Last year I finished my bachelors degree Economics on the University of Amsterdam.
I have always loved the game of poker but since maybe 1.5 years I have taken it more serious.
I had a subject during my bachelors named Game Theory which I found really interesting and that increased even more when I found out it had a lot in common with poker.
So I was very pleased when I found out they offer a Master degree in Game Theory and Behavioral Economics which I will start next week.

Are there more people here who are doing or have done Game Theory on the University?
Do you think it will be easier because I have some poker theory understanding?
Or will it be the other way around and will my poker-game improve because of things I learn during class?
And do you think it is possible to keep my thesis poker related? What are some serious academic researches I could do considering poker and game theory?

You should research how sample size and getting to the long term affects profitability when using game theoretic vs. exploitative strategy.

Basically gto is supposed to be unbeatable in long term but doesn't mean it's most profitable.

There are papers on poker in the mathematics literature concerning simplified models. I don't think poker is an interesting game theory problem essentially because it is solved: up to having a big enough computer. Thus the problem is more one of computer science, of implementing the big calculations. There are interesting applied mathematics questions, again connected to how one might organise the calculations efficiently.

An economics or game theory course will focus more on those topics rather than implementing the biggest, worst case of a well-understood problem. They are quite fun topics and not that hard either. My guess is that study will improve your poker but knowing something about poker probably will not help in such courses too much.

</Engage disappointment>
Your thesis will be related to the areas your advisor understands, otherwise they can not advise you. If you do not follow an advisor's topics then you will most likely fail because your work is wrong, already done etc etc. I know that is disappointing and rather dull but it is sadly true.

Really?

You cut the sentence at the grammar which gives a disingenuous quote. Nevertheless I think the statement is fair. We can argue about what is interesting to a game theorist for a long time of course. (In honesty I only know one game theorist but I do know a few decision theory people too.)

Among the first things you learn is that Game Theory Optimal is not the same as Game Theory. It is GTO which has been solved for Poker, just about.

I haven't studied game theory, but I have spent a very long time with Poker Theory and I've had an interesting conversation with a Game Theory lecturer. From what I can tell, the logic matches up very well. To learn Poker theory is to learn Game Theory.

OK
Really?

It's not clear to me whether you are trying to ask an honest question or tell me I don't know what I am talking about. I'm content with the later.

I would say poker is solved, in the sense that we know how to solve it. I would say it's not solved in the sense that we don't have the solution here in front of us. Whether you consider it solved or not will kind of depend on your viewpoint.

To a mathematician a problem is solved once the method is clear, the actual numerical number might not be of interest and the mathematician might not care that the answer is practically hard to get.

To an engineer a problem is not solved until you have an answer you can implement

There are techniques that I think will clearly solve any HU poker game given enough computing resources. On the more practical side are techniques like counterfactual regret minimization. But also, brute force is always an option, just not a practical one.

But it's a little like saying "sending men to mars is a solved problem" which is true, yet, we don't have men being sent to mars.

Rusty standing in line believing the lies as usual.

----

On topic:

You could construct some examples of collusion that disrupts the nash equilibrium in multiplayer poker situations and show how to counter exploit such situations.

Rusty's pretty smart, and his argument makes some sense, though when he says, "given enough computing power", I don't think this is accurate. Today, "enough computing power" is practically infinite and past the size of the entire economy many times over, especially for more complicated games. There are, in my opinion, missing algorithms still.

Here's the flaw with Rusty's argument. Mathematicians have long known that computers could solve these kinds of problems "given enough computing power", so poker was solved decades ago, under Rusty's argument.

Collusion is NE in multiplayer games. That's my understanding.

https://en.wikipedia.org/wiki/Folk_theorem_(game_theory)

I clicked and skimmed, but didn't see any mention of the possibility of counter exploitation of collusion.

I think if we offer the opportunity to counter exploit as part of the nash equilibrium in multiplayer games, that such alliances will dissolve to the point that the collusion should never occur in the first place.

I agree. I was making a mars joke relating to this statement.
Which is clearly ignorant of the truth. Not only do we have humans on mars, but on the moons of the gas giants as well as the larger asteroids.

To do a brute force analysis for HU NLHE is probably practically impossible. Even for LHE it might be practically impossible.

But for methods like CFR I don't think that's true. I think a few orders of magnitude from now will probably be enough to crack HUNLHE for example. Possibly only for limited scenarios - such as, limiting bet sizes to, say, 1/4, 1/3, 1/2, 1x and 2x pot size or something like that, and/or limiting stacks to integer values of chips up to a maximum.

I think most people don't consider a game solved until a solution exists that you can point to and use, but I was just trying to explain how to a game theorist, the methods to solve poker are essentially known. I don't actually know if a lot of effort or interest exists to try to to improve approximate solutions for the game at the academic level. Obviously, there is a lot of engineering effort going into it, but that is by no means the same thing.

You have said this on multiple occasions.

It is not true.

It reflects a misunderstanding of what Nash Equilibrium entails, especially in a poker setting.

Muppet's answer seemed reasonable to me.
Not me.
Possibly, but I think it's very possible that your game theory course will make no mention of poker at all. Poker has way more variables than the 'games' that I believe are used in the teaching of entry-level game theory and behavioural economics. I'm pretty out of touch with what's taught in universities these days, though.
EDIT: I'm aware that M.I.T. had a poker module as part of its Game Theory syllabus, but I don't think it was compulsory.

I don't get it. You people say Poker is not solved. What does nlhe solvers do then?

Its not solved. The goal in Poker is to win chips, and so to 'solve it' you would need to know how to make the maximum amount of chips possible. And that has not been done and cannot be done.

They try to find GTO. This is a strategy which everyone knows is not the best strategy.

Solvers find maximally exploitive strategies for heads up situations, given ranges, stack sizes, and predetermined betsizes. Since optimal preflop ranges and betsizes are not known at this time, we cannot say that poker is solved.

It may not be the most profitable strategy, but I think gto would do quite well in the modern poker environment; what with people 3 betting KTo and other such horrible hands preflop.

If you let a chess engine think long enough, it will solve chess.

Yeah I agree, but finding a strategy that does quite well is not solving the game.

finding a strategy that beats any other strategy is...

In the context of game theory, which is what this post is about, "solving" a game has a very specific meaning. Some games have solutions that satisfy everyone. Many games, like poker, do not. As many people have noted over the years the word "optimal" in GTO is an unfortunate choice because it does not meet the intuitive defintion of the word optimal. When I studied game theory, those kind of solutions were usually called "sub-optimal" and were found in games where optimal solutions weren't possible, such as the classic single-iteration of "The Prisoner's Dilemma"

Finding a perfect exploitative strategy to a specific fixed strategy is actually very simple and does not take nearly as much computational resources as finding GTO strategies. It's also not particularly interesting from a game theory perspective.