Quote:
Originally Posted by samooth
From the 538 article above:
First of all, the "more atoms than.." line is getting old very quickly, Second, and more importantly, HUNL isn't that complicated of a game as those researchers are trying to sell it (and it's not even close). That ~10^160-170 number that you read in articles and academic papers was artificially created by the ACPC, explained in detail here.
No sane person would describe 200bb HUNL as that big -- adding tons of sizes to the betting abstraction that are strategically indifferent to a LOT of other sizes and don't add EV doesn't make a game more complex. In the format played in this competition, a player has 19800 potential 3bet sizes after a button minraise, for instance...
Are we sure they are strategically indifferent? i was of the idea that the optimal preflop strategy could converge to a mixed strategy of combo-weighted bet sizes (depending on position at 6max), like opening (for ex) 30%) 2.3x , 25% 2.6x etc, with most combos in most opening sizes.
It's wierd to dismiss that automatically and calling that strategically indifferent.
It could very well be that GTO HUNL strat is opening 32 different sizes, and betting many many kind of different sizes to for small but significant on the long run gains.
Maybe this isn't true but until we have a "close-to-GTO-enough-bot" how can we tell? and that would require no bet astractions to try and search for those mixed-bet-sizes strategies preflop and on the flop.