[2-4]: ATs OOP in a 3bet Pot

About the bolded part, I have two comments.

One is just, is there a better public forum, or do you just mean that private groups could be higher quality?
And the second, imE a really important factor about the usefulness of posting online (which I think you're neglecting) is that it forces you to justify what you did. In coding, there's this thing called Rubber Duck Debugging, which... well I'll just quote the introduction

So the idea is that just explaining yourself is actually already helpful even if you have 0 useful feedback, and you will have to do this if you post online. And in practice I think this actually works even better in a forum because even bad replies will force you to think about aspects of the hand that you wouldn't have thought if you were trying to use a rubber duck strategy for poker hands. So yeah I think the usefulness of posting hands actually punches significantly above the average response quality.

I think almost every hand I've posted so far felt like it's been worth it in terms of pure strategy improvement (... whether it's worth the tradeoff of having a lot of people be rude to me is a different question), with maybe one or two exceptions (probably including this hand). The AQ hand I mentioned was a great example of this because I found the thread to be extremely helpful despite still thinking almost all replies were wrong.

@Everything else: well I find this list very interesting even if I don't yet agree. Drilling the solver for particular spots you encountered in your hands is definitely much better than looking at GTO in the abstract (because it solves the problem of it seeming unrelated to your play and then you forgetting it before you can ever apply). I probably should do it at least more than i do it right now.

My final point I guess is just a sort of abstract concern I have that focusing on the solver all the time will train you toward really bad habits where you automatically start to think about every spot in terms of a solver, which I think is a very unhealthy way to approach a game where the majority of your opponents are super weak. This is also my concern about online play because online people will play much better, so it trains you to play against strong players, which is a different mode of thinking (and I think defaulting to that is quite bad, like I think if you get into the habit of treating your opponents like online players, you're going to leave sooo much value on the table).

I got this impression with Jonathan Little a lot. I remember watching an old webinar before GTO was a thing where he talked for several hours about cash games, and it seemed super sharp and practically oriented. Whereas nowadays everything he releases is incredibly formulaic; it's just picking some kind of spot and then looking at a bunch of GTO charts. It just seems so lazy and uninvested. Which I guess works if you're playing super high stakes, but way degrades usefulness for soft live games, imo.

sorry third reply, but this just seems extremely unintuitive to me. ImE (and I do have a decent amount of experience playing other games on a decently high level) just grinding is already not a good strategy in most games, but relatively more useful in other games. Like, in chess or starcraft, I don't think grinding all day is optimal for improving, but at least it will help some (with pattern recognition and raw mechanics, respectively). But in poker specifically, it seems to me it's all about quality over quantity, and playing 200 hands but writing five of them down and then reviewing them is way more helpful than playing 10k hands on autopilot. I honestly don't know if playing 10k hands on autopilot will have any training benefit at all.

Getting a data base is useful, that's sort of separate thing, but even then not clear how much you can apply any results from your hands to live play.

Right, yeah, against strong players specifically (which Villain from this hand was), I mostly agree. I definitely should do more GTO grinding to be better against regs. (In fact I'm making a mental note right now to do that in the future.) But again, that's a pretty small proportion of live hands.

i think its an interesting conclusion to come to that getting better at poker will cause you to play worse against bad players

most of this stuff (grinding, database leak, drilling, solver output to a much lesser extent - that one is good to learn poker theory) is oriented around finding your leaks and giving you a path for direct improvement. also grinding bunch of hands gives stealth mental game benefits

end of the day re 97 you spent how long to conclude what? that if you give precise assumptions you can alter the ev output to suggest whatever action you want. would whatever time you spent on that have been better off on a more recurring node (possibly 3b pots oop!)? that was kind of my point. its like a hyper specific line / node lock and it took ages and assumptions you can't possibly ever test. also i think solvers can node lock multiple streets but i have never tried (apparently pio can but not gtowizard).

i like the rubber duck theory and thats how i justify my time on this forum to myself too but realistically this place exists for community / egoic validation. basically everyone in the world is doing some form of everything mentioned here if they are good / aspiring to be good at any variant of poker. what are the chances you are right about the improvement process and everyone else is wrong lol

So I think the biggest thing that gets miscommunicated in these discussions is that a lot of related but distinct terms get collapsed into one concept: GTO = theory = doing something with a solver = math. Honestly, you can even throw some modes of speech in there, or like people talking about realization are probably the "GTO people" whereas people using implied odds and RIO are the exploit people or whatever. Even nodelocking and MDA get lumped in even though they're quite literally the opposite of GTO.

