Basically the title.

64 gigs of ram, 5900x. So, most configurations shouldn't be a problem, but of course time is a factor.

Also, as far as bet sizing is concerned, there's no point in having a 25%, and a 33% bet size for example. That being said, are there spots where a solver would want to bet 50% pot, but not 25% or 75% with a similar range? Or in other words, is there a significant EV loss in having only a 'small' or 'large' bet without a 'medium'? Does something like 25, 75 and 200 seem reasonable if thats the case?

What about flop bets specifically? Are we losing a significant amount of EV by not solving with an overbet size on the flop? I know there are spots where a solver will choose to overbet the flop.

Edit: what about raise sizes? Something like a ~33% and a 100%?

Edit 2: I guess it goes without saying that preflop action matters. Let's say BU vs BB, single raised pot for starters. Curious to know how you would change your tree based on preflop action and positions though.

Any ideas?

This topic comes up a lot and the answers are never very satisfying. The truth is that there isn't any hard public data on this, and the EV difference is very small and hard to measure over a large flop subset.

You'll want different trees for different SPRs, and also different trees for when OOP is the pf aggressor. It also kind of depends if you're looking for efficiency or accuracy. If you're trying to solve 180 flops you'll generally want to use something efficient that solves quickly. If you're studying one board then you can invest in a more detailed sim. You might also want to create custom trees tailored to each spot by editing nodes directly.

One tip: always provide both players a comfortable path to all in by the river. This means you usually want overbets on the turn and river in single raised pots. Enable all in as a sizing choice for the river.

I guess I'm looking for a more general tree structure so I can solve many flops. At the same time, I don't want to sacrifice EV by not including certain sizings that solver will prefer.

Do you have any recommendations for such a tree? again, BB vs BU SRP 100bb deep

(30, 75), (75, 150), (75, 150) would be fine

Really? I've been doing 25,70,200 (or something like that, maybe its 75 instead of 70 etc), and one thing I've noticed is there are TONS of spots where the solver loves to 2x pot on the flop. For example, AJ3tt, AK2tt, although maybe it's not a significant EV loss to not have those sizings? I kinda like having the 2x flop overbet in my range.

Also, no small bets on turn or river? Are we not losing much EV there either? Maybe I'm overthinking this, but I would think a small bet range on the eriver would be worth having? IDK about turn

BTW, is there a reason you go for 150 and not 200, or whatever other overbet sizing? Or is 150 just sorta 'standard'?

Finally, regarding raise sizings, right now I'm using 33 and 100, so basically a min raise and a huge raise. Should I just leave the default (or set a default to be like a half pot raise (50)?

I'd recommend going to zenith.poker and downloading some of their free sims from the library. They use huge trees for some common spots that will give you an idea of how certain spots play. They even have 4x turn probes in SRP.

With that said, it isn't really necessary to have a million spots everywhere. It is better to play a strategy you can execute.

As long as you can get stacks in by river it's not going to have a significant impact on ev.

Having a solid strategy that you can execute well is more important than striving to gain 1% of ev.

there are 2 ways to do it

1. you allow a million bet and raise sizes on all streets and end up with a 100gb+ tree that is not even practical to drill
2. you think in advance of a simple but not dumbed down model that covers most of your needs and you work with that, adjust as you experiment with it and see fit

As complex as possible to allow for as accurate of a sim as your computer can handle. Of course not with the intention that you will ever be able to execute such a strategy, but with the intention of being able to extrapolate concepts and ideas and general tendencies that you can actually apply to your own game.

For example if you never ran a sim with a 1.5x flop overbet you might never know that on some type of boards IP is suppose to play an extremely polarized flop strategy and once you consider game theory fundamentals it becomes obvious why that is so and it's also very easy to actually apply in game. All sorts of quite easy to spot and understand lines will come up when you're studying accurate sims.

Talk of "ev" when considering solver outputs is very misleading and pretty useless in my opinion. It comes with the underlying assumption that a human could ever properly execute even a simple 1 bet size solver strategy. Sure for something like the flop you could get pretty close, but that flop ev is only guaranteed assuming you can play all possible turns and rivers exactly like a solver, which you can not.

In reality ev comes from knowledge asymmetry and you're not going to create that asymmetry by trying to play the simplest and easiest lines that every other reg in the pool knows and has studied.

Yeah that was my thinking as well. It's of course harder to intererpret and attempt to implement a flop strategy with 2, 3 or even 4 bet sizes than it is with 1/3 pot or check, but people will have also studied those lines less, and make more mistakes against them

So no small turn or river bets? No Flop overbets? Would it be worth adding 1 more size to each street, that are different enough to where there would probably be a clear distinction on when to use which sizing (of course some hands will use both)?

For example, 25, 75, 200 (or 150?) on all 3 streets?

I've been running sims on 3bet pots BTN vs BB with 25, 50, 75, 100, 200, and jam if less than 400% pot. It's the not knowing that kills me. I'm curious if there are spots where half pot is strictly better than both 25 and 75. Is that why the 75% sizing is chosen? As a nice middle ground that accomplishes roughly the same as a 50% or 100% in most scenarios? Is it becuase the odds you're offering villain by betting 50% are not so far off the odds you offer them when bettin pot? having to win 25% of the time vs 33%. Another way to phrase that is: would our betting range look very similar when we bet 50% or 100%? And would the same be true about villain's continuing range?

How's 25, 75, 200 on all 3 streets? And of course the good ol' jam if pot is less than 400% or something

turn/river sizings won't influence flop strategy "that" much, just if you don't allow large bets on turns/rivers you might end up betting more often or something and not allowing small sizes on turn/river will just mean you end up checking more.

IDK how much turn/river studying you're going to be doing with these trees. If you want to study turns/rivers with these solves then by all means include as many sizings as you want, but just understand that it will end up increasing the tree size by a factor of 2-3x potentially which isn't always worth it. A full tree save is a couple gb (depending on the sizings/ranges used) whereas just a flop save is like 1mb or less.

File size is the least of my concerns honestly