poker correlations

[poker1O1]

I play full ring MSNL. I created an excel sheet w/ all players over 2500 hands, and found the correlation between bb/100 and every stat that PT could export. Here are my findings:

highest correlations
won_wsf : .454
turn agg
turn bet%
fbet_won_wsd
flop raise %
turn raise %
wf_no_fold
steal att : .362


lowest correlations
turn check % : -.380
p_check
flop check %
flop fold %
turn fold %
wf_pf
pfc_fold_no_sd : -.237


conclusion: based on these findings, it appears that aggression, in one form or another, is the biggest indicator of a successful player.

[jukofyork]

[ QUOTE ]
I play full ring MSNL. I created an excel sheet w/ all players over 2500 hands, and found the correlation between bb/100 and every stat that PT could export. Here are my findings:

highest correlations
won_wsf : .454
turn agg
turn bet%
fbet_won_wsd
flop raise %
turn raise %
wf_no_fold
steal att : .362


lowest correlations
turn check % : -.380
p_check
flop check %
flop fold %
turn fold %
wf_pf
pfc_fold_no_sd : -.237


conclusion: based on these findings, it appears that aggression, in one form or another, is the biggest indicator of a successful player.

[/ QUOTE ]
A couple of years ago I attempted to make a set of PT rules for 6max limit and found it very interesting just how much passive (post-flop) players lost:



It appears that the extreme LAG players didn't lose all that much and most of what they did lose was to the rake.

The most juicy donks of all appeared to be the players who raise alot pre-flop then are passive post-flop; loosing at a phenomenal 16BB/100! (while only paying 7BB/100 in rake).

From that point onwards I always tried to get at least one of these LAP "super donks" on my table if possible (I think other good players would tend to hunt down the not so profitable LPP/LAA players (wrongly) assuming they were the biggest fish...).

Juk [img]/images/graemlins/smile.gif[/img]

PS: You might alot find this thread of interest which also found that post-flop passiveness was the biggest indicator of a losing player (the same guy went on to write the "SixthSense" table selection software [with it's NL "fish score" equations])

[Abbaddabba]

This kind of analysis can go wrong really easily.

People get aggressive when they have strong hands.
So it shouldnt be a surprise that the people who are up money tend to be aggressive.

If the sampling is done properly it shouldnt influence it, but it's very difficult to get the kind of data you would need.

[jukofyork]

[ QUOTE ]
This kind of analysis can go wrong really easily.

People get aggressive when they have strong hands.
So it shouldnt be a surprise that the people who are up money tend to be aggressive.

If the sampling is done properly it shouldnt influence it, but it's very difficult to get the kind of data you would need.

[/ QUOTE ]
Yep, very true and at the time I used any player with > 100 samples (see here from my method). This does introduce a significant systematic bias because of players running hot or cold over their sample. Even so, my plan was not to generate statistically correct results; it was simply to try to create a better set of 6max auto-rate rules than I could make from an educated/arbitrary guess. I went on to play another 500k hands of limit 6max after creating these groupings and I'm pretty sure it helped my table selection (and thus winrate) significantly.

Juk [img]/images/graemlins/smile.gif[/img]

[Abbaddabba]

The rule of thumb is: dont sit at tables with people who run good.

[img]/images/graemlins/wink.gif[/img]


I think that if you bumped up the min hand criteria by a factor of 10, it would sort out a lot of the problems with your sampling. As long as one of the figuress you're looking at converges, it'd offer some insight.

[jukofyork]

[ QUOTE ]
The rule of thumb is: dont sit at tables with people who run good.

[img]/images/graemlins/wink.gif[/img]


I think that if you bumped up the min hand criteria by a factor of 10, it would sort out a lot of the problems with your sampling. As long as one of the figuress you're looking at converges, it'd offer some insight.

[/ QUOTE ]
I did think about this at the time, but it turns out that it also introduces another systematic bias caused by the fact that alot of the fish tend to bust out well before they've played 1k hands and the ones who do last beyond are usually running on the hot side of variance anyway (this a similar bias to what can be observed in personal PT databases where the winner:loser ratio is well off at something like 40:60, etc).

Juk [img]/images/graemlins/smile.gif[/img]

[Doug Funnie II]

How can you export PT stats into excel?

[jukofyork]

[ QUOTE ]
How can you export PT stats into excel?

[/ QUOTE ]
I think you would need to make a custom SQL query to do it using PT - I used another app called "PokerManager" to do it, but sadly it is no longer supported by it's author.

Juk [img]/images/graemlins/smile.gif[/img]

[Shandrax]

[ QUOTE ]
People get aggressive when they have strong hands.
So it shouldnt be a surprise that the people who are up money tend to be aggressive.

[/ QUOTE ]

At first this sounds like a key observation, but since we are talking about a sample, the hand strength should equal out and in the end we got players left who just play more aggressively than others.

[filsteal]

Ideally, what you would like to do with this is to apply weighting factors to the different opponents based on how many hands you have on them. Obviously an opponent for whom you have 3k hands should count more than an opponent for whom you have 150 hands.

There are also like a zillion problems you can when trying to do this sort of thing. Remember that correlation does not imply causality.