It doesn't show it, but all the plays the solver makes preflop are already the best knowing bunching effects and removal, that's why i said your example is bad. if the solver says something is mixed, it means it's mixed given bunching effects and all that

Not sure about other solvers, but you can display bunching in HRC by right clicking a node, node details, bunching.

Here's some interesting stuff on the bunching effect:

https://www.holdemresources.net/blog...ching-effects/

https://forumserver.twoplustwo.com/1...ffect-1792290/

I did not know that thanks, also I can see evidence for this.

I'm looking at 6max vs 9max.

TT always 4bets in 9max but not always in 6max because of bunching. The same example I was using earlier. We 4bet more in 9max because BB has more Ax/Kx which means his range is weaker.

6max: SBvsBB3BET



9max: SBvsBB3BET

That's just the rake. If you make them both 50nl they're literally the exact same range which seems off

Is it just the rake? I had any rake setting.

Let me check 500nl rake. 6max



500nl 9max



Big difference here and the same rake. So I'm pretty sure it's the solver calculating the bunching effect like anero said.

I think 50nl rake is an outlier situation because rake is 8-9bb/100 or something silly so it just jams both.

Usually when people collude they share cards. A random example: You 4b shove AK preflop and the cheater calls with 66 because his partner had Ax, which increases the equity of 66 vs your AK heavy range. In this situation you will run under the EV longterm.

Since when? Over a sample of ~2mil hands, the difference between my EV and BB/100 is only around 0,08bb/100.
+/- 2BB would be an absurdly high difference.

In late positions, AK can gain extra ~4% equity vs QQ when you apply the bunching effect. This means that the player who tends to get it in preflop with pocket pairs more often, should consistently be running under the EV.

There are also spots postflop, in which a good reg will take advantage of card removal. For example, a reg might fold the nut flush draw vs a shove in a 3-way pot, just because he assumes that a rec has a flush draw too and is blocking the reg's outs. This gameplay would make better players run above EV postflop and worse players under EV.

In conclusion: BB/100 stat is definitely more accurate than EV BB/100 over large samples. You guys are just arguing about the exact value of the difference, which wasn't the question of the thread.

it says 50nl

woops.

I only have the starter package for GTO Wizard so either they don't have 500nl 9max ranges or I don't have access to them.

Do you have 500nl 9max ranges or some way we can compare?

The 50nl rake solution is troubling because they shouldn't both be jamming at 6max or 9max unless rake is that big of a factor.

Okay here we go.

50nl rake.

6max simple solution 50nl rake.



9max simple solution 50nl rake.



***** I think you are right.

These should not be the same and goes against solvers taking bunching into effect.

Anero/ZK what do you guys think?

Even thinking about it more, they are the EXACT same EV's. That definitely can't be right.

Okay now I go back to my original belief that solvers don't take into account bunching.

tldr: Poker is hard.

Yeah thanks for breaking it down.

Next question is:

Do solvers actually take bunching into effect because I have evidence that they don't.

Can you specify the exact situation, spot? Are you suggesting that because someone has a partner holding Ax, you can start calling with 66?


I thought the question was about smaller sample sizes. For example, if someone has 2 million hands, it's obvious that the focus on their EV or BB win rate isn't as critical. I assumed the question was more about wanting to understand these metrics over a smaller sample size, like 100k or 200k hands.

@↑DooDooPoker, could you clarify what you were referring to in the original post?

HRC (what I believe GTOW is using now) definately takes bunching into account.
But it seems GTOW's older preflop sims doesn't take bunching into account, which is weird since I'm pretty sure MonkerSolver also takes bunching into account.

It is rare to take bunching into account for postflop sims.

Maybe GTOW created the 6-max sims by simply running a 9-max sim and then removing the first three players?

My original question was what was a better metric to use. I didn't think about sample sizes at all though.

Like most concepts in poker, when you ask one question you start to think about more questions to better understand the concept and we get to go down the rabbit hole.

Good information.

I'm wondering if GTO Wizard has any articles or information on exactly how they run their sims because that 50nl 6max/9max preflop sim definitely does not take bunching into account.

The preflop solvers that I used did take bunching into effect. I've tested it.

Most likely GTO Wizard owners just took some shortcuts to avoid running the "same" sim multiple times.

If his partner folded Ax preflop, then pocket pairs like 66 will have higher EV for getting it in preflop.

If his partner had something like JT, however, getting it in with AK/AQ preflop is more profitable for the colluder.

He has this extra information and you don't. Therefore, if you play vs colluders, you will run under EV longterm.

Interesting thread.

I still think EVBB/100 is way better for smaller samples. In the long term it suffers from systematic bias, but in the short-term the variance reduction is worth it.

Most preflop solvers account for bunching (including GTO Wizard). Most postflop solvers do not account for bunching. If you want to experiment with postflop bunching, I'd recommend HRC, Jesolver, or WASM.

Re: card removal

Is there any software that, by providing preflop ranges for everyone, calculates the likelyhood of cards showing up postflop (or by giving ranges to everyone and a fixed preflop formation, how likely YOU are to get dealt an A, a 5, a 2 etc?).

I coded an excel vba script a while ago to check it (it might be right, or it might be wrong and I missed smth), for a 9-max spot where BTN opens and BB calls. It's a little more than 3x more likely for any card on the flop to be an A than a 2. The A is the card less removed from the deck. Interestingly, with the ranges I gave (a simplified version of GTOwizard with 100% frequency for every combo but matching the RFI%), 8 is the second most less removed card, thus the 2nd card more likely to appear.

Would love if someone has another tool to check this up, or did the work and found something similar.

Thx Tombos.

So you think EVBB is better for a 100k hand sample but not for a 2 million hand sample?

These are my thoughts as well: there could be a small impact on EV from colluders, but in my opinion, EVBB performs much better on smaller samples.

I think $/hr > EVBB/100 > BB/100

And if you want to overconfident assassin meme it out even further it's probably brain energy EV/hr (BEEV), cross referenced with expected brain energy EV (EBEEV; a.k.a. how objectively taxing the spots were based on psychology metrics) > $/hr > EVBB/100 > BB/100

Then you take your EBEEV/BEEV calc and overlay that with what kind of day you're having (WKODYHEV) to get a remotely accurate picture of how much you’re wasting your life. But you might need invasive brain surgery first. (IBSEV)

Everyone always overlooks the WKODYHEV

I believe for the 9-max sims they just reused the 6-max ranges, so bunching will be off for those sims.

Solvers definitely take into account bunching. GTOwizard is a library of sims, not a solver. They can run sims in whichever way they like without it necessarily alligning with gto