How much would a 500nl reg lose against the top players 3 handed?
https://twitter.com/George_ymb/status/17...
Saulo and YouMadBro had this discussion/argument on twitter.
Per
35k hands analyzed. Average EV loss per hand compared to GTOW: -0.02 = 2bb/100 GTO only gains EV vs pure mistakes, which is why you can play far from GTO frequencies and still lose no EV vs GTO.
that's not how it works at all, the frequencies are to balance out your entire range, just because a hand doesn't lose EV based on frequencies you will still lose WAY more than 2bb/100 because your frequencies are so ****ed on a range level
that's not how it works at all, the frequencies are to balance out your entire range, just because a hand doesn't lose EV based on frequencies you will still lose WAY more than 2bb/100 because your frequencies are so ****ed on a range level
That is how it works. The "GTO" strategy is static. It doesn't start exploiting your frequency mistakes.
35k hands analyzed. Average EV loss per hand compared to GTOW: -0.02 = 2bb/100 GTO only gains EV vs pure mistakes, which is why you can play far from GTO frequencies and still lose no EV vs GTO.
I really doubt the GTO Wizard EV loss is accurate at allThis is all 10-25nl when I barely studied any theory and would've surely gotten wrecked by even 500nl regs
Honestly I don't think this feedback is particularly helpful from GTOwizard, but it can help you identify spots you *think* you're playing in a pseudoGTO fashion and adjust. If anything, a low loss rate compared to GTO at low stakes is likely indicative of a strategy lacking exploitative play--which should be the primary goal vs. weaker opponents.
I would imagine top players got the knowledge, the experience, and are smart enough to have already learned how people screw up trying to play GTO, like, how their river ranges in x or y textures will get messed up and they will end up over/underbluffing, or over/underfolding, because of how humans on average play when they try to implement flop overbets, or whatever strategy that is not trivially simple. Idk, but I tend to think that learning how people, including ourselves, would screw up trying to play smth, and how to exploit it, is easier than actually implementing that optimal smth. Obv the risk being this is exploitable. Who knows, poker is hard 😃
That doesn’t sound right.
MDF doesn’t care how you range construct but you have to defend properly. You can’t expect to pure fold 0 EV mix-calls and achieve MDF unless you pull calls from another portion of your range.
That is how it works. The "GTO" strategy is static. It doesn't start exploiting your frequency mistakes.
Sure but you’re still losing because you’re playing a GTO bot and you’re making frequency mistakes.
That doesn’t sound right.
MDF doesn’t care how you range construct but you have to defend properly. You can’t expect to pure fold 0 EV mix-calls and achieve MDF unless you pull calls from another portion of your range.
He’s saying that you don’t need to achieve MDF against a GTO bot, as long as you don’t make EV mistakes with hands your frequencies can be very off and you won’t lose any EV against the GTO bot
That doesn’t sound right.
MDF doesn’t care how you range construct but you have to defend properly. You can’t expect to pure fold 0 EV mix-calls and achieve MDF unless you pull calls from another portion of your range.
Sure but you’re still losing because you’re playing a GTO bot and you’re making frequency mistakes.
Zamadhi is correct and honestly he explained it pretty well.
He’s saying that you don’t need to achieve MDF against a GTO bot, as long as you don’t make EV mistakes with hands your frequencies can be very off and you won’t lose any EV against the GTO bot
Correct. If at equilibrium two actions are mixed, then they are equal EV. Therefore, doing either action at a different frequency vs. the static equilibrium strategy will result in the same EV.
This is also true. They are taking EVs from approximations of an equilibrium solution with many, -but not all-, bet sizings along with many, -but not all- stack depths. There are going to be situations in which the sizings or stack depths are different compared to whatever the constraints (bet sizings/raise sizings/stack depths/rakes) used to generate the dEV's.
If at equilibrium two actions are mixed, then they are equal EV. Therefore, doing either action at a different frequency vs. the static equilibrium strategy will result in the same EV.
Oh cool so I can take every mixed 3bet/flat and just 3bet it at 100% frequency and not lose any EV vs equilibrium. Same with postflop raises and river calls vs jams too, time to up those all to 100%. Awesome.
Where do I sign up for your coaching you seem like you have a deep understanding of GTO?
Oh cool so I can take every mixed 3bet/flat and just 3bet it at 100% frequency and not lose any EV vs equilibrium. Same with postflop raises and river calls vs jams too, time to up those all to 100%. Awesome.
Where do I sign up for your coaching you seem like you have a deep understanding of GTO?
Yes, you can. Matt Marinelli talked about crushing midstakes on Ignition with like 30% 3-bet.
The only thing stopping you is your opponents ability to recognize what your imbalances and properly exploit it.
Oh cool so I can take every mixed 3bet/flat and just 3bet it at 100% frequency and not lose any EV vs equilibrium. Same with postflop raises and river calls vs jams too, time to up those all to 100%. Awesome.
Where do I sign up for your coaching you seem like you have a deep understanding of GTO?
Brokenstars and Zamadhi are correct. Mixing mistakes do not lose EV against a fixed GTO strategy, but they do leave you exposed to exploits.
The idea that you need to be perfectly balanced to prevent losing EV against GTO is a myth.
This is a consequence of the law of indifference; baked into the game theory of poker. If a hand mixes between two actions, those actions should (theoretically) have the same EV against your opponent's strategy. If you assume their strategy is fixed and unchanging, and continue to play your hand the same way against them, then it must yield the same value.
In other words, yes, you can take every mixed 3bet/flat and just 3bet it at 100% frequency without any EV loss against equilibrium. Yes, you can do the same with postflop raises and river calls.
