I just run poker variance calculator for -1 bb/100 winrate and write the results in %Probility tab.

Than look at the GGPoker data for reg player who have 200-300k hand sample. There are total 67 players like this in that data. %Pool is the % of the players from that 67 players.

You can see that how close the numbers pool vs poker variance calculator. I wish we have more data but we don't.

Of course this not show anything but its very open to question like that

What if the difference between regulars are so small and who have better results vs the population actually the one that run hot?

250000 are indeed not that many hands, results over that sample will be very contingent on luck

Yes its true. But most of the coach in our community share results even lower than 300k sample.

but usually coaches dont show a -1bb or +1bb winrate over those samples
if they a true 300k sample (their last 300k hands for instance and not a warped or specifically picked sample) with a winrate of 10bb you can trust they are a decent winner, it still could be 6bb actually, but it shows they beat their games

Not sure if I got exactly what you're asking here. Yes, I think above 10nl or so the rake gets so high on GGPoker that, coupled with people being more or less disincentivized to bumhunt) the average reg is losing pre rb, probably that -1 bb/100 you mentioned if that was what you got over a larger sample.

Any difference in a reg ability compared to his peers, big or small, should be reflected in his true winrate.

Last, for any given standard deviation, you will always have a sample big enough that the winrate will trend closer to its true value. Being big or being small or being negative doesn't change that.

Yeah I think its a disgrace. Anything who tries to promote coaching and shows a curated graph of 100-200k hands where things went well should be a red flag.

especially given how long online poker has been around. If you've been crushing for years, why would you only show a tiny sample? Pros who have been playing long enough to become coaches have millions of hands, some more like 10 million by now.

You can always ask for a bigger sample to prove they are winners over very large samples, or just refuse to do business with them if they don't provide that to you. If there is no single coach able to show you a big sample of crushing, you can just say no to everyone and not be coached at all.

Poker community doesn’t have a clube about variance.

Biggest winner makes 25 buy ins at highest stakes over 20k hands and people think he is goat because he won 2millions. It’s pathetic lol

Just take a look at comments…

exactly, everyone has a 20k hand sample where won 2 million, nothing special there
and then there are people who think those players are noteworthy!! what a bunch of ****ing idiots

Any of those top 15 guys who consistently play highstakes can make a run of 25 buy ins over 20k hands.

It literally happens all time.

See NBA. People were talking Jokic could have a better career than lebron… but now people talk Jokic sucks because he never beat a 0.500 team in playoffs.

Here at 2p2 it’s exactly same situation.

The question of whether 250,000 hands is a large enough sample size for reasonably accurate statistics can't be answered. That's because no information is given about the size of the standard deviation. Coaches, or whoever, who are posting graphs of how well they have done, should also post the appropriate estimate for their standard deviation for their play.

The best formula that I know of for estimating the standard deviation, known as "the maximum likelihood estimator for the standard deviation" can be found in the chapter titled "Computing Your Standard Deviation," starting on page 75, in my book Gambling Theory and Other Topics - Expanded Edition.

When looking at a sample of 250,000 hands, I would want it to be the last 250,000 hands that this person played. The reason for this is that over time games can change or something that was once quite effective may not be as successful as it once was.

In addition, you may also want to look at the chapter "Win Rate Accuracy" starting on page 84 in my above mentioned book.

Mason

What do you guys think then of highstakes coaches providing graph for those few 100k hands, but also key statistics from his game, maybe a few HHs here and there. Assuming his potential clients would be midstakes regs, they should be capable of judging his coach by these parameters right?

Take this highstakes player whose stats are public (won't mention his sn), filtered for 6-max (NLH):

778K hands, 3.7 bb/100, main limit NL2K, plays up to NL40K.

VPIP 33, PFR 24, 3bet 12.4, F3B 51, 4Bet Range 4.9, WTSD 30, W$SD 49, WWSF 51
Flop Cbet 61, CB IP 64, OOP 51, FvsCB 40, IP FvsCB 32, OOP FvsCB 42, Overall Flop Bet 42, FlopXR 17, Flop Agg% 42
Turn Cbet 54, CB IP 57, OOP 46, FvsCB 48, IP FvsCB 40, OOP FvsCB 51, Overall Turn Bet 40, Turn XR 11, Turn Agg% 37
River Cbet 56, CB IP 56, OOP 54, FvsCB 52, IP FvsCB 48, OOP FvsCB 54, Overall River Bet 44, River XR 13, River Agg% 39

Would you take him as a coach if given the chance?

Please don't leak his name but these are my thoughts.

I would say No pretty quickly. What happens if I want to play 2knl? He probably can't extrapolate all his knowledge to me so I am looking at less than 3.7bb/100 winrate on average. That should be a floor not a ceiling.

My view from the students perspective is your coach should be at least twice as good as you wrt to winrate and more likely 3x as good as you. If I'm coaching a 50nl player and his winrate is 4bb/100, I better at least have a double digit winrate at 50nl to coach him. Otherwise it's not worth it for the student.

For myself, if I am looking to get coaching. I'm only hiring a world class player because it doesn't make sense for me to hire someone that is only slightly better than me. I'd rather just pay the $1k/hr and hire someone like Jayser since you know you will learn from someone that good.

I dunno, to me feels like 3.7bb from nl2k to nl40k online is pretty worldclass, maybe he battles a fair bit

i wont "out" him either but cmon we all know who it is and he doesnt even coach
btw he could also add some HU sample 98k hands 7,4wr with a 655k profit

his results are clearly impressive, that doesnt make him a good coach though. But the framework is there, at least its somebody who beats the game well and has a public graph/stats.

I would pay a lot to have a few conversations with this person and anyone trying to get to his situation who wouldn't isn't thinking straight.

I've learned from better players than me even by hearing them say things I disagree with/think are wrong

You can check your standart deviation from Pokertracker.

The main reason I opened this topic was to point out how close people's winrates are to each other in high rake situations. Someone like me, who has spent a long time examining databases with high hand count, knows how different the situation is for Pokerstars and GGPoker. In pokerstars regs can hit for 8-10bb winrates in a very big sample. But i never see any reg in GGPoker have 5bb winrate in a very long run.

In 1mil hands -2/+2 winrate can happen. In 2 mil -1,5/+1,5 can happen and that kinda differences are very big when you have 5- bb winrate. But those difference is not that high in Pokerstars because good players can have better winrate.

If a player post 500k sample 1bb/100 graph from GGPoker our community see him as normal player.

If a player post 500k sample 4bb/100 graph from GGPoker our community see him as a crusher.

Two quick points.

1. How is Pokertracker computing the standard deviation? The reason i responded in this thread is that it seemed as if the claim was being made that 250,000 hands was not enough. This is easily answered by knowing what the standard deviation is for the particular statistic of interest.

2. In limit games, the blinds are much larger. So when comparing roughly equivalent games of no-limit to limit that can produce approximately the same win rate in total amount won for expert players, you would expect there to be a big difference in the number of big blinds won.

Mason

They must use standart standart deviation formula for hand. Usually players stay between 80-100 in NL Games depends on how much they use run it more than once future.

Do they have on their site exactly how they're calculating the standard deviation? I haven't looked at this sort of thing in many years, and I assume that the Pokertracker people are probably doing it correctly, but I have seen examples in the past where this was not done correctly.

And when you say 80-100 in NL Games, I assume you're saying 80-100 big blinds per 100 hands.
Mason

I also check for how they calculate it but cant find it. But i think its very unlikely that they are making mistake. Yeah you are right its for bb/100.

Just to point out something. If the sd is for big blinds per 100 hands, then a sample of 250,000 hands is actually a sample of 2,500 groups of 100 hands. And the square root of 2,500 is 50. So, you can divide the 80-100 sd estimate for 100 hands by 50 to see what happens after 250,000 hands.

So, let's say that your win rate is 6bbs per 100 hands. So a 95 percent confidence interval (2 standard deviations either way) and going with the upper end of the standard deviation is approximately 2bbs to 10 bbs.

Mason

Yes you are right.

Yea and in highest stakes poker no one posts that number of hands. Its pretty obvious that variance is going to play a bigger role than actual skill edge in the highest stakes games that only run for a few thousand hands at most before they break because the fish leaves again.

Im fairly sure its common conception that someone like Barak Wisbrod is running super hot rather than being an actual top 3 player skill wise for instance.