The Future of Poker

Egyéb | olvasási idő
2017. június 16.
It’s official, AI has been created that is undoubtedly better than the world’s best humans at No Limit Texas Hold ‘Em, a game with both luck and psychological components that should have given the humans a chance. We take a look at the implications of such a stunning loss in this article as it represents the first time humans had been bested at this poker variant.
The key element leading to the success of poker over the course of centuries is the fact that, while luck can occasionally come into play, it is a game of skill in the long run. It is true even in the case of online poker, which arguably has more of a focus on understanding the math of the game than even the live game in which there is more of a psychological component. A relatively amateur player may win a few hands, perhaps even a tournament, but will be a loser in the long run if he or she plays against players that are superior to him or her.

With online poker, there is also the concern of straight up getting cheated. While rare, there have been cases of individuals (sometimes even in ownership of the site in question) that have taken advantage of hole card information. There effectively ceases to be a skill component in such games because, in the long run, you have no chance if your opponent knows what your hole cards are. You have absolutely no chance if there are no further cards to come out, your opponent already knows whether or not he has the superior hand.

The closest thing that a player could come to know someone else’s cards is to play a computer that plays the game near perfectly. Such a computer, of course, would understand the math behind every decision and be able to instantly compute the effect of removal of certain cards to the hundreds or thousands of a decimal point understanding its effect on the probability of drawing certain hands.

For whatever reason, humans have set about trying to create computer programs and machines incapable of losing to us at a myriad of games. One of the first endeavors of this regard was an attempt to create a computer that plays chess perfectly. Other games have included checkers and tic-tac- toe. One of the primary differences in these types of games is the fact that there is no luck component whatsoever and very little psychological component. Granted, fatigue can become a factor in chess, as we see in the course of many professional lengthy matches, but not so much psychology.

Furthermore, there is no way to bluff in any of those games. Granted, you can occasionally make a move that is a gambit in chess, hoping for your opponent to make a mistake (perhaps by being too aggressive) but you either have a favorable position from one move to the next, or you do not.

The fact is that the two are totally different games whereas one is entirely skill and the other is mostly skill with a luck component as well as a psychological component as being an aspect of having skill at the game. That is particularly the case in No Limit Texas Hold ‘Em due to the ability for a player to have a wide range of betting options (the minimum at the level of play all the way to an all-in) as opposed to Limit Texas Hold ‘Em in which the amounts that may be bet are fixed. For that reason, Limit Texas Hold ‘Em, as far as skill goes, is not only an arguably easier game but also has more of a focus on the pure mathematics behind each decision. To wit, there are very few opportunities for a bluff in a limit game.

The result is that No Limit Texas Hold ‘Em, as compared to most other games, should have much more of a “Human component,” to it. While probabilities and a solid mathematical understanding will get you pretty far along, other skills are required for a player to have a complete all-around game.

Either way, several programmers, often professors/students have made several attempts to create a machine that is simply unbeatable at No Limit Texas Hold ‘Em. While there have been a few failures in the past, just this year, they came up with a successful device that wiped the floor with a handful of the world’s best players.

Even if there were to be a computer that is mathematically superior, we would think that some of the world’s most skilled players would be able to compete against it somewhat meaningfully. Think about it this way: As a below average chess player, if I were to go up against someone such as Nigel Short, you might as well call the game, “I lose.” I’m not going to beat him with luck or psychology, in a hundred games, I would lose one hundred times.

On the other hand, I could theoretically beat someone such as Daniel Negreanu in a heads up match just by way of pure luck.  I probably wouldn’t, but it’s possible.  I could, even perhaps with two out of three, four out of seven, but if you were to push it to the best of 101, his vastly superior skill would definitely win out in the end.  

Thus, the computer winning by a few games wouldn’t necessarily be meaningful, but if people were to get destroyed by it, then it would have meaning.  

Earlier this year, Carnegie Mellon University School of Computer Science brought the fight to Rivers Casino Pittsburgh against four poker professionals.  One of the opponents is an online specialist, Dong Kim, known as, “Donger Kim,” is a specialist in online No Limit Texas Hold ‘Em, so the game being played (120,000 hands total against four players) was right up his alley.  

Jason Les may be a more well-known name to poker enthusiasts as he has cashed in at a few of the more major live poker No Limit Texas Hold ‘Em poker tournaments that have taken place over the years.  Unlike Kim, he doesn’t necessarily specialize in online play, but he has the mathematics of the game down cold and is certainly one of the top one or two hundred players in the world.

Jimmy Chou (also known as Jimmy Zhou) is another Heads-Up NLHE specialist who placed first in Macau at the APPT 2015 Asia Championship of Poker and gave it a respectable run the following year placing fourth in his defense of the title.  The two cashes, after conversion to USD, were worth approximately a million bucks combined by themselves.  He’s also known as a strong cash game or sit-and- go player.

Finally, Daniel McAulay focuses on cash games and has not participated in as many tournaments as the others.  Even with that, he did place seventh in 2016 $1,500 No Limit Hold‘Em Shootout, so he’s no slouch by any means.

Liberatus, to win, had to play a much better game than the previous machine designed to beat these players.  In the last effort, almost two years prior, the humans won by an amount that was almost statistically significant.  There’s not much need for any drama in the results in this case because the human competition was mercilessly destroyed.  

Ultimately, Liberatus would have had to have won by at least $7.70 per hand for the result to be deemed, statistically significant, and it almost doubled this requirement winning by a total of $14.70 per hand and trouncing its human competition by a total of $1,766,250.  In short, it wasn’t even a competition.  

What’s particularly alarming about the result is that such poker professionals were handled with relative ease!  Just ask yourself: What would happen if Liberatus played you?  Given that a few of the best professionals were handled with such ease, it stands to reason that Liberatus woul mop the floor with an amateur player.  Perhaps more worrisome is the fact that a less skilled artificial intelligence would be able to trounce an average human competitor with ease still.  

While it will be a while, certainly someone with just a little computer know-how, in terms of programming, could create an AI that can be fed hand data and the bets put in which would then be able to develop a strategy to beat human opponents in an actual online game.  If that happens, then a number of countermeasures would have to be undertaken to ensure a fair (human v. human) game.  The poker sites might have to develop or purchase AI of their own in order to analyze the play of human players to determine whether or not the person is performing, “Impossibly well,” against the competition.  Of course, there would be little incentive for this additional expense other than the fact that humans getting trounced every time they sit down would be much less likely to make deposits in the future.  

One interesting thing that could happen is to have AI v. AI games in which there are tables specifically for people who have created an artificially intelligent system to do the playing.  I would almost be the poker equivalent of, Robot Wars, so such games would certainly be fun to watch and, of course, could be played with actual money.  

If nothing else, such systems could probably be created to analyze an individual’s play after the fact.  That would be much in the same fashion that chess programs critique a human player’s move after the game has been completed to let the person know when he or she has made a mistake or blundered.  More rudimentary forms of such things have been discussed here.

One way or another, it’s only a matter of time before computers have an even more meaningful impact on the online game.  

Kapcsolódó cikkek

0 hozzászólás