Every now and again a highstakes pro will casually mention something that challenges your assumptions of game theory. In this video, Uri Peleg states that you should actually 3bet smaller as you play deeper.

https://youtu.be/FKOwoUfdmgk?t=81


My intuition was always that you want to raise bigger as stacks get deeper, adjusting to the SPR so to speak. Perhaps it's very sensitive to the tree construction, or perhaps this doesn't generalize outside highstakes heads up. Or perhaps 3bets are smaller but then 4bets are much larger. Either way, I'm still curious about the underlying logic.

What are your thoughts?

I learned today that cold 4bets should be smaller than regular 4bets. That was something that was not intuitive to me.

As far as 3bets go. You are supposed to size down the more people left to act - so an UTGvsHJ 3bet should theoretically be smaller than a COvsBTN3bet.

When 200BB deep. Our OOP 3bet percentage goes down and our IP goes up.

Sizing is interesting. I'm not sure on that topic. I can see arguments for both. I'm going to try to find out the answer though.

Will report back

@DooDooPoker how do you find out an answer like this on your own? Can a solver like gto+ solve 200bb+?

You can't really.

You'd need the PIO preflop solver to go 200bb+ deep and that requires a really strong pc that a home pc can't do. So you'd have to rent a server.

I know a good amount of players better than me so I just ask a bunch of people and see if there is any consistency with the answers.

Getting information from a bunch of different sources is a good way to go about it I think.

Word. I was genuinely curious because I have no idea how one would go about solving deepstack nodes.

As I played and learned, 3bets are normal, 4bets and 5bets should be cautious because ev and odds change vs stack depth, high pairs don't bear the same weight. Use the same ranges, but hold off on that AKo meta.

Also, QQ and JJ lose more value here.

I tend to go larger and with a tighter range OOP/lower freq

I tend to go same sizing and with a wider range when IP

I agree slight deviation is positive when deeper when IP. I do extend my range to calling 3bets when IP, trying to focus on effective stack size.

Let's try and simplify this to one variable. Heads up, out of position, BTN/SB opens to 2.5bb. Does your size go up or down with stack depth? Or is it more complicated than that?

For me personally I might increase my sizing slightly so I'd go more like 12bb 3b here, but I'd also develop a flatting range, so therefore my 3b range would be stronger.

This has something to do with cold 4-bets being merged (unless we have cold call range, which usually we don't)? We don't need to force as many folds as we would with a polar 4-bet strategy.

This is assuming we're at a depth that a 5-bet will be all-in.

Sorry this is a bit of a tangent to the original discussion.

Zenith 3bet ranges interestingly enough are neither larger nor smaller. They only increase if RFI chosen to raise larger first. Also Zenith ranges suggest the solver wants to increase overall RFI size in deeper scenarios.

Uri is right here. We do go smaller OOP but we go the same size IP when deep.

I didn't watch the video.

Is it just because we don't want to get ourselves into marginal spots OOP in potential 400BB+ pots, but we want to put our opponents in those tough spots when we're IP?

I don't know the exact reasons - he doesn't say in the video.

But I'll speculate - I think it has to do with SPR thresholds though. Like let's say we go 13bb instead of 11bb for a BB3bet vs BTN.

Now the SPR is 7 instead of 8.5. There's almost no difference there when it comes to strategy.

At 100BB Effective in 3bet pots - we get to jam turns at 4 SPR at some frequency. We never get to do that at 7+ SPR.

Brokenstars makes a cool point about developing a flatting range in the SB. I'd like to look more into that as well

when you're very deep playing big pots OOP is not good

Very interesting. Don't have much to add, except that I like these sort of threads a lot better than the bad beat hand histories at 5nl.

I would speculate that, in addition to the reasons others have mentioned, a lot of common sense 100bb pre-flop theory comes from the fact that 5-bets are always a jam, and that 5-bet jamming is an important weapon the 3-bettor has. So as stacks go up, we don't just want to 5-bet jam (or just want to do it less often), thereby making 3-betting large less appealing, since we're no longer setting up for a jam

Correct, as I have been taught (obviously, one might want to make exception for AA, but I watched a big stack stumble vs 60+ BB shove with 66 vs AA).

But, I generally keep the 3bet sizings around the same, I just widen my options IP (presuming there are no others left to act).

Well said, I think this is the root cause.

100bb deep, we 3bet larger to force a good price on a 5bet shove. But if we're playing way deeper then we can't force a good price on a shove, and therefore the large 3bets become less efficient.

I'm gonna run some tests and publish the results here over the weekend. It's easy to defer to common wisdom or backwards rationalize what pros say. But if there's one thing I've learned, you have to try to find the answer for yourself to understand it.

Well, it took a lot of time to render, but I think I've confirmed this effect. The following is taken from HU solves, BTN can open to 2.5bb, BB can 3bet to 3x/4x/5x/6x, lots of 5bet sizes, and 6b is always all in. No rake.

100bb effective stacks seem to be in a strange "goldilocks zone" that makes extra-large 3bets especially attractive in order to force a good price on a 5b shove.

At 50bb and 75bb effective, you want to 3bet smaller as we can 5bet jam even against a min-raise. At 200bb+ we can't always get a good price on a shove, and therefore a smaller 3bet starts to become more attractive.











Detailed sim results below:

50bb effective
75bb effective
100bb effective
150bb effective
200bb effective
300bb effective

Something else I've noticed is that the BTN seems to adjust their 4bet size, such that they size down against the large/polar 3bet and size up against the smaller 3bet. Now that makes sense logically, but I also think it has to do with optimizing the 4bet size to make BB's 5bet less effective.

Nice work mate

Tombo's - what kind of software are you using to get these sims? Also, what ranges are these for 200bb/300bb?

Awesome stuff.

I'm using HRC Beta, an experimental preflop solver that allows people to run pf sims without the insane memory requirements.

Regarding the ranges, the solver chooses all actions, but the opening range doesn't change too significantly from 50bb-300bb.

very good stuff tombos. we can also see the difference in the composition of 3bets in shorter vs deeper stacks

most pros i've seen size up when deeper so this is really interesting, thanks! i would imagine its not necessarily the same IP or OOP either, and definitely not a linear scale. i've always thought of it in terms of target SPR, where as stacks get a bit deeper its nice to size up such that you can still comfortably get it in with your overpairs, but once stacks get big enough that that's not really an option its good to go smaller and avoid awkward middle SPRs