this

^

Actually, I think I've stumbled onto a red herring regarding the computer chess program vs computer poker program argument.

Just because you can write a computer program for one thing but not another is not proof one is more complicated than the other.

FOr instance, what is more complicated, doing calculus or reading this image

I can teach a 5 year old to read this image or any image like it, however you could not program a computer to reliably do it.

But you can program a computer to do calculus. So does the fact that you can program a computer to do calculus proof that calculus is "easier" and "less complicated" then reading wavy tilty and offset images???

So basically, my new argument is simply that the lack of NL poker computer program (equivalent in skill to chess programs) is not proof that poker is more complicated than chess.

I can think of a million things that we would think of as simple that we can't write a computer program for.

I think I could write a computer program to solve HU poker situations, though, it just wouldn't finish in my lifetime. But that's beside the point, as you said, the complexity of a problem is not related to how hard it is to write a program to solve it.

This is exactly what I was going to say, so I won't.

It's comparing apples and oranges, really. They involve very different skill sets. There are many ways in which poker is more complex, and vice versa.

Chess is much more creative than poker.

One of the specialties of chess is the composition of endgame studies. These are chess puzzles that are simultaneously challenging, instructive and quite beautiful—like works of art.

Here is a database of chess endgame studies:

http://www.hhdbiv.nl/media/folder.pdf

Poker has nothing comparable to this.

I think you would be hard pressed to be able to do this, even given infinite time, especially in No Limit poker - unless you had a very very large amount of storage space (more than you could possibly have available).


I disagree. Can you think of a better way to objectively measure the complexity of a problem???

If we just allow that "Problem X is not complex if X is easy for a 5 year old child to do" (as in the captcha image recognition example), then we are basically giving NO credit to billions of years of evolution and 5 years of repeated learning experience for having found a solution to a very difficult problem.

The captcha image recognition problems are both complex and difficult. They just *seem* easy to humans because we've got billions of years of evolution (the most creative problem solving effort that any of us are witness to) working for us. Our brain has been fine tuned to perform these types of tasks... Plus each of us has many years of exposure to visual stimuli under our belts, a continual learning experience since the moment we were born... These things have made each of us amazingly versatile in our visual pattern recognition ability... But this doesn't make the problem of solving captcha images a simple task.

If playing chess well (as opposed to responding well to visual stimuli) were one of the primary selective pressures that led to the evolution of humans, we might then see chess as a "simple" problem, and most humans alive today would play chess at a much higher level than any computer can. Instead, it turns out that being good at doing visual pattern recognition was important for our evolution. So solving captchas is easy for humans but difficult for computers. And chess remains difficult for people.

What is easy or difficult for humans is a bad way to measure complexity IMO.

You actually make my argument.

We are talking about complexity as percieved by a person and thus complexity as percieved by the human brain. Your argument is that since the human brain is this highly evolved "thing" that we need to account for that as relates to why a computer can't do what our brain does......

Thus, by that argument you negate the effectiveness of using a computer as a metric as relates to measuring complexity as determined by the human brain.

If the human brain can do Task A, B, and C with minimal effort

and a computer is unable to do Task A, B, and C then you cannot use the computer as a metric to gauge complexity as percieved by the brain.

Basically, if there is such a fundamental difference between the two, you are talking about apples and oranges on a HUGE scale here.

I mean, this is like trying to determine which animal is "faster" a dolphin or a cheetah.

A dolphin's land speed is probably 0.5 mph as it hobbles along but its speed in the water is like 40mph

a cheetah runs at 70mph and probably swims at 2mph

The two are just so fundamentally different that trying to use one to compare against the other or trying to define a metric based on one vs the other is just not possible.

I submit the same thing is "similar" between computers and the human brain as relates to trying to define "complexity".

When I see the word "complexity", I think of something that can be objectively measured, and hence is not subjective, and hence is not related specifically to someone's personal judgment. There are many different kinds of complexity -- in computer science alone, we have space complexity, time complexity, and also things like algorithmic complexity.

There are various other uses of the word in other sciences. But the key point is the word "complexity" tends to be intricately linked to a scientific study. And science is all about *objectively* measuring things. So "complexity", in my mind, should be an objective kind of thing.

So we're asked the question - "What is more complex, chess or poker?"... It is clear that the question is not about the games in and of themselves, but rather about solving the problems necessary to play them well... In my opinion, then, the best way to respond to this question objectively and scientifically is to look at the algorithms that one might develop on a computer to play these games. If we do this, it is strikingly obvious to someone like me who is well versed in these algorithms, that poker is much much more complex.

----

But there was also the question, "Which game requires more analytical skill - chess or poker?" This seems to be more human-centric. It's similar to asking, "What requires more athletic skill, basketball or soccer?" The problem with this question, however, is the "analytical skills" required to play chess and poker are very different from each-other. Now we're comparing apples and oranges. There is no good objective way to answer this question, whereas I think we can be very objective about comparing the "complexity" of the two games.

First off, solid answer and I can definitely see the reasoning logic behind it but it is off.

We run back into what sort of "algorithms" a computer is efficient at solving vs what sort of algorithms the human brain is efficient at solving.

The reason why chess may be more complicated to the human brain vs the computer is the # of combinatorics which is almost infinite as far as the human brain is concern.