Someone in another HH just put out an open question to all the GTO folks in the thread, and I think it's intended for me, but my entire justification in that thread was exploitative so I'm not even sure how to answer the question.

I don't think the big divide is over methods or GTO vs exploitation. If you pay close attention to the back-and-forth between the "solver people" you'll notice that's what we're arguing about amongst ourselves.

The big divide between the "solver people" and others is over whether or not you expose your priors to some level of falsifiability and rigor. An equilibrium sim is one way to do that; a nodelocked sim is another way to do it; MDA is yet another; running an EV calc based on some rudimentary assumptions is another; building a custom script to do the same thing is another; etc.

(FWIW, HH review does expose your priors to some degree of falsifiability, though you'll get so many contradictory responses that you'll likely be able to validate your priors one way or another. And the degree of rigor is, um... YMMV.)

I think you squarely fall on the side of people who has some standard of rigor--in fact, I think your misgivings about a lot of modes of analysis actually come from your standards of rigor being (IMO) TOO high.

I think yours and submersible's disagreement over building a custom script is far narrower than you're taking it. I really just think he's saying building your own script is inefficient because solvers have built in functionality to do that for you, which is called nodelocking.

Which I think brings us to why all of the people who put a premium on falsifiability and rigor are all "solver people" to some degree or another even if they have differing perspective on methods or theory vs practice: because solvers are really powerful all-in-one tools that have subsumed a lot of tools of yore. You could never run a sim, and still get everything out of a solver that you get out of PokerStove, Flopzilla, etc.

Is the argument that the grinding helps the previous lessons from the solver work (which otherwise got forgotten, per your earlier post, and I agree) sink in, and become rote?

AIUI, is the efficient learning pathway:
1) Do solver work, especially in common areas like CO v blinds, LJ/HJ v BU, etc, and on common boards. Which will establish a baseline heuristic and plug most of the glaring leaks.

2) Grind micros OL where possible, to instill the lessons and observe common deviations in the player pool.

3) Evaluate observed and hypothetical deviations with node-locked solver work, with a mind to trying to determine which V characteristics must be present for the deviations to be preferable to the GTO baseline.

Learning is supposed to be painful. I don't mind making myself look like a dumbass in front of all of you, so long as I can learn from it, and I appreciate the contributions from those of you far more studied. Thanks!

Another good post, in a thread full of them. Just chiming in to agree on the "Explaining yourself," concept with the observation that, in study groups way back in the long ago Dreamtime of undergrad and grad school, I didn't learn nearly so much hearing someone explain something, as I did when I tried to systematically and slowly break a concept down as I taught someone else.

It's easy to jump to a conclusion for a given hand's set of decisions posted here. Much more difficult to break down and explain the logic behind the decision, and frequently in doing so, I find I got the concept backasswards.

Right, and I was mostly commenting on this thread 😀

Although I think it can be generally useful to apply GTO to hand review of most HU (i.e. easily solved) situations....

This is where nodelocking or just analysis is useful. No, your Villain has probably never studied a GTO postflop spot, but either:

(a) you have legitimately no idea what Villain is doing, in which case you should be trying to play balanced GTO because it's unexploitable, or
(b) you have some idea of what Villain's strategy will be, and you can compare it to GTO and/or nodelock V's responses to see what the maximally exploitative response would be

As a very hand-wavy example, range betting most/all flops as the preflop raiser IP is not actually a GTO strategy, but much of the reason comes from OOP's x/r range and if you node-lock OOP to x/r never (or only a small number of very strong hands), range bets suddenly become attractive. Conversely, if you node lock IP to range bet with impunity, OOP suddenly starts x/r very aggressively. Real players won't do this. This is the kind of information you can get that "isn't GTO" but comes from studying with a solver.

Maybe 12 hours total to conclude that I tend to overvalue slowplaying nutted hands. Which I think is pretty generalizable. And I do tend to be more pro slowplaying than almost everyone else; I've noticed that pattern a bunch when talking about hands with other people. So roughly speaking, now whenever I flop a really strong hand, I ask myself how much of that lesson applies and try to adjust my estimate, and maybe I decide to fastplay more because of that.

But I got the action I didn't want! I spend two pages in a thread arguing for flatting the Turn even though everyone was telling me how wrong I was, and then the simulation concluded that I was wrong. So I don't really understand how you can accuse me of tweaking the assumptions to get the result I wanted :(

I guess two things

(1) You have to be careful not to conflate these things if you want to argue against the position I actually have. You lumped statistical analysis in there with GTO, but I'm completely pro statistical analysis, like finding out which spots are overfolded by the population and such, and also finding statistical leaks in your own games. This is compatible with being skeptical of GTO.

