Well this is what I mean to point out, that was unable to be discussed I think years ago. What does it mean that she was a top pro (lets hypothetically only consider mtt play) in a game in which you cannot prove that a good winrate is skill based.

Rather she was a POKER STARS pro which means something completely different. What she was, was part of a group, that sold to the players that higher rake is better and creates more profitable games.

But this was in comparison to her experience or other stars pros who were making their income not from the profitability of the games. It's dramatically different to call yourself a poker pro and to show up at a live even as a phenomenon, and be a marketing investment, who doesn't really understand the game (like phil helmuth). Does phil understand the game when he flies off the handle at how his opponent plays and what cards his opponent has? This isn't how the game works, thats not considering range at all.

He's selling himself as a product and thus he can play games that are unprofitable for himself and yet make money. But its not profits from playing profitable poker strategy.

Thanks for that clarification, I think I got your point now. You're saying that no matter how big the sample, you cannot guarantee someone is a winning player and thus a good coach, ignoring their coaching abilities etc? Your argument is there is no correlation between winrate and specifically HUD-reading skill, as you have a very large database and it is still way too small to draw conclusions from.

Assume a spot in an SNG where you have a short stack. A random decision vs a GTO shoving range-based decision will do worse in that spot and we do not need a sample size to determine that. There is a correlation between which of these two decisions you make and your winrate. That also means there is a correlation between winrate and being better/worse at poker and being able to say someone is better than someone else is all you need to determine someone can be a better coach than someone else. None of this relies on any database-driven decisions, I'm simply concluding that the person who clicks random buttons will very likely have a losing winrate and that the GTO-approach will very likely have a positive winrate.

Now, perhaps your point is that it is only possible to state someone 'likely' has a higher or loser winrate and that their real winrate is subject to variance. I agree with this and what you're saying is that there can be no source, such as long-term winrate, that with perfect authority can truly prove someone is a winning player. That's true, but I fail to see how it does not give an indication with a degree of confidence. the -50% roi player is more likely to be bad and the +50% roi player is more likely to be good. That's because the +50% player is the above GTO-based player and the -50% clicks random buttons, as I attempted to correlate above.

There's no such thing as an authority on knowledge, so you may as well take the best possible predictor at hand, so that you pick a coach that is more likely good than bad, as even though there is no definitive authority that can claim someone is winning, it's better to take the action that you have a higher confidence in than a random one.

Do you agree with this?

Yes we are speaking to the same direction. I'm not complaining that no matter how big the sample you can't be perfectly confident tho. To be clear Im saying that in my games you could only reasonably play 30k games a year and its nowhere near a reasonable sample size for confidence in regard to the winrate observed.

I'm suggesting anyone claiming their winrate in such games makes them a good player is actually showing they have no proper understanding of the math.

Now as to your point, I think you perfectly understand, except now speak to a general idea in which the players trend towards optimal poker. ie there is no one who is comparatively -50% less good than the other players. Not sure how to say this. But the observation changes as the game gets harder or the field tends towards GTO.

And yes there are spots that we can identify as being solved such that we can judge a strategy but we are intrinsically implying that we can't do this for all spots else it would be difficult to argue there is any profitability left in game with an accessible solution.

This thread is dominated by 1 person making 90% of the posts/ Like wtf?

My only disagreement remaining is that I am using that simple spot to imply you can do this for all spots, given poker is a game with finite possiblities and the person who trends the closest towards this strategy will likely have a better winrate. I am also saying there is an accessible solution to poker (GTOwizard for a mere $90 a month, or run your own sims); this is very much the reason winrates become lower and lower, and over time you are right there is no profitability left. That's entirely accurate, but we're not at that point yet as not every player is implementing that solution, despite it being accessible. I think you can make a similar argument for players that play exploitative, but the GTO one disproves the argument of winrate being no predictor nicely and we only need one argument against it to prove it inaccurate, so I will stick to that one. Furthermore, perhaps my argument of 50% worse is excessive; however even 1% better or worse proves my point. As long as you agree people can be better or worse than another, we will end up with the conclusion that winrate is an indicator of skill, albeit no perfect predictor.

Given the better players will more likely trend towards the higher winrate as we established before, that means your best bet for a coach is still one that has a positive winrate. The winrate is an indication towards being a better player and therefore if you are looking for a coach, you should still judge by winrate, keeping in mind that it is not a guarantee that the person showing the winrate is indeed that good.

