When I read the line of thinking of pokerstars and the trip reports, i frequently encounter “the best plo players are still able to grind nice winrates” or “larger winrates for plo players compared to nlh”. I really doubt this and fear I think there is a huge error in analyzing the data.
I feel like Stars presents a list of the biggest winners on PLO over x hands, to show people still win. Below I show you 10 runs of 100k hands with an expected PTBB/100 van -3BB (bruin straight line) and a standard Dev/100 of 175. I think these settings are very realistic if you play 100bb deep tables (although based on player styles, pokerstars will have a lot of players with a way higher stdev, increasing the likelihood of a large win).
As you can see, even in this 10 runs, you have 2 decent size winners. So at 100 plo you expect to be town $6.000,-, but two of these players are up about $15.000,-. I think the above is a huge UNDERESTIMATION of variance because:
- It assumes a normal distribution, which is very unlikely if you take into account tilt and varying stacksizes (Winning or loosing those 3 >1000bb pots you play in this sample is a huge effect on your winnings)
- It assumes a fixed expected winrate. However your winrate over 100k hands itself shows huge variance. I’ve personally had 100k samples with a winrate of <-5bb and >10bb. Its such a huge effect if you flop good in 3 / 4 bet pots or hit your draws often or not.
Another important issue is that players running good are way more likely to grind out a large sample size, compared to players running bad (they simple bust halfway). So selecting on sample size is a huge selection bias in itself.
If you analyse the amount of data that Pokerstars will have, you will easily find many outliers of players performing very well and I hope you did’t base any Isle of Man discussions on data like this?
What should be done is that you select 2 sample periods. In the first period you determine a group of players, most likely to be very good players. In the second period you check how they do.
For example, an analysis I would really like to see can be constructed like this:
Select the 100 biggest winners on plo 0,25-0,50$ 6-max games during Jan-Jul (makes it likely these are mostly good players). See what the average winrate is of these persons on PLO 0,25-0,50 6-max games during Aug-Nov. I would be very curious to this and if these guys have a positive winrate, I think rake is not too high atm.
Last edited by joeri; 12-01-2012 at 06:35 AM.