Pluribus Overbets on the Flop (Poker AI)

Pluribus overbets are probably the most talked about behavior of the Poker AI. Yet we don’t know that much about these and, boy, are they ugly ! In the poker games I play, there are quite a few overbets on the turn with a combination of nuts and draws, overbets on the river with nuts or air. However overbets on the flop are super rare, other than the fish wanting to go all in preflop and as it doesn’t happen, shoves on the flop with his JJ or AKo whatever the board… or overbets on limped boards which well, family pots, who cares? But overbets on a normal heads-up single raised pot, I really see few of them.

Pluribus bets 200% pot on almost 10% of her Cbets. Now 200% is not a number Pluribus gradually optimised based on machine learning training or anything like that. On single raised pot, the researchers only allowed for 3 bet sizes: 50%, 100% or 200%. So the actual size really means nothing.

So in total, on the flop, Pluribus overbets the flop 27 times. In the following chart, I detail the kind of flop (single raised pot, 3bet pot, or limped raised pot), whether Pluribus was the agressor preflop, the position with detail, Pluribus hole cards, the flop cards, comments I could come up with, and range vs range equity for the flop (based on standard upswing ranges, not my estimated Pluribus ranges):

Pluribus Poker AI Overbets
  • What can I add to this ?
  • The majority of Pluribus overbets are IP vs BB , hence with a significant range advantage on most flops giving the opportunity to bluff at a given frequency
  • The range advantage of these specific flops is not particularly better than for random flops at the same positions.
  • The out of position and limped/raise Cbets seem super logical
  • The 2 overbets while defending are just super weird to me, especially the one donking multi way, looks like what the few remaing donk fishes still do in 1-2 live games. Probably just low low percentage AI weirdness.
  • Interesting to note, only 1 of these 27 overbets got a call and none a raise
  • Looking at this, it looks super exploitable to me, like reraise each time Pluribus overbets as the preflop agressor in position in single raised pot.
  • Obviously, I just don’t get it… Not a behavior I’ll add to my game 🙂
  • Tell me what you think, I can’t wait to have someone more competent than me analyse this !

7 thoughts on “Pluribus Overbets on the Flop (Poker AI)”

  1. Strange.
    Most likely, it’s highly related to the stack depth of exactly 100 big blinds PF.
    Over-betting really takes away the OOP player’s ability to check-raise the flop with bluffs and made hands, then make substantial bets over two more streets thereafter for value or as a triple barrel semi-bluff. In other words, it takes away a lot of the OOP players leverage when $550 goes in PF and players have $9,750 stacks.
    Likewise, it takes floating away from the OOP player as an option, then bluffing or check-raise bluffing the turn on a good semi-bluffing card.
    It’s actually a hard strategy to play against when carefully executed. A lot of the time you have to fold the best hand bc you have to put in so much when your opponent could have a set or top pair good kicker. You probably have to fold middle pair, bottom pair, and top pair mediocre kicker. It’s too expensive to reraise all-in and if you call, your opponent can push on the turn when you are beat and with strong draws and take a free card to beat you with weaker draws.

    1. Let me make that a bit clearer. By betting $1,100 into $550, if the OOP checkraises to $3,000 or $3,500, Pluribus gets the decision with bluffing equity to push for $10,000 or fold for $1,100. Likewise, if the OOP calls, there is $2,750 in the pot on the turn, and Pluribus can strategically make two pot-sized bets on turn and river, or push all-in on the turn, or take a free card to hit certain draws.

      1. Note that even if you have top pair, top kicker, or an overpair, you don’t really want to put 17.7x into the pot. You’d like to win two or ideally three pot-sized bets, something like between 4x to 13x. But when more goes in you are likely losing to a set or two pair. When Pluribus puts in 2x with top pair good kicker over betting that’s more than reasonable for it. It can easily give up. But if you call that and twice more or reraise, top-pair top-kicker probably loses you money. Even AK on a king high board.

  2. To add an additional thought, expanding upon my earlier ideas, it clearly loves to make this play in a very specific scenario: When it’s in position head’s up, with exactly 50% to 62% range equity, or maybe 47% to 62%.

    My hypothesis is that it’s protecting itself against lots of small check-raise semi-bluffs and/or call then bet the turn semi-bluffs. The best response, according to Greenstein and Sklansky, to someone who semi-bluff or bluff bets or raises a lot is mostly not to call them more widely but to semibluff raise or semibluff reraise more widely.

    At this stack depth, continuation betting twice pot greatly takes your opponents ability to to this away. Notice thar hands like 77 on AT6 and Q4 on KJ4 have some equity. You’d actually like to bet to protect them bc they are usually ahead. But they are way to weak to call a check raise or to bluff reraise back. Given that it can semi-bluff and protect it’s smaller made hands so effectively with an overbet, it has to play its best hands this way as well.

    Pluribus is effectively hoping that it’s opponent starts reraising liberally as you suggest bc then it keeps stacking it’s opponent when it flops two pair or better.

  3. You’re already doing it in your games lol.
    On another page you mentioned that you like to bet 1/3 on the flop and follow it up with a 2/3 to 1.5 pot bet on the turn. The dynamic that makes this work well is when you opponents raise your 1/3 bet on the flop with way too many of their strong hands. Leaving their call range capped. Allowing you to pound there CAPPED call range on the turn with Large bets. Because large are extremely effective vs capped ranges.

    The AI is doing the same thing but on the flop. Take A Q 3, when an opponent just calls (in a GTO environment) they might not have any combos of AA, QQ maybe not even AQ, AK, A3, or 45. Making a hand like AJ nutted. When our opponent is capped we have wayyy more nutted hands than usual. There for we now have way more bluff we can use to balance those new nutted hands.

    1) In general Large bets work well vs capped ranges
    2) The larger we bet the more bluffs we are allowed to have
    over potting allows for the ai to maximize the amount of bluffs they can have, and build larger pots for their nutted range.

    I think if you click back often with a range that is extremely capped, the AI is going to have plenty of calls and 3 bets ( which you wont have too many hands to defend against their 3 bet range) .

    In your defense there were a few boards where I’m not sure how capped the caller would be like 963 and 983. Not sure why these boards would be over bet unless it has to do with opponent checking instead of leading, or if specific opponent was short stacked.

    1. You’re right, will eventually update my post to take into account. Thanks.

      In these hands opponent is never short stack by the way. All stacks are 100bb at start of hand always. I’ll also check the boards where we don’t see obviously why opponent would be capped

Leave a Reply

Your email address will not be published. Required fields are marked *