I agree with you that the coach's claim ought to be 'I am likely a good player given my winrate' and not 'I am a 100% good player given my winrate'. I disagree that using one's winrate as indicator towards being good at poker is wrong.

well yes obviously I'm trying to be like those gimmick accounts that come on with a weird framework but maybe that gets players talking about something that was previously taboo. I'm just trying to fan the sparks a little until someone comes in an says 'there is no way your field was such that there wasn't a large enough sample available to tell who were the winning players versus the losing players'. And then hopefully someone more knowledgeable actually shows how the math works.

But its sort of a difficult thing to discuss with the community because if its true, and we don't understand it as a group, then we are full of people that are guilty of this 'leak'. It means they are those coaches claiming their winrate proves their knowledge when actually making that claim proves the opposite.




Have you spent time analyzing the available data and working with relevant variance formulas tho or is that just pure sentiment? My understanding is that you absolutely can't use winrate to determine who is a good player or not, in my field, and its a very telling error to believe that you can.

"Backwards corollary?" Do you mean contrapositive?

For what it's worth, your question was "can this coach be good at teaching profitable poker," and the answer is yes. Not all of them will, but some can. Notice that you could replace the phrase "this coach" with almost anything, and the answer will remain yes.

I gotta agree with this. There's minimal math involved (just a little arithmetic) for most people. Then again, I notice that I'm agreeing with every smoothcriminal reply in this thread.

Consider blackjack instead of poker for a moment. Someone had to work out the probabilities to determine basic strategy. That person had some mathematical or statistical expertise. But a person using the BS card in the lobby needs no such expertise. A person learning BS needs no skills at the maths.

Now move back to poker: the majority of players are more akin to the person reading the basic strategy than they are to the person who first wrote it. Someone else did the math – the extreme example being those who worked out the GTO solutions – and the rest of us now take a ride on that.

* * * *

To the OP, your discussion is interesting, but can you clarify one part of your post? Specifically, this:

What were the figures for this? For that matter, who were the "top players" for whom you examined results? (Not looking for names, just trying to determine the sample you used, especially since you felt the need to put the word "top" into quotes.)

Also, when you say Katie Dozier "wasn't winning much," are you talking about SNGs in particular? Besides, being married to an author on SNGs doesn't mean she herself will be similarly adept at the format.

You're asking us to accept your conclusions, while providing us very little in terms of the reliability of your premise.

I think so ;p


There is a nuance here tho. BJ is solved and much more just a cheat sheet. I AM referencing short stack/turbo poker for my experience so between 0 and 75 bbs kinda. But that game is not really solved. And as far as the strategies that coaches will teach, they are NOT teaching players to play the solution.

* * * *

To the OP, your discussion is interesting, but can you clarify one part of your post? Specifically, this:


Ya. Its sort of purposeful. I think you can use any field especially for mtt and mttsngs. Maybe start with Katie's stats. Yes just because she partned with a guy that wrote the book on sng's doesn't necessarily mean she should be the top. But it's low/mid stakes. It was one of many flags for me. I also play with players with sort of silly winrates who were silly players, I was one of them at one point. I thought my winrate meant I was awesome. So I started to read strategy by these poker stars and full tilt pros. Being unpredictable and crazy is best. Being able to play any hand from any where is hard to play against they say.

It doesn't make any sense. But it seems to, until you consider what I referenced in the op as a new line for observation. Consider that the entire field is naturally trending towards optimal poker. Then let us discuss what we can deduce. I conjecture that we can extrapolate something that hasn't previously been brought to light.

(by "top" players its hard to distinguish between players that play the best poker, players that are thought of by the community to play the best poker, and players that have the highest observed winrate)

Hi! I don't know if this is some elaborate bait or I just don't understand the points you are making, but I have a background in statistics so I'll try to explain.

If a player has a 25% roi in 100k 180-mans, the 99.7% confidence interval for their winrate bottoms out at 20.39%. If they coach a player to play exactly like them, it is overwhelmingly likely that the player they are coaching will become a winning player. Your sample size point is well-taken in live MTTs or streamed cash games, where we need to infer some element of skill based on play, but I think your understanding of variance is off by a few orders of magnitude.

I actually think the exact opposite of what you are claiming is the case. If a player with 25% ROI over 100k games is telling you to make a play that is non-solver approved, the most likely explanation is that it is a good exploit vs. population, not that they are a bad/losing player.

In your OP, you claim that having a high winrate doesn't mean you are more likely to be a player who wins at a high rate. This is false, and the whole point of variance calculators is to quantify the uncertainty in winrate.

My point is more if the player sees your post and says, 'oh tmanto ya they think they know about variance but they don't understand how variance works', and then they have a high winrate, maybe 25% over 100k games, and you are suggesting you would have confidence in their ability to coach. I'm suggesting they might be a natural player, but that they clearly can't understand the decisions they are making and why.

