Delayed Pooh Bah / Carpal Tunnel post. Random empirical observations.

[Oink]

Ok, so I have been pretty active in this forum the last year or so. You guys are the main reason I have become fairly succesfull at this LHE.

So thanks!


Anyways. Vmacosta and rzk asked me and a bunch of other regs for some stats which made me look at my winrates for all the marginal hands outside the blinds.

So I have spent the whole day doing this and I thought why the hell not post it and call it a pooh bah post or whatever.

I