Quote:
Originally Posted by PokerTracker
25% may in fact just be an inaccurate blip on radar\if there is not a big enough sample between the two opportunities, in general we believe these two stats will correlate better than you suspect - but alas as shown below we could not come up with a sample size big enough to prove us right or wrong. PokerTracker 4 includes scatter graphs which are very helpful to see how large of a sampel size is needed to get any effective data, and also allows you to correlate the difference between two stats.
In the example below we added a filter so we are only looking at hands from the BB. The X axis shows Fold Blind to Late Position Steal (our existing stat) and the Y axis shows BB Fold to Button Open (this is a steal as well, but we use the term Open due to PT4 nomenclature). The sample size exceeds 500,000 hands, with a minimum of 100 hands played and 500 hands played respectively.
With a minimum of 500 hands played we only have 9 data points to show and there is some variance between the two stats, however they correlate very closely as I predicted in my earlier post (there is little difference between 85% and 92% over a sample size of 500 hands).
With a minimum of 100 hands we can see how the datapoints greatly increase, but the variance also increases due to sample size.
The only thing we can truly take away from this experiment using the Scatter Graph built into PT4 is that the sample size required to accurately determine the difference between folding to a LP Steal and Folding to a Button steal may need to be as big as 10 million hands with 2500 or greater minimum hands. Of course you can also use the scatter graph to debunk many theories about stat ranges that are commonly claimed without a proper sample size to determine if the claim is correct, that could be a major benefit next time your arguing against a strategic line in the 2+2 strategy forums ;-)
Please post your own graphs once the new stat has been included, we would love to see how your own results correlate!
- TT
I'm glad to see the different stats, although of course it's easy enough to create homemade stats for such things (as I've personally done). I just wanna mention some considerations about the discussion of diference between the stats.
I liked the fact that you took the time to give reasoning behind the initial discussion, and to suggest that people might be looking at too small a sample size or skewing the results in some other way. Just gonna put forward the other side's arguments though. Rather than look at stats, there are logical reasons for why these stats can be useful to at least some degree.
The stats given: are you looking at two distinct stats here, or is there overlap? It should be "BB fold vs CO" compared to "BB fold vs BTN", not "fold vs LP" vs "BB fold vs BTN".
Having said that, more difference might be observed from the SB. From the BB, you can often call with a reasonable hand that you can't profitably 3bet - neither of which is a fold - whereas from SB people tend to call a lot less, so hand that aren't 3bet are more likely to become a fold, and adjust this stat. So just looking at the BB is flawed, I think.
Also bear in mind that it does depend on the games analysed. Difference in stakes, PLO, NLHE, CAP, etc, all change the dynamics. In CAP, for example, some players tend to 3bet or fold from the BB in addition to the SB, and so the fold %ages are a more direct reflection of the opener's RFI% in those games (and therefore more affected by BTN or CO differences).
Any set or subset of games where people have a larger difference between BTN and CO open should show more different results - in some games, the difference will be minimal, but in others it could be a lot more important. So looking at large, general samples may not show any useful information.
Anyway, I do agree that having the different stats is useful, so good work on that... sorry if this post is going off-topic somewhat but I felt it was worth saying. :-)