Range c-bet BU vs. BB: Why do Peter Clarke and PioSolver disagree?
Posted by whiteshark
Posted by
whiteshark
posted in
Low Stakes
Range c-bet BU vs. BB: Why do Peter Clarke and PioSolver disagree?
Background:
I am currently studying Episode 7 on selective vs. unselective c-betting of the From the Ground Up course. I loved the episode and I do feel like I aquired an understanding of when to range c-bet as BU vs. BB in SRPs and when to go for a selective/polarized strategy. The general idea that Peter Clarke outlines is that the BB (preflop caller, capped) starts with an equity disadvantage vs. the BU's uncapped opening range and seeks for any kind of flop that has equalization potential.
So BU prefers:
1. high card flops that interconnect well with opening range (increasing range advantage)
2. dry flops that leave little room for the BB to catch up (preserving range advantage)
The BB in turn prefers middling cards and connected boards on which he can equalize Hero's value hands, most often of the one pair type, through a lot of two pair combos and flush/straight draws.
Peter Clarke's take on high card/dry flops:
On favorable flops for the BU where we have equity and nut advantage such as Q73r or AK3r, Peter Clarke advocates for a range c-bet of 1/3 potsize with all of BU's holdings. He even explicitly states that "a solver would endorse these strategies".
PioSolver:
I ran these flops in Pio in order to train myself in when to range c-bet vs. when to go for selective betting. The issue is that basically no matter what kind of flop texture I insert (even the most favorable boards for the BU such as HHLr), Pio chooses a selective strategy, never goes for a range c-bet, and checks back up to 40% of the opening range. This pattern is robust for different boards, range widths, and allowed bet sizes (only 33%, 33% and 67%).
My best take on this is that we humans need to simplify and go for range bets even with wide ranges when the spot is favorable while a solver has no reason to simplify. Still, that seems to contradict the episode of FTGU. What's your take on this? Here's a solution for a AK3r flop.
Loading 13 Comments...
It's hard to see which board you used in PIO in the image.
I would say Peter Clarke's statement that Solvers would endorse this is either wrong or interpreted wrong by you. Been a while since I've seen the video. Nut advantage and equity advantage isn't what I use in SRP IP cbets as the driving factor. I'll explain with an example in the next paragraph onwards. Exploitativly I don't mind though. Small size range cbets are somewhat exploiting microstakes population. They don't defend properly against this with checkraises etc.
I generally use A84r and AK4r as two examples to explain what drives our cbet sizing and frequency.
In a single sentence it is "We bet often and small when our range contains a lot of marginal holdings with incentive to bet".
A marginal holding is hand such as 2nd pair, 3rd pair. Showdown hands basicly.
Incentive to bet generally means that when this part of your range is betting it is making your opponent fold a decent amount of equity.
Let's look at both flops. A84r we have a decent amount of marginal holdings such as 8x, 4x, some pocket pairs. If we look at BB's reaction his folding range consists of a lot of overcards to our marginal holdings, even against a small size. Which means that we have a lot of marginal holdings with incentive to bet.
This leads to a lot of betting for small size. We include strong hands here as well. After that we then see if we have enough strong hands to generate a big bet size. You'll notice solver does have it but a small amount.
On AK4 we also have decent amount of marginal holdings. Kx, 4x, pocket pairs. The incentive is less though. There's less 4x in our range. And BB's folding range doesn't have a lot of equity against Kx.
Due to this we go for a big size. You see this happen with strong Kx hands as well. This is because BB's range is wide enough that BB needs to defend with a lot of weaker hands.
Thanks a lot lllCitanul for taking the time to elaborate. The point that range c-bets are driven by the incentive of marginal made hands to bet is interesting. I am trying to make sense of why the incentive of marginal hands to bet should drive our decision to range c-bet:
In the Upswing poker lab, Doug Polk categorizes all hands into four types. 1) Value hands, 2) Marginal hands, 3) Good Air/Draws, 4) Weak air.
Summary: So since we always c-bet value hands and good bluffs (given that we don't check back our entire range), it is the strategy of marginal made hands that determines whether we range bet or not.
Is this reasoning right? Or are there other reasons for placing the decision over selective vs. unselective betting on our marginal hands?
To put this reasoning into practice: On a flop like 8d6d2s when playing BU vs. BB, as the BU we don't have a strong nut advantage neither is this flop particularly dry, however we will have a lot of 6x and 2x that profit from making the BB fold a hell lot of overcards.
So here although there is no strong nut advantage and this flop is rather wet, we go for a range c-bet as determined by the incentive of our marginal hands to c-bet.
I'm familiar with this strategy. I was a lab member a while ago.
8x and 4x would fall in category2. So we are actually betting 1,2,3 and 4.
Yes. Coming from the 4 categories concept you can look at it from category 2 hands. First question should be "Do we have a decent amount of these?". Second question would "Do they want to bet?"
For the A84 example, this is BTN vs BB. But if we open UTG and BB calls we can go through the same questions.
