Quote:
Originally Posted by Cangurino
However you seem to suggest that this does not hold for equilibrium strategies in multiplayer games. A few strong players appear to disagree with that.
My suggestion in multiplayer games is still to play defensively, but just to be cautious with the results of analysis and not to depend on equilibrium strategies to necessarily be strong.
For example, if you have a decent approximation to optimal in a headsup game, you can basically just blithely play it against anyone, because the guarantees are so strong about how bad things could be against you. But I don't think this is true of an multiplayer equilibrium strategy; also I think this problem can be exacerbated when you have special knowledge unavailable to your opponents that leads you to the NE strategy.
For example, consider lategame NL SNGs. Suppose that the population of these is like nearly 100% people who are using software to calculate some approximation to a NE. Then probably playing the NE along with them is quite right.
Now suppose that it's earlier in time when software wasn't available or whatever. Suppose the NE includes a bunch of looser-than-anyone-expected overcalling along with loosish calling. (I'm just making this up, any resemblance to reality is coincidental). Now if everyone plays a more standard tighter strategy, you'll benefit in small chunks from their tightness, but get hurt badly when you overcall with much weaker hands. In a sense they are implicitly colluding against you without even trying.
I think this kind of stuff happens often in practice in multiway games, especially those games where analysis hasn't mostly converged the basic strategy. So I like to draw a distinction between the strong guarantees of HU strategies and the weaker ones of NE. None of this is to say that the strategies themselves will necessarily be weak. They will often in fact be quite strong.