If their student played exactly like them sure...but how does a player ignorant to the variance implications of their winrate pass an otherwise accidental strategy on?

That only seems slightly high to me at a glance. But 100k as a sample and 25% roi I believe are too high. The players in that field didn't really reasonably have access to that large of a sample I think (except over about 3 or more years time which I believe ran back to blackfriday) and perhaps I'll be corrected but 25% wasn't really near a sustainable winrate. I do remember some players that had such high winrates but only over say 5k games believing that the higher their winrate the less of a sample one needs (its true to an extent but looking at the sharkscope data I remember I was kinda dumbfounded).

I guess we could ask who in the field could have reasonable confidence in their winrate that they were making a reasonable profit, because A LOT of players were claiming to. There were many stables and coaches (maybe still are).

I didn't mean to say something that disagrees with what you said here. More about what you implied with 100k and 25% roi. Maybe I'm mistaken on the available sample and the necessary sample for reasonable confidence. I might have looked at the math wrong. But I took time to look at the actual available numbers and felt like the general population misunderstood.

Here is 2014 for example mttsng 5-15$


I can't remember the parameters here perfectly but many people know them. The 15 dollar games ran way less, and you could bink a few and be waaay ahead for the year.

also poker stars changed the stakes/buyins so the games became not perfectly comparable (or this could be something we could consider in general happening).

But my point is players could have played more games than are on that leaderboard but you can't play waaaayy more games.

Then you can say, well each of these players could get a 100k sample over a few years. But if all these players played 30k games, their average roi %'s would naturally drop (assuming in general top multi tabling regs are better than recs etc).

The 180 man games I think had the greatest volume/liquidity. I can't remember why I am claiming that this extends down to snggos with less players and tables, have to concede that isn't true for now. Obviously it is greatly so for mtts.

tmanto is right... I have a feeling we're getting leveled in a major way here, and I'm falling for it.

Okay, that was a giant whoooosh. The point I and others posited is that a person does not need to be strong in math to be a winning poker player.

Let me make it easier:

1. The people who figured out a basic strategy for winning at Game X used some kind of math to do it.
2. The people who use the aforementioned basic strategy for Game X could be someone with little to no math skills. They're just following instructions that the first people create.

Indeed, Game X could be blackjack. It could also be Go, Stratego, Connect Four, checkers, backgammon, chess or goddamn Battleship for all I care. It doesn't matter if the game is solved, somewhat solved, or unsolved. The point is that the people who develop the strategies are the ones using some level of mathematics (obviously higher for some games than others). The people implementing said strategies can do so with minimal use of math expertise.

Poker is the same way. You do not need to be highly skilled at math to win at poker, because someone who is good at math has already done that part for you.

But what stats did you use for Katie? Her SNG stats specifically? Collin's book is about SNGs. If you looked up Katie's MTT or cash results (or data confounded by MTT and cash results), then your conclusion is meaningless. If you only looked up her SNG results, then your conclusion is, well, she might not heed the advice of her husband.

Not for nothing, but Gisele Bundchen probably can't throw a football 20 yards. Yet if her husband was giving me tips on throwing a tight spiral, I might still listen.

You didn't answer an important question, which was this:

If you examine the actual top players, and not just the people who are "thought of" as the best, are the win rates more consistent?

I ask the first question (and the one re-quoted above) because I'm not sure what you mean by "consistent" for the purposes of this discussion. Take any player with any reasonable amount of volume, winner or loser, and their graphs swing all over the place - especially in the short run. But overall we see a general trajectory that is more informative.

Gisele Budchen is so athletic looking though. I'm 5'9" which is model height about 1 inch shorter than kate upton and not that i'm super sporty or athletic but I can throw a football pretty far. I don't really know how far 20 yards is but I can throw a football across my entire backyard.

Just trying to help here. 20 yards = 60 feet. 60 feet is close to 9.5 Tom Bradys laid head to toe.

She walked away with millions in net profit and she isn't giving it back.

OP You're being incredibly reductive. This one comment proves that. You're describing, in lay terms, the single biggest argument in the poker world. It's almost like you're not aware that this is an ongoing, as of yet unresolved debate, with opinions scattered across the extremes, each camp having stalwart absolutists as well as advocates for reconciliation. At this point though, your closing statement is simply not true. Poker does not it have a GTO strategy, because, with the exception of heads up limit Holdem, Poker is not solved. What it has is the development of strategies that over time come closer in their approximation of GTO. This does not mean that the player pool is doing the same thing. This isn't to say that a higher percentage of successful players aren't embracing GTO , because they are, but this doesn't reflect the overall player pool. Exploit/adjust is definitely still the default skillet being used (profitably or not) by the majority of players even trying to employ skills, and as long as the game is beatable, this will necessarily be the case. This isn't a guess, it's an observation.

As to your main point about observed WR being incapable to tell if one is a winner or not, this is also a solved wheel. We can estimate, with varying levels of confidence, the likelyhood of our true WR being within a variety of different margins. Because of this, this is a point where our results can tell us the likelyhood of our being winners or not. And somehow this has to do with it being uncool that somebody promoted the idea of rake being good.

How does any of this reductive repetition make Selbt's comment invalid? Yeah, the two finalized are better then her at the moment as she doesn't keep with the curb, but her point is still valid. Attenborough definitely plays very exploitative, situational poker. His Norwegian opponent didn't try to outplay him. Idk if she was criticizing this and saying he should have. It was probably better that he didn't cause the Aussie ended up outplaying himself. This still leaves pages of repeated red herrings, irrelevant venting, and a ton confirmation bias. Honestly I still don't get what OP is saying (aside from the many unrelated things youre saying). Oh yeah and this totally proves that OJ did it, because Tai chi is not a real martial art and gree tea isnt ad healthy is it pretends it is.

Whoa ok I just said the same things a bunch of other people said. This is a huge troll, there's a bet on how many people will see this, explain specific things, and not address the hoax.

I don't want us to speak conjecturally when we have data. 100k games as a sample and a 25% rio is irrelevant. There is no such sample for 180 mans mttsngs.

I think I had around 100k games with say 14% roi. But the games had changed because that was over 3 years period. I determined I might have been trending (significantly) towards being a break even player.
To look at 100k game sample and have 90%+ confidence in your winrate is not really correct. There are other very relevant factors.



I was looking at katie's stats for sng's she played a lot in my field. I think this doesn't really apply to physical sports. I don't know what excuses we want to accept that the partner of the guy that 'wrote the book' on sng's doesn't have strong results in low/mid stakes mttsng environment. Its just one player. I have zero counter examples in the data I looked at.

I looked at mttsng's from around the year of what I posted there (2014).

I'm commenting on the ACTUAL available data and attainable sample sizes. Counterpoints shouldn't really be theoretical in that sense (ie 100k games with 25% roi doesn't help us when no one can achieve that.)

..]

It's not a troll in the sense posters are accusing. I took the data, and I look at it all the best I could. I didn't just think in my head like you guys are doing.

Can someone take the data and show me what I did wrong? I'll start playing poker again maybe.

Lemme put it another way. Can someone here please take a set of top player results from a year of mttsng 180's around 2014. Combine their total games played and average their rio's and tell me if the game seems profitable from the result.

I did this experiment. But we don't like my math. Can someone show the result?

How would we determine the top players? Or you mean highest actual winners?

There is no consistency in winrates yoy. If you ACTUALLY look at the data its silly to consider.

when I played 180's I had a 30k month and then almost a losing year rofl. I remember for a year I was the top or top 3 180 player then I just ****ing lost it for one reason another . While I certainly had some mental collapses from huge downswings and the format changed a bit for the worse I understand what your saying the variance was silly and there was always a new "best 180 player" Variance is just insane epically in 8bb turbo 180 mans.

Its truly impossible to know your winrate in some formats because you just change your game too much often for the worse like I almost certainly did to go from the top of the world to a shitty depressed **** reg. then you got 179 other people changing

I remember my strategy was to be far more aggressive then what is considered gto push/fold now or even then. It also allowed me to play more games. any ev I lost from shoving too wide was probably made back from keeping fold equity for guys who didn't know how light they needed to call and building stacks to run over final tables in a super top heavy format. This worked great after black friday for a couple of years I steam roll alot of ft and get alot of top 2's which is all that mattered at the time.

Years later fish are now calling their 3bb stacks from the bb with any 2. There are more % regs in the field and they know my ranges, they correctly call me light from experience of playing with me for 2-4 years now I am just punting by using the same style and everyone knows to call me light. Finally they change the format from 18-27? Now my strat makes no sense because its no top heavy. Do I change my game to adjust? no why would i sharkscope says I have 20%roi but really I am punting in the current game?

Do people pay me money to coach? sure why not I have a 20% roi! now we have more punters.

how many people saw Venessa's terrible preflop play and thought that was sometimes ok in poker? meme

Alot of what he said is true. winrates were never revealed

Sorry, but there is no point here. Even if winrates are not provable with certainty, it's some of the best data available. You cannot seriously argue that if you are picking a coach, you are as well off selecting a random punter in the casino as you would be selecting someone who has an historically high winrate.

Its just stunning to me that otherwise intelligent people don't grasp the basic concept of edge in any game.

(1) What is my edge over the field?
(2) What is the total rake (in terms of rake, expenses and tax implications)
(3) Can 1 overcome 2 to any degree to make a decent living for my troubles?

The vast majority of otherwise sane individuals seem to overestimate their edges in events where they are playing shallow stacks and you have to win a lot of flips!
How can a person possibly have a huge edge in such circumstances. Its just not possible. Its purely high variance/low edge.
AND.... you still have to overcome rake and expenses.

Online you have the additional problems of playing against Bots, RTA and collusion.

IMHO the vast majority of people would be better off doing something else than trying to crack this nut.
I'm sure someone will post the stats from a +4SD sun runner to counter my points.

Feel free to ignore all the above.

This speaks to the point, its part of it. One could bink a few 180's in a higher tier and then sort of stop loss all the way down to the bottom of the tier and then try to justify a sample size. I think people would be surprised how often this happens (or how many players here have done that)


The language here can be nuanced. It's not that winrates aren't provable with certainty. That's just obvious. Rather, if you look at the certainty based on the available and reasonably possible sample sizes, the results are unusable.

I don't know how you are differentiating between a random punter in a casino and someone with a historically high winrate. That would be a horrible thing to use to select a coach. I actually am flabbergasted.

I have two points and point B is contingent on point A. I don't want to argue point B unless we agree on point A

A) The available/achievable sample sizes in the most liquid games around 2014 are such that winrates cannot be used to determine skill
B) Not understanding/knowing this turns into improper strategy at the table.

You need to get back on your meds, man. Your response to my post is literally just a jumble of words that mean very little. It's worse than postmodern philosophy. If your position is literally that I am equally well off hiring some drooler in my local 8/16 game as a coach who has been losing for decades than I would be hiring someone who has been a consistent winner for decades, you're beyond hope and not at all connected to reality.

Jumping in becaue this sounds like fun and there's a couple pet peeve misconceptions being brought up here. Feel free to ignore.
There's a qualifier that makes this statement iffy. What does 'reasonable' mean in this context?
Of course you can use Chebyshev's inequality and see to what confidence level does an (observed) high win rate correspond with a true edge. While you can never say it with 100% certainty you can get arbitrarily high in this confidence given a large enough sample.That is why when evaluating someone's claim you need both: win rate AND sample size. The cutoff for how you determine the probabilities as acceptable is up to you. There is no mathematical standard for it. (Many scientififc papers use 95% or 90%..but those are arbitrarily chosen and are being used by convention - not for any fundamental reason)

On the flip side the entire premise is flawed for two reasons:

1) (as has already been mentioned) It posits a (past) observed win rate - if born out as a (past) edge - translates in a similar (future) expected win rate/edge.
Players improve. This doesn't work because neither the own skill nor the skill of the field is constant. But it also doesn't compeltely invalidate the past win rate and sample size as an indicator. It just means that the age of the hand samples is relevant.

2) Sample size is not equally meaningful between stakes. An average win rate of 10bb/100hands at miocrostakes is not the same as 10bb/100 at the highest stakes. If you go to a coach and want to learn mid stakes poker and they show you that they played a million hands with 10bb/100 win rate at miocrostakes over 500k hands then that's an entirely different resume as if they show you 10bb/100 win rate at the highest stakes over the same amount of hands.
If everyone else was trying to play mathematical poker then: yes. But that is not the case. GTO doesn't mean you win most no matter how others play - just that you will not lose if your opponent plays optimally. Those are two very different things. If you're not prepared to play exploitative when the opportunity presents itself then you're leaving money on the table.

In short: "Best poker play" is not defined by making consistently mathematically best plays but by being most efficient at winning money (again: over a large enough sample size that satisfies your personal level of confidence). There is no prize for style points. Knowing what the mathematically best play is is just a fallback when you're out of ideas how to play exploitative in your current situation at the table. It's the base skill (no one has yet mastered) and only becomes the be all/end all once everyone has mastered it fully.

Is survivorship bias a thing in coaching? Certainly. You can use the same Chebychev inequality to calculate how many percent of coaches are just deluding themselves (unwittingly). Problem is that it's again a probabilistic game which of all the coaches these are (but that, again, is skewed by their win rate and sample size. So you can make a good - though not perfect - choice when selecting a coach)