This is attributable to a multitude of factors and computing power is not one of them. Keep in mind that computer chess is much further ahead in terms of development. Computer poker isn't nearly as far along.

With that said, there are still bots that play a very solid game. These bots are good enough to beat mid stakes NL grinders for a solid winrate (in limit poker, the bots are soul reading demons). Now imagine IBM getting into the fold with a team of high stakes players for input...If they actually put effort towards developing an NLHU super bot, I'd put my money on it against any HS pro.

Ok, now on topic. Chess is much more complex. I feel like anyone who's spent time learning both games would be of this opinion.

Funny how this is the computer age yet so many people are clueless as to how computer programs really work.

Chess computers use algorithms developed by masters. Go to your local library, grab a big book on chess with the subject "play chess" or "play against the masters". Those books are giant algorithms, you can make a move then look in the index of your move and there will be a counter move. You can play a ton of games "against" the book. Computers are little different you can just pack more "books" into them.

Ok, to all the people who say that chess programs are so strong because of the length of development and grandmaster level knowledge incorporated in these programs - this is mostly hogwash.


What I know is this -- although I have never written a chess playing program before (I have written other game-playing programs though), within a week or two, I could personally put together a chess program that uses pretty much no Chess-specific high-level knowledge, and I am absolutely certain that this program would beat me and most amateur chess players (but definitely not high-ranked chess players). This program would use techniques in a field of AI called "Reinforcement Learning" to "learn" how to play chess by playing against itself... The types of things I'd do to make this program stronger (creating an opening book and endgame evaluator and refining the search algorithm are the most important things) are all, relatively speaking, simple tasks.

The thing about a program like this is if I allow it to play on my desktop machine, it will be at one level of strength. If I run it on a much stronger machine, though, it would dramatically increase in playing strength.... So if I want my program to beat the best chess players in the world, all I would have to do is wait for computers to get faster. In 15-20 years from now, writing a grandmaster-level chess playing program will be child's play IMO.


Writing a strong poker playing program, especially deep no limit, on the other hand, is WAY more difficult. I mean, I could put together a decision tree and use all my personal poker playing experience to say "if you are in position X, do Y". After plugging in hundreds or thousands of these statements, this program would get to an ok level. But any decent player would quickly learn how to exploit the hell outta this program. It is impossible to write a program in this fashion without leaving gaping holes... And putting this program on a faster computer would do absolutely nothing for it.

So, say we go back to the ideas we used in that chess-playing program -- search, evaluation, and self/reinforcement-learning.. Well, search is immensely more difficult to program (the search space is enormous when we consider all of our opponent's possible actions and our actions and possibilities for their hole cards and for the board runout). A good evaluation function is also way more difficult to learn - this is mostly because a good search algorithm is so difficult to create... And the only way around these problems is creating different types of shortcuts, all of which leave pretty big holes in this program and make it much more exploitable.

----

Anyway, to sum up what I've said:

In 20 years from now (I dunno, maybe longer, but not much), your average/advanced computer science major will probably have little difficulty in writing a chess-playing program from scratch, in a relatively short amount of time, that runs on a desktop machine and can beat the best chess-players in the world.

There will never be a time when it's a simple task to write a poker playing program that can beat the best poker players in the world.

What does this say about the two games???

Actually, it doesn't quite work that way. Unfortunately.

It turns out that masters are very poor at describing what it is that they do in a way that is usable to programmers. Partly due to much of what they do taking place on an unconscious level. As a result, programs use a very different process to arrive at what move to play. A human player might look at about 1000 positions during an entire game. A computer program will look at millions of positions per second for every single every move. Humans are very selective at what they consider by very many orders of magnitude. Programs take a brute force approach.

Most chess programmers are weak chess players. Playing chess well hasn't been realized as a significant edge for programmers. There are some strong players that have written successful chess programs, but this has never been an important qualification to write a strong chess program. It has happened in the past that a below average player has won a computer chess championship with their program.

All those chess books are great for human players to read, but not as useful as one might suppose for a chess programmer. Chess books are not useless to chess programmers, but they are hardly a blueprint for success in developing a good chess playing program.

The most useful chess books are opening books where programs can simply regurgitate their contents during play for as long as the opponent plays the expected opening moves, but after the opening phase, the program is on its own.