So this may make a brute force algorithmic (trial and error) approach ideal as far as a computer is concerned but yet that same approach is beyond the mental capacity of humans.

Now take poker. From an algorithmic (trial and error) approach this may be a nightmare for computing. At any given time in poker you are presented with one of 4 options: Check, Bet, Raise, or fold. This work with a variable betting amount vs randomized cards on changing streets... Unlike chess, all the variable/parameters are not known (i.e. cards) and the variance makes the same brute force algorithmic approach that a chess program may use impractical for poker (i'm guessing a bit here)

But as far as the human brain is concern, we process the "unknown" on a subconscious level and its child's play to us. Similarly those 4 options are something we can more easily wrap our brains around (vs the 32 pieces we have to wrap our brains around in chess).

The more I think about this argument, the more i'm convinced that we can't assign common metrics here and without common metrics and definitions I just don't see how we can use computer "complexity" as a metric for determining whether are not A is more complex than B.

What we can 100% agree on is that it is harder to write a computer program for A than B. Okay, I can buy that argument.

But that doesn't mean A is harder than B.

It just means that A is harder for B for a computer.

Thought I'd try a different angle and just argue directly for why I think chess is much more complex.

In my mind, it comes down to what I think of as "significant combinatorics".

So what makes something significant? Well, if we are playing NL, eff stacks 100bb, pot is 40bb, then there is no real difference between betting 10bb vs betting 11bb or betting 99bb vs betting 100bb. You could group bets together in ranges that you consider equivalent. So you can reduce the number of so-called parameters by orders of magnitude from a standpoint of doing calculations.

In chess, you got 32 pieces on a 64 square board in a 2-d geometry with each piece being mobile and interacting with each other. In my mind that results in more combinatorics than I can even begin to imagine. But what makes it even more "complex" is the sensitivity the game is to mistakes. The simplest mistake can lose you the game. Moving 1 pawn 1 square instead of 2 squares can lose the game for you 20 moves down the line. This is chaos theory incarnate where the most minor "thing" can have a HUGE consequence later.

In NL poker, you have 2 cards, 4 options (check, bet, raise, fold), 4 streets (preflop, flop, turn, river), and chips. This in my mind is far simpler and results in a lot less combinatorics by a factor of probably about a million. Poker is no where close to the chaos that is inherent in chess (I mean the mathematical/physics definition of chaos). If you are in a poker game, you will rarely encounter a situation where betting 32bb instead of 33bb costs you your entire session. The chaos is low to non-existant in poker as compared to chess.

anyways, its late, I guess we will agree to disagree. I'll think more on this and continue the argument when in a couple of months someone reposts the chess vs poker argument because it comes up fairly frequently on here. So i'm sure we'll cross swords again on this issue.

99% of those excellent in any technical activity will say that it is FAR more complex than other technical activities. There is a strong psychological tendency to think that what one has understood is more profound than what others have understood. This is why physicists look down on engineers and engineers look down on physicists.

Yeah I dunno maybe I enjoy poker more so I trick myself into thinking it's more creative =p. I definitely enjoy the psychology, mental control and logic aspects a lot in poker.

Or maybe your'e doing that too as well haha. I'ts fun to compare them and think about the differences, but as others have said it's like comparing apples and oranges.

Chess is complicated combinatorially. On the other hand there exist formalizable heuristics for eliminating from consideration many of the moves possible at a given moment. Also at every moment each player has complete information about the opponent's possible future moves and past play. Computers can be made to play chess better than people because the combinatorial difficulties are surmountable and heuristic scoring of potential moves is quite viable. Chess has also been studied quite intensively in a formal way. A hundred years ago the chess champion Lasker was also a professional mathematician.

Poker is trivial combinatorially. At any given moment one can analyze the exact probabilities of the possible hands to come. The difficulty comes from the lack of information about what opponents hold, and about how they have played in the past and will play in the future (deceptive play has no analogue in chess). A Bayesian statistical approach is almost obligatory in assessing things like expected value, and plays cannot be considered in isolation (as they can be in chess), but rather must be considered in the context of thousands or millions of repetitions. Poker has not been studied intensively by serious academic professionals (e.g. mathematicians). From the point of view of a professional mathematician a book like Sklansky's is quite primitive (this is not meant as a criticism). For instance, there is not much explicit mention of variance and there is no attempt to formalize heuristics for guiding play. The literature on something like bluffing is surprisingly limited. There are related contexts in which a lot of useful formalism has been developed - the theory of auctions, for instance - but little serious effort to link them to gambling seems to have been made. The subjective nature of the assignment of probabilities in the presence of incomplete information is clearly an obstacle to programming poker. (Also, who wants to write a program that beats people at poker? In the long term that's a money loser because it kills the game.)

There is a big cultural difference between chess and poker. Chess is a socially acceptable thing, smart kids are guided towards it, Nabokov wrote a book of poems and chess problems, etc... It is part of "high" culture. Poker is gambling, occurs traditionally in casinos, cardrooms, and shadier places that children are shouted at for going near, etc... They attract different sorts of people. The sort of person who becomes a professional engineer, like the chess grandmaster working down the hall from me, is more likely to play chess than to play poker, and this may affect how they get studied, and who studies them. On the other hand, there is more money to be made in poker than in chess, and there may be more incentive to keep secret what one knows about how to play well.

I don't take too seriously any of what I just wrote, but they're some thoughts.