"Do we have a decent amount of these?". We do not. 8x and 4x are very uncommon in UTG opening range. Hence not a range bet.
This just happens to be a bad example. These lower flops are generally better for the BB. BB has more of these lower cards then BTN does so makes more paired hands. It's even possible for BB to create a donking range here.
Another factor we think about when deciding about betsizes is what would be our main advantage over our opponent. Which hands do we have that he does not have? In this case it's a lot of overpairs. Our advantage comes from overpairs.
You are right we do not want to go big as we don't want to isolate against sets. So we go smaller than big size. We also don't want to go range cbet because we miss the board very often. Which leads to more of a half potsize bet.
You have a solver. I would run a bunch of flops for BTN vs BB. Try applying the marginal hands principle to the flops before looking at the solution. If the solution is different from your expectation, look into the why. If you can't figure it out, feel free to post.
whiteshark lIlCitanul Fyi upswing very quickly moves away from the hand categories in their second part into this textured approach.
I think you answered your own question at the end, and its the same answer I would give as well. Humans extrapolate principles and heuristics from what we observe and experience, then we apply them. Solvers are calculating and recalling numbers, they cannot explicitly teach us things. We only observe the trends and apply reasoning to them so it makes sense to us.
Peter is using the best reasoning available to him at the time to create the story that justifies his actions. I don't know how long ago the series was made, but he has probably updated his thoughts with more nuanced reasoning on when to break the 1/3 pot range bet on dry board and deviate to different sizes by now.
Solvers don't need to spend the time rummaging through heuristics to learn the same nuances and to learn when to deviate/randomize. This creates the gap that you are seeing between Peter and solver. Human brain can't possibly memorize every detail, so we must create general principles, learn the exceptions, reform our principles, and repeat. AI/Solver can skip this process and get straight to the answers and regurgitate them, but cannot tell us the big picture of why it takes specific lines with specific bet sizes.
I don't see this as a contradiction, just a gap in learning. Solver would probably think Peter's 1/3 range bet is fine and +ev, but it knows its going to be even higher ev when you check back 40 percent. Not a contradiction at all, its just that solvers are already fine tuned and optimized to the max while human brain is struggling over time to get there.
Thanks a lot for clarifying this!
I'd say its purely for simplification and because the EV loss would be minimal.
Can you try to solve when BTN can only c-bet 33% or check and one where solver can only c-bet 33% (no checking) and see EV differences between 3 solves? Should make things clearer.
Don't forget that solver is going to play "perfectly" as OOP. I would look at how PIO defends (BB's shoes in this sim) and see if you think that BB is going to come anywhere near a perfect defense (hint, he won't). So, that being said, we are likely to force OOP to over fold quite a bit by having a high frequency betting strategy.
One way to look at this is to note down the EV/Equity of the strategy you posted and then lock PIO to bet the same range 100% of the time and see how the numbers differ between the two strategies. If there are fractions of a % between the two strategies, then range betting should be fine (not saying it is best though).
HawksWin
''and see if you think that BB is going to come anywhere near a perfect defense (hint, he won't). ''
That depends alot on what stakes you are playing, if you play somewhat high, regs will have a very good understanding how to defend against high% cbets.
It's basically one of the first spots you start studying when you get a solver because it's such a common spot.
People are more likely to missplay less common spots. Like Probes, delaycbets. Also facing uncommon sizings like 80-110% is harder to play against because its less used and people priority to study against the most common sizings.
Im not saying rangebets/high freq small size is bad, its still the most used strategy, but if you watch very good players, i would say very few of them are limited to a small cbets. Its very easy to play against someone who always using 30% or checks. And even if you can defend properly against odd sizings, i belive more people will get out of their comfort zone.
Robbiish I agree with you 110%. I am in the "selective c betting" camp but we can still employ a high frequency "selective" strategy and make BB's lives miserable. I also like your take on using some larger sizes to mix things up vs the more competent regs and exploit the calling leaks of the rec/fun guys.
Do you have donking allowed for BB?
I recently ran a bunch of sims and made BB nodelock check 100% to save ram.
I compared my sims to GtoWizard sims and noticed a difference in cbet frequency of around ~15%.
Well, some % difference is to be expected since their sims use way more sizings and ram.
But whatever I changed in my sims I couldn't reach a higher cbet % like the Gtowizard sims.
So I allowed BB to donk and now BU did start to cbet with a higher ~15% frequency even though BB only donks like ~4% of the time. Not allowing BB to donk kinda fucked it up.
But if you run larger subsets of flops you will find some textures where BU does like to bet a very high ~80% frequency. Like on these textures, at least in my sims. (check% is last row)
Yep I looked into this a bit closer with a buddy of mine. We studied 149 flops and a more extensive game tree and we also found quite some boards on which the BU is incentivized to bet >85% of hands. So yes, I do think that my initial results were also just due to my simplification of the game tree.
Be the first to add a comment