PDA

View Full Version : Help with % of winners in PT

rigoletto
09-20-2005, 04:54 PM
Every Pokertracker database seems to have the number of winning players at 35-45%. This can obviously not be true, but can someone explain why?

edtost
09-20-2005, 05:18 PM
various confounding variables:

-winning players more likely to have multiple accounts (counted multiple times in PT's calculation)
-winning players more likely to play more tables/hours (so more likely that you've played with them before)

etc etc etc.

timex
09-20-2005, 07:05 PM
Because its over such a small hand sample. Ie. if you play headsup against players, and only play one hand against each player, the average win % will be right near 50%, but that doesnt mean 50% are winning players. If you play 2 hands against each player, percentages will will get closer to average. Basicly, in poker tracker lots of the losing players have such huge standard deviations, that 20 or 100 or 500 hands doesn't determine whether or not they are winning players.

09-21-2005, 02:57 AM
Yes, but they wouldn't skew to winners disproportionately.

I think every DB has 35-45% of players as winners because 35-45% of players are winners.

barongreenback
09-21-2005, 04:55 AM
There was some discussion here. Link (http://archiveserver.twoplustwo.com/showflat.php?Cat=&amp;Number=1390393&amp;page=&amp;view=&amp;sb=5&amp; o=&amp;fpart=all&amp;vc=1)
I'd like to see someone apply some mathematical rigour to the topic. I changed my view half way through the thread but at the time I was happy with the explaination I gave in my last post. I'd welcome further discussion.

James

rigoletto
09-21-2005, 07:09 AM
Thanks. I think jtr hit it on the nail in that thread:

[ QUOTE ]
Let's be radical and assume that there are no winners. We model everyone as a -2 BB/100 loser, with a 12 BB/100 s.d. Then assume you've got a big poker tracker database with 10,000 of these people in it. On each of them you have 1000 hands (10 x 100 hand sessions for the purposes of calculating their winnings). They'll all expect to lose at -2 BB/100 but the standard deviation will induce some variation.

If you looked at the winners / losers figure even for this database, you would see 31% winners.

[/ QUOTE ]

09-21-2005, 10:12 AM
Imagine a table of 10 people each playing slightly better than the one on their right (except the last guy, obviously)

Over time the 5 best players win and the 5 worst players lose. But because of the rake the 5th best guy, even though he wins from 5 players and loses to 4, is still struggling to break even.

Moozh
09-23-2005, 03:21 AM
[ QUOTE ]
Yes, but they wouldn't skew to winners disproportionately.

I think every DB has 35-45% of players as winners because 35-45% of players are winners.

[/ QUOTE ]

That's because everyone has a small sample size of players.

Say that there are 10% winners... even 30% or so of the rest will win over a small sample of 1000 hands (even if we know they've already lost much more).

Everyone has small snapshots of everyone else. The only way for it to converge on reality is for someone to have the full picture (every hand everyone has played).