(2) I don't follow that many pro players and ~never watch interviews. I consume a lot of material from a few players, rather than a bit of material from a lot of different player. The one who I've spent most time watching is stuff his Alexander Fitzgerald, who talks about statistical stuff all the time (like this spot is overfolded, this line is underplayed, etc.) but almost never about GTO. So I guess my point is, to the extent that my view is contrarian, it's definitely not on purpose.

But I'm also not really going to change it based on it being a minority view, that doesn't seem like a good idea.

This is the least important point, but I reject this framing. You're equating playing GTO with being good at poker. But game theory optimal is just a misnomer. The GTO strategy is the strategy with the property that no other strategy has +EV against it. That's it. That doesn't make it optimal, or right, or "good" poker. In fact it's literally only optimal against itself.

I'm definitely much more sympathetic to the "grind a lot" advice with this context. Like if you start by wanting to learn a certain strategy, and then treat grinding as a tool for getting a lot of reps with the strategy, then it makes much more sense that it actually helps. In this case, I would imagine that your playing also includes frequent looks back at the strategy in between hands.

I mean I think the unique issue with poker is that playing itself doesn't give you a lot of feedback about what you're doing wrong, which is why I'm so skeptical about just grinding. If you don't combine it with first looking at solver stuff, then you could arguably play the same situation 20 times, and just do it wrong 20 times, and don't get away anything. But your 1)-3) pathway mostly fixes this.

too much back and forth. i suggested ~8 things you could do to improve and you told me you can't do any of them because {reasons} and the players you play against are so bad that improving at poker would make you play worse vs them.

meanwhile 12 hours to figure out you need to build a pot when you have the nuts vs these horrible players that you have figured out so well you dont want to jeopardize your understanding of poker against

re podcasts, mechanics of poker is quite good and may help you to break out of whatever echo chamber of 1 you are currently in (mmasherdog did one recently that id recommend)

:c

Just a "quick" clarification that GTO isn't just about avoiding exploitation; optimality is in fact a huge part of it.

If there's no money in a pot and someone blind raises, you can just fold 100% of hands without being exploited or allowing any strategy of theirs whatsoever to be +EV. If you plug that scenario into a solver, it will not simply find that banal unexploitable strategy. It will maximize EV gains while minimizing EV loss in the worst-case scenario.

In other words, GTO is a minimax strategy. If you were to 3D map out all possible strategies against all possible opponents, GTO is the saddle point. There are other localized maxima, some of which are relatively broad-based where you can attain the EV gains even with relatively large margins of error on where precisely you are in the distribution, but others are blips surrounded by moats of despair.

So in an environment where you know nothing, GTO is optimal. (Of course we do know quite a lot about our opponents, hence why we target other localized maxima, blah-blah-blah, this was just supposed to be a quick clarification, not address all points at once.)

Another example of a theoretical construct that provides contingently optimal strategies is the natural bet sizing table we talked about elsewhere. Even if you presume you don't need to balance your various bet sizes, you still need to know how big you can bet before you start value owning yourself. The truth is no one really knows the biggest size a fish will call with a capped range on some node/board/config combination that for all I know has never been played on before, so the theoretically optimal bet size for the hand you happen to hold is a pretty good baseline to work off of!

In other words, this construct doesn't just keep you from getting exploited; you're IP OTR with SDV, you can just check it back if you don't want to get exploited! The theory helps you win as big a pot as possible without getting exploited.

I've been playing since the 2000s and I can tell you no one in the universe was guessing they could bet sizes anywhere near what's on that table just based on population reads.

Yeah so, given the last few replies in this thread, I'm not sure this applies to this debate in particular, it kinda does feel like there is at least also a real disagreement about the usefulness of GTO.

But, if if we zoom out, this makes a lot of sense to me as a general axis along which people debate. And I can definitely see how from the perspective of a "solver person", it's easy to lump hopefully-genuinely-valid-critiques-of-GTO together with general-excuses-not-to-update-to-protect-your-own-ego, which I do think are incredibly common in poker discussions. Like if 90% of the time a critique of GTO is just an excuse, then it's natural to just round up and assume all skepticism of the solver is BS. ... but, I do think that's a big mistake.

right, except that node-locking still assumes Villain plays GTO past a certain point. I mean letting the solver pick a strategy for Villain is the entire purpose of using a solver; if you hand-pick all strategy from Villain, then Hero's strategy becomes a straight-forward math problem that you don't need a solver for anymore. Which I think relates to the next point

yea agree with this. part of this goes back to what I said earlier, which is that I think the absolute skill required in poker is actually super low compared to every other (e)sport. Like, you can learn how to be a decently winning live player in a year. That's crazy! Unless you're 1-in-100-million talented, you definitely can't do that in chess. but the flip side is that improving is more difficult than in any other game. so I think that view justifies having super high standards of rigor.