Some other things to add. First multiplayer games are in general much more complex than two-player games.

Also, poker is more a game of exploitation than game-theoretic optimal play. Although chess players may sometimes look for ways to exploit the specific opponent they're playing against (e.g., playing a specific opening vs. a specific opponent, but a different opening vs. others), a chess player is usually just going to make the move they think is strongest. There is no real need to create an "opponent model".

In poker, however, the best players create a model of their opponents -- incorporated in that model, the best players try to reason about how their opponents think, their opponent's specific motivations in the context of the game, the way their opponents reason, their opponent's emotional state, etc. And this is taken even a step further and poker players start to reason about the way their opponents are modeling other people. This type of reasoning, these "leveling wars", are much more complex than anything that happens in chess.

To anyone believe that chess is less complex; do you play chess???

I think you have misunderstood Nash equilibriums, they are based on certain assumptions about the game that must be fulfilled for a Nash equilibrium to exist. One is that there can be no difference in skill level between the participants which I believe everyone will agree there is in poker. Another note is that Nash equilibriums doesn't care about winning strategies, it 'only' care about dominant strategies that is the best strategy you can choose at the given moment. This dominant strategy can (and often are in real life applications) a losing strategy but a less losing strategy then any other chosen.

I may have been a bit rash in my other posting about creating a chess program. I guess to amend my point, what I meant to say is that chess programs are more than just brute force calculation. Yes, brute for calculations are "part" of it, but they are still based on algorithms/decision trees develop by master/grandmaster chess players and that plays a critical role.

Its not like the movie War Games where you write a computer program and the computer just teaches itself to play chess to the point where it can defeat a grandmaster. No, the fundamentals and "backbone" of the algorithms chess programs use come from previous, algoritms, decision trees, and games developed by master and grandmaster chess players.

FTP did

How are you defining "complex"?

One obvious method (mentioned above) would be to count the number of paths in the tree of possibilities for a game.

To see why this is a bad method consider the game of darts as played by perfect throwers (ie: players who never miss what they are aiming for). For a 2-player, first to 501 game, there is a huge tree of possibilities (more than both poker or chess), but nearly all the paths in the tree are dominated.

Juk

As mentioned by pocketzeroes above, it's actually possible to tune these parameters via self-play using "Reinforcement Learning", see here:

http://citeseerx.ist.psu.edu/viewdoc...=10.1.1.113.79

Again as mentioned by pocketzeroes above, it's actually quite possible that somebody with almost no domain knowledge of chess could write a very strong engine (the algorithms required are non-specific to chess and well known). The only domain knowledge required would be the identification of evaluation function features and even that is arguably not needed if you can search deep enough (see: OliThink for an example of a strong engine that uses no domain knowledge).

Juk

Chess is much more complex. Its not even close.

No, these are not properties required for nash equilibriums. (Except that of course a NE for a given situation can be -EV.

It's vaguely true that NE doesn't care about "winning" strategies, in the sense that a NE is a strategy that can not be beaten, but is not necessarily one that will win. Some games have NEs that won't win against non-NE opponents, some do. All that's guaranteed is that it can't be beaten.

This is a very difficult question to answer. Certainly chess players use more processing power per second and have way more time because they have a game of complete information. They have to map out and test each possible decision against one another to use their past experiences and or unconscious learned memory. Whereas poker is more of a diverse game. Involving logic, math, psychology. Certainly I would say pure like memory is much more important in chess. Whereas in poker it's sort of a more creative game in each hand you have to plan ahead, factor in psychology, metagame, math and logic for 3 streets and figure out their ranges, as well as trying to conceptualize how they're thinking about the hand to arrive at the right choice.

Having said that it's clear that the top chess players are better than the top poker players at each game. I'm not sure if it's because the top chess players study and play the game more or if it's simply that their brain structure in some way is like a super high IQ or not even that but chess seems more like a specialized game. It requires super high metal processing powers. Poker seems more like you can be strong in certain areas like logic, or psychology or math and be successfull. Or like a myriad of each. It definitely seems reasonable to think a top chess player could fail in poker, and a top poker player could fail in chess, I just think they involve different skill sets.

I don't play chess seriously only once in awhile I feel that I should throw that in =p.

Chess and its not close