Hard to say. Bots look to find *an* optimal solution. Just because a GTO bot takes one line doesn't mean another line can't also be GTO.

There's also the fact that we are playing a human. It could be a human will adjust poorly to our limp. If limp/call is only marginally worse, deviating has advantages of putting villain in a unaccustomed spot.

Last GTO bot I looked at raises SB v BB with Q3o and K2o. Not sure how you got Q6o to be a fold.

i just tried to find a bot online and ran across "neo". prob not one of the real bots that you're talking about. i limped qto in sb and the bot c/'ed option with a9o and raised my bet on qj6, i called, he checked a 7 turn, and i bet an ace river and he called. wtf is that all about?

which bots are you guys looking at?

phunk-> how many effective local maxima (can't think of a better phrase to describe this) do you think there are in a gt problem like this? now, don't take this to mean that i think the logic dd was using is perfect, i just wanted to point out that this probably isn't a case of many potential solutions (so there probably isn't a solution, according to the bots in question, where EV is maximized and there is an "open limp/call sb" bucket).

now, to deal w/ the question begged here (does it make sense to build our game around what bots don't do/never do/sometimes do?), that's pretty interesting.

first, i don't think we're capable of playing like bots since we can't have the same perfect memory and ability to randomize so easily. nor can we work backwards from the trees that bots build and fill in the subsequent buckets correctly (especially not in game time, but assuming we did this work away from the table and applied it in game time). as a result, i don't think it makes sense to take one or any number of things a bot (also assuming the bot is a great/winning one lol) does and just add it/subtract it from our play. the reason is that, assuming the bot "solved" the outcomes by maximizing EV in conjunction with its whole game, we'd have to fully understand the bot's (bots') entire game and how the action in question affects the other buckets and actions.

to quickly address phunk's point here ("just because a GTO bot takes one line doesn't mean another line can't also be GTO"), that actually can't be true (making the usual assumptions about how the bot got to its solution etc.) b/c the O in GTO is, by definition, one singular strategy. as i mentioned at the beginning, i don't think there are tons of local maxima running around here where lots of various strategies can result in GTO outcomes. in fact, if we could list them all and their expectation, we'd of course find that only one is ACTUALLY GTO (b/c any others w/ lower expectation would be less optimal, thus failing to live up to the moniker GTO).

since limit holdem can't actually be solved this way though (trees too big, too many unknowns etc.), we are left to guessing a bit.

so back to the questions in this thread:

1. sb opening play: i wouldn't limp/call or have an sb limping strategy in general not solely because dd points out that well designed bots don't do it therefore it can't be optimal, but rather because it opens up my game to a) requiring a LOT of at the table adjusting/thinking, and b) being worse EV-wise since i'm more libel to make mistakes in designing and applying the buckets (i.e. if i'm open limping the sb, i need to also have a lrr'ing bucket and frequency, a limp/call bucket/frequency, a flop donking bucket/frequency - much moreso than when i raise and am 3b by the bb-, and all of this has to then also be aligned with my sb open raising buckets). i don't know about all you folks, but i'd VASTLY prefer a slightly worse simpler strategy that i can easily apply in game time. so since i'm willing to actually take a small CUT in EV in order to have an easier to implement strategy, the fact that not having an open limping bucket may actually be superior EV-wise anyways seals it for me.

2. sb raising range: on the margins i'd prefer a fluid strategy. against some folks i'm happily folding Q6o and against others i'm raising 100% of the time. that's the type of hand that is annoying to play against very strong opponents and against tricky/agro types. i may end up playing "too tight" w.r.t. bots and really good players, but i don't think my loss in the blinds is much worse for opting to release those hands when i do.

that said, i'm absolutely raising Q8o across the board and probably almost always raising Q7o so Q6o can't terrible either way. similarly i'm 100% folding Q2o-Q4o so coming at it from that side also shows it's a super marginal hand (literally meaning on the margin rather than the traditional "mediocre" meaning of marginal).

anyways, that's a rambly enough post.

This has been mentioned here before, and it's BS.

Have you read Newell's books? Are you calling them BS? They look at bots lines and interpret why it does what it does.

Basic example: Hero BTN raises, BB defends. Flop comes A93r -- Bots have a low cbet for KK, QQ. Flop comes 456tt, bots have a much lower cbet across their entire range compared to other flops. Those are huge lessons people should be incorporating into their game. It doesn't take a genius or infinite memory to figure out why a bot does that.


I think its very possible for one bot to never limp SB, and another bot to have a limp SB range, and both bots to play close to GTO.


If you have no framework on what range to open SB v BB, a good bot's range is an excellent starting point.

Since DD first mentioned bots, I'm assuming he finds them a useful resource. Hence, when I see a bot raising Q3o, it makes me wonder why DD is saying he folds Q6o.

Maybe because when the bots are raising Q3o they have position?

I was doing some re-reading and discovered I have been using the sb ranges in heads-up games for my raising range in the sb in a bvb situation.

well, raising almost every hand is still fun, and good marginal-zone practice

anyway, can anyone point me in the right direction to toy with bots for playing bvb? is there a website or software available you guys are using?

Reread. I've seen Sb v BB bot opening Q3o.

I have never even heard of a bot that simulates strategy for a full ring game where everyone has folded to the blinds and the small blind is OOP. Where could I find that? Would be really interested to try it out.

maybe off a bit, but certainly not bs. i haven't read newell's book, but i have read chen's. and sure there can be a few easy things to incorporate (as i'll note later in response to your posted examples), but major things with multiple adjustments/bucket changes needed definitely align with my point.

agreed, that's actually where i started out thinking that my cbetting across the board was too high in some spots (lmp steals for example it was close to the bu steal cbet % which was clearly way too high. when i saw the increasing cbet % by position, that helped spark that change for me). so i'm not saying that it's impossible, but for highly specific and complex issues (not the easy one you point out here), it's not a good idea.

so here is where we part. i do not think that's easy to reconcile, nor do i think both bots are playing close to gto. that implies that there are TWO solutions very close to each other in terms of EV. i think the always raising bot vs. the sometimes limping bot will have a higher EV ceteris parabis (higher not by like 0.02 or something like that, more like 0.2-0.3, which is a lot).

again we agree. yes, that's an easy starting point. and again is where i learned quite a bit. but it's not a complex multi-street multi-bucket requiring spot like the opening a sb vs. always raising that same blind is.

i agree they are a useful resource. i can't explain why DD does what he does but i can tell you why i don't open Q3o despite bots doing it: you lose very little in positive EV (if it is positive) and reduce the difficulty in implementing your game proportionally by far far more. playing Q3o oop vs competent people is no fun. it's barely fun playing Q7o in that spot lol. so i'm folding it and i'm assuming DD folds it for similar reasons but i can't speak for him.

(i think) he's saying you likely saw that in a HU spot where the sb is the bu and acts first pf but last postflop.

i can't find one either. i guess it's a secret or something.