But I also agree with the second part, at least with respect to nodelocking; I'm totally sold on this being useful and will start doing it a lot going forward. Even if it doesn't tell you everything about what you should do in a spot, it's just a really cheap way to get an at-least-reasonable answer.

yea I think this is true as well, at least among the set of things that are useful at all. Which is a pretty important caveat though because I think most people spend most or even virtually all their time just doing things that aren't useful at all, like mindlessly playing or watching pro games.

(I kind still feel like it's better than all the above? At least initially? I definitely never felt like I was learning as much as during a good hand review.)

But regardless, you're right about the diminishing returns. My hand reviews nowadays are much less useful than they were 4 months ago. It used to be that we'd find that I grossly misplayed about half a dozen spots each session, and now it's more like, one spot is grossly wrong, maybe three spots were suboptimal, another three spots are arguable and I'm still not sure afterward, or something like that. Which is actually the main reason why I'm more interested now in expanding more to other forms of study.

Sorry if you feel like this was a bait and switch; I guess we all agree that GTO has some degree of usefulness, but usefulness isn't a binary. Two people who agree something is useful could have a higher degree of difference over how useful it is than the difference between someone who thinks its a little useful and someone who thinks it's totally useless.

Maybe that sounds pedantic, but I've been having the same argument for 10 years and I can assure you the same people tune me out when I say "the population tendency on this node is to overfold" just as much as when I say "I plugged this hand into PIO and said to bluff this combo." "Node, " "overfold" and "population tendency" are modes of speech that imply that some math stuff is going on, and math = theory = solver = GTO, aaahh, begone with you!

You may be asking yourself "How does any of this affect me?" It doesn't. I'm just turning this into my own personal therapy session. Carry on with the thread.

You can nodelock to your heart's content with any real solver. GTOw is more a repository of sims than it is an actual solver, which is why you the functionality is quite different (for better and for worse).

The main reasons people don't nodelock every node is because it's tedious, but it takes less than 12 hours to do 😃

If/when you get your hands on a solver and you dig your fingers as deep in the dirt as you do with your own NE calcs / scripts / etc, then you'll be shocked how quickly you become the world's foremost expert on exploiting certain tendencies. (Or at least in the public-facing anglo-phone world). HMU when that happens, I wanna learn what you learn.

I mean, "straight-forward" is a load-bearing word here. Everything PokerStove and Flopzilla do are far more straight-forward than calculating EV in any multi-street scenario, and those are still extremely useful tools.

(I kind still feel like it's better than all the above? At least initially? I definitely never felt like I was learning as much as during a good hand review.)

But regardless, you're right about the diminishing returns. My hand reviews nowadays are much less useful than they were 4 months ago. It used to be that we'd find that I grossly misplayed about half a dozen spots each session, and now it's more like, one spot is grossly wrong, maybe three spots were suboptimal, another three spots are arguable and I'm still not sure afterward, or something like that. Which is actually the main reason why I'm more interested now in expanding more to other forms of study.[/QUOTE]

Agreed on all fronts here, and I'm glad you've found it useful.

Maybe it'd be more useful to say HH review has a harder ceiling than the other stuff listed.

But maybe that isn't even true. I lowkey feel like StupidBanana's gonna be ruling the world in no time off of nothing but being the King of LLSNL HH reviews.

Maybe there's truly no such thing as pointless study. Somewhere out there, the King/Queen of Poker Vlogs has an Anki account full of every exploit they've ever learned from ChicagoJoey and they're absolutely smoking 5/T with them.

What makes a lot of modes of "study" less useful than others is the passivity and lack of absorption. I wouldn't stake my money against anyone who's actually putting the work in, whatever that work is. The truly appalling thing is how little actual, active, engaged study anyone does (at least any public-facing people in the anglophone world, etc). Myself included tbqh.

I don't know if this contradicts anything I've previously said (probably) and I don't even know if I believe everything I'm typing (probably not), but these are some arguments I'm open to.

RaiseAnnounced is the poster i see when i look in the mirror