Using Bayesian inference, you can get far more accurate winrate estimates over small samples.

Here, I've plugged in the NL200 rate distribution as a prior, then updated it with DDPs results over the 64k sample.

Assuming DDP is a random player plucked from the population, his true winrate is probably about 6 bb/100.
There's a 75% chance that his winrate is greater than 3.7 bb/100.



If we instead plug in 200NL reg data as a prior, his true winrate would be closer to 7.3 bb/100.
There's a 75% chance that his winrate is at least 5.1 bb/100.

Excuse my ignorance, but isn't this the same this as what the primedope calculator does?

I assume it's basically the same reason here?

I have a hard time understanding why anyone would think the verification is about poker competence. What other free platforms where people offer their services do that? Does Angi's List send out people to check if someone really is a good carpenter or mason? Ratings are always based on customer feedback.

Takelessons.com (nowadays owned by Microsoft) does background checks for new teachers. That includes crime databases and if someone is in the national sex offender registry but not if they're good at teaching piano or saxophone. Teachers have ratings because their students reviewed them. Not the platform.

@tombos21
his sample is anything but random

@madlex
Its fine its just identity verification. I would call that verified user and put some criteria on verified coach think (maybe big winning sample or wait for few good reviews before), maybe its obv to others that verified coach means you just give ID not that you know anything it was confusing for me at first.

Nah, primedope doesn't make any prior assumptions about what the winrate "ought to be". It's a frequentist approach. So it would just take the winrate at face value.
Bayesian approach uses population data as a prior, so you tend to get more accurate results.


That's why I included the second section where we pull using reg data as a prior, rather than assuming a random sample.

But I guess your point is that he cherry-picked the sample?

Yes he wont post if he was winning 2bb, so there is selection bias.

Your analysis works if someone commits to play x number of hands and do the analysis regardless of the result. Most posted graphs are not random, there is a reason why someone is posting which makes it non random.

Also i think DDP SD is higher than 100

Thanks for the analysis Tom!

Yeah everyone thinks I cherry picked my sample so now I'm doing a BR challenge .

That's probably true lmao

‘Coach’ is too broad a category to require graphs though, there’s ppl offering coaching for $25/hr.
Anyone can peruse the coaching page and see the huge range of price points (from $25 to $1000/hr) which should make it pretty clear that verification is a forum requirement, not a value indicator.

Money goes one way in coaching. So presenting a winning graph is good for getting clients, they aren't going to get money from the coach when the graph turns. Keeping interest up as long as possible is the key.

GL

Why did you choose BR challenge over putting in more hands at 200nl?

4 main reasons.

1. I don't 100% trust my money on Ignition (maybe 95%?) and I don't want to put too much of my net worth on there.

2. I don't know how far the bots have infiltrated the poker economy and I still want to study their play more (for cheap).

3. I want to see what winrates are possible at lower stakes using heavy MDA strategies as an example for future students and to boost coaching resume.

4. Crypto

Basically, all this talk of variance and standard deviation is a derail. The little bit of Bayesian vs Frequentist back and forth is amusing (Pro tip: In a Bayesian vs Frequentist challenge, always bet on the Bayesian), but it too omits the main point.

The OP is true, and at the same time it is ridiculous. If you are playing poker at a table full of people with skill equal to yours, you are doing it wrong.(*)

If the sixth best poker player in the world sits down with the five better players, things will not go very well for them; but if some kid who has read, say, Exploitative Play in Live Poker goes straight from their job flipping burgers to the local cardroom to play 1-2 NLHE, they stand a pretty good shot of being able to beat the game long-term.

Poker is fundamentally about playing against weaker players and exploiting those players' weakness to take the money. This is why game selection is a fundamental skill of successful poker.

Poker is, at root, a hustle, a way for the smart money to take advantage of the dumb money.

In the artificial situation described in the OP, by construction no one can hustle anyone else. Does variance matter in poker? Of course it does. But successful players can fade the variance, and in the long run they are going to prevail over the fun players.

--------
(*) Yes, a home game with your pals can be well worth playing when no one has an edge.

I think the point is more so that the observed winrate deviates so much from the true winrate. An initially breakeven player may first run at -10bb or vice versa, which influences the trajectory of their career vastly

Yes he is missed the whole point of the thread.

Is this post from 2005? Had no idea!