Learn more:
- The Three Laws of Indifference
- Does your range affect your strategy?
- GTO vs. Exploitative Poker | Weekly Coac...
Tl;DR
Mixing mistakes do not lose EV against a fixed strategy, but they are exploitable. Pure mistakes are what lose EV against GTO.
Exceptions:
Spoiler
There is an exception to this rule in raked pots. For example, if you mix more continues against a GTO strategy, you will inflate the total rake paid. This extra cost comes out of the aggressor's pocket. So you can lower the EV of a GTO player by playing suboptimally, e.g. by calling more often, but the only one who gains is the house, not the suboptimal player.
Just putting it out there that the result for always playing rock, 100% of the time, vs a GTO rock-paper-scissors villain is:
Spoiler
You break even
Ok apologies to Brokenstars for the sarcasm, I was incorrect here.
Intuitively it seems like the massive switch in frequencies that would occur there eg BVB 10 -> 60% RFI can't be happening without substantial EV loss but I guess that's the f'd up magic of gto
Ok apologies to Brokenstars for the sarcasm, I was incorrect here.
Intuitively it seems like the massive switch in frequencies that would occur there eg BVB 10 -> 60% RFI can't be happening without substantial EV loss but I guess that's the f'd up magic of gto
You will get wrecked by a human with a brain because they will adjust. You will also get wrecked by nodelocked solves that exploit your leaks. You will not get wrecked by a static gto solution that doesn't adjust to account for your mistakes.
This is the core of the whole thread, actually. Saulo (and strong 500nl regs) can play somewhat reasonable approximations of gto that won't lose much ev against a bot. But being actively exploited by strong players is a completely different ballgame.
Just putting it out there that the result for always playing rock, 100% of the time, vs a GTO rock-paper-scissors villain is:
Spoiler
You break even
Haha, this is a good example. Gonna steal it myself in the future!
Isn't playing a 100% perfect GTO strategy unexploitable? Hypothetically if I played perfect GTO poker you'd have to play perfect GTO poker as well just to breakeven pre rake? What I'm trying to understand is how could high level players have significant edges over 500NL winners unless the 500 NL winners were playing no where near perfect GTO strategy. And if that's the case, following GTO especially at lower stakes is a bad strategy given basically no one is playing close to ideal GTO strat.
Isn't playing a 100% perfect GTO strategy unexploitable? Hypothetically if I played perfect GTO poker you'd have to play perfect GTO poker as well just to breakeven pre rake?
If you played perfect gto poker your opponent would have to play without making ev errors to breakeven pre rake. Your opponent could make a bunch of frequency mistakes (doing things too often/not often enough) without losing ev.
What I'm trying to understand is how could high level players have significant edges over 500NL winners unless the 500 NL winners were playing no where near perfect GTO strategy.
Decent+ human players are pretty good at not making ev errors too often. However, they will make a bunch of frequency errors that the world's best players could exploit for a lot. The world's best players will likely also use sizings and lines the solid regs are less studied in and will struggle against.
And if that's the case, following GTO especially at lower stakes is a bad strategy given basically no one is playing close to ideal GTO strat.
I don't think anyone advises to play like a gto bot at lowstakes. Learning and understanding gto is still very valuable for a slew of other reasons.
Oh cool so I can take every mixed 3bet/flat and just 3bet it at 100% frequency and not lose any EV vs equilibrium. Same with postflop raises and river calls vs jams too, time to up those all to 100%. Awesome.
Where do I sign up for your coaching you seem like you have a deep understanding of GTO?
Ok apologies to Brokenstars for the sarcasm, I was incorrect here.
Intuitively it seems like the massive switch in frequencies that would occur there eg BVB 10 -> 60% RFI can't be happening without substantial EV loss but I guess that's the f'd up magic of gto
No problem. It is a very common misconception.
i definitely learned something new, intuitively it doesn't make sense
The only thing I wonder is that sometimes the sim doesn't actually converge to a high enough accuracy to truly separate the mixed hands from the pure hands, so if a hand is like 99.99% one action and 0.01% the other GTOw would still consider the 0.01% an equal EV action, although it could be a problem with convergence. Just doesn't make sense to me that anyone would only lose 2bb vs GTO, but I need to do more research.
https://twitter.com/George_ymb/status/17...Saulo and YouMadBro had this discussion/argument on twitter.Personally as only a 500nl reg who could be wrong, I think YMB is right the best players would have a big winrate although he didn’t need to make the case in such a rude wayMy argument is theoretically if you add up all the different leaks a 500nl reg has like be
Oh well.
Yes, interesting developments. Though most of the volume wasn't 3-handed, see my last point.
There are a few things to take into account:
- he already plays highstakes with good volume in nl2k
- he's probably backed with part of the action, so that takes part of the pressure off, particularly since he could grind back in nl2k, and not NL500 (1/20)
- Another point I just thought - since he grinds a lot in nl1k and nl2k in the last few years, I'm guessing he has a lot of experience playing 6max with full tables. And most of the players who play high stakes are playing mostly 3 to 4 handed most of the time, so maybe they aren't as fine tuned as the best 6max players in EP...
And then there's the natural variance in small samples of course
Of course the samples are far too small to actually take seriously (and drug is one hellavu drug in the case of stefan), but at the very least you have to appreciate the cosmic comedy of it:
* Not only does a former saulo student win the whole thing
* Linus ends up dead last
* Stefan ended dead last in the nl500 championship
"Fate, it seems, is not without a sense of irony"
Obviously they were talking when steffan was peak and not a shadow of himself and 3h is vastly different than 6h
