Quote:
Originally Posted by Avaritia
+1
LP16,
I have read elsewhere that AI naturally came to the result of overbetting the nuts often.
Can you explain why AI comes to these conclusions? Is the idea that "when you have the nuts you want to put infinity in the pot" the natural learning AI developed and therefore over ships entire river 3b range here?
It depends on the situation. If our range is perfectly polarized then the optimal bet sizing on the river is all-in. This can happen when the nuts can't be shared and we have bluffs which block the nuts, or when the opponent's range is capped. There are plenty of situations where we shouldn't overbet the nuts.
Take the obvious one which live players seem to never get right. The board is KQJT5, no flush and one player bets 80% pot. The opponent can defend with the nuts only, so the bettor gets no value with his nuts and cannot bluff profitably because the nuts is such a big part of ranges. The correct bet sizing here is likely very small with both nuts and bluffs.
Another example: Maybe we have a range of thin value hands on the river that maximizes value by betting relatively small. If we bet with a capped range the opponent can exploit us by raising aggressively with hands he knows can't be beaten and a balanced number of bluffs. However, if he does this often enough then now we have an incentive to also bet small with the nuts.