Quote:
Originally Posted by eldodo42
The math shows that this would be more than meaningless, assuming there are at least 5,000 regs. It would be meaningful if, for example, there were 1,000 regs or less. Here is the math:
Let's take a player whose "true winrate" is 0bb/100, that is breakeven. What are the chances for him being a 4 ptbb/100 winner over 160k hands?
Assume a usual standard deviation, for PLO, of 80ptbb/100. This translates into a standard deviation of 3200ptbb over 160k hands. (To get this, we compute 80*sqrt(160000/100).)
A winrate of 4ptbb/100 translates into 6400ptbb over 160k hands. So two standard deviations. The chance of a random variable to be at least two standard deviations above its mean is around 2.2%. (This assumes the random variable is normal, but you'll get similar results for other well-behaved random variables, and besides, the winrate over a large number of hands should be approximately random due to the law of large numbers.)
Therefore, if we know that the number of regs is at least 5,000, then it wouldn't at all be surprising if the luckiest 100 of them are winning at least 4ptbb/100 over 160k hands. If the number of regs is, say, 500, then the existence of a hundred players winning at 4ptbb/100 over 160k hands would indeed be surprising.
If you give me relevant numbers to check I can do similar analysis for them.
Edit: by the way, it appears to me like there are few people involved in the discussions here with solid professional background in statistics (I mean Masters-level or higher). No offense meant: obviously this is something one has to learn. I think such background is really important for the purpose of discussions with pokerstars, as well as to understand for ourselves what the data actually means. I think the topic needs a dedicated statistician. But if no one steps up, I have some relevant background (Ph.D. in math) and would love to lend a hand. I'll be available in this thread and in PMs, and when the player reps get selected for the pokerstars meeting, I'll contact them directly to ask if they want some one-on-one discussions about statistics of poker vis-a-vis rake. My offer is standing to just about anyone interested.
Thanks alot.
Very intrested overall in this, partly casue just very recently thought move into PLO and also cause like numbers i guess.
Math and logic might be faulty below, but anyways will alteast try contribute with something:
If we take the biggest winner in the sample:
VillainA: 355K hands at approx 4ptBB
We get 80*sqrt (355K/100) = 4766ptBB/100
Since the bit of sample we seen so far feels like it raises a question if anyone is really able to beat rake in this game,
so therefore put VillainA winrate at 0ptBB/100
He has made 3,7*(355K/100) = 14200ptBB
14200ptBB / 4766ptBB = 3 SD from Mean
To run 3SD from mean is 0.23% or 2.3 in 1000
Is all this correct? Although its hard to draw much conclution from it since ppl quit who dont run good, most ppl dont get this sample appearently (i thought we would see alot more grinder in this sample tbh but i dont play Stars but was bit surprised).
Anyways, if we say the best regs can just breakeven in PLO100, we wont have a big number of the best regs and that we have one at 355K hands that hit a 0.23% variance likelhood would talk in favour of this winrate being set to low, and most likely best regs have alteast a postive ptbb/100 ?
**Although bumhunting or very hard table selection might skew this. Im not sure , as i guess this was regular game, if we would choose what sample we wanted to analyse how beatable the games are wouldnt a possible sample of Zoom/Rush poker be the ultimate
Last edited by plunn; 03-30-2013 at 04:36 AM.