skill. There are enough hands in most games that the luck factor is fairly diluted. Between players of similar skill, luck is much more important.
Re:So coming back to the age-old question (Score:4, Insightful)
Texas Hold'Em is a game of statistics. In any short-term run, luck might triumph over skill. But if you play long enough, there are a variety of strategies that consistently prove to be better than naive play. The simplest one to model is deciding whether to go all-in pre-flop. That's straightforward enough that papers like Universal statistical properties of poker tournaments [arxiv.org] have worked on distilling it down to a simple function.
My favorite example of odds-based play involves completing a flush. If the 3-card flop comes out, and you have 4 cards to a flush, the chance you will complete that flush is 35%. Many new players think "I have a 1/4 chance of getting a card of any one suit each time, so the odds I'll finish this flush are 50/50". That's wrong; it doesn't take into account that you already have 4 of the 13 cards in the suit. You have to play a fairly large number of hands to distinguish that the odds are really closer to 1/3 than 1/2 though. That's why someone who is betting based on an incorrect assessment of odds will bleed money over time to someone who bets appropriately, the edge of skills over luck here. It is a long-term edge though, and luck dominates the short-term game.
David Sklansky's writing is a good starting place filled with statistics based poker observations. The Theory of Poker [amazon.com] is the standard text on odds-based play. Sklanky's career training was as an actuary, which is one reason his numberic analysis of the game is so strong.
Re: (Score:1)
Re: (Score:2)
My favorite example of odds-based play involves completing a flush. If the 3-card flop comes out, and you have 4 cards to a flush, the chance you will complete that flush is 35%. Many new players think "I have a 1/4 chance of getting a card of any one suit each time, so the odds I'll finish this flush are 50/50". That's wrong; it doesn't take into account that you already have 4 of the 13 cards in the suit. You have to play a fairly large number of hands to distinguish that the odds are really closer to 1/3 than 1/2 though. That's why someone who is betting based on an incorrect assessment of odds will bleed money over time to someone who bets appropriately, the edge of skills over luck here. It is a long-term edge though, and luck dominates the short-term game.
That seems like a pretty basic mistake to make. Even instinctively, I wouldn't think I had a 50/50 chance of getting a flush in that situation.
Re: (Score:3)
It varies with the specific poker game, but I've read estimates that Texas Hold'em is about 60% skill, 40% luck. That's considered a nearly ideal split for skilled players. Any more tilt to the luck side and there's no percentage in it for the good player. Any more tilt to the skill side and unskilled players seldom win and don't find the game fun. Ideally, the unskilled players will donate most of the time, but will win just often enough to keep them interested. As they say, you can shear a sheep many time
Re: (Score:2)
Is there non-anecdotal evidence for poker being (predominately) a game of luck or a game skill?
It depends on how good you are. If (like me) you have little knowledge beyond the basics and what you've picked up from watching TV, and you are playing against people of a similar level, then luck is very important. I play with people I know for pennies, so I don't care. It's just fun.
If you're a top pro, you'll always beat me. Although there is always some element of luck involved, this will become irrelevant over a long enough period of play. Clearly, if you put the best player in the world against m
Non-standard bet sizes (Score:1)
Jonathan Schaeffer, Checkers and Hyperborean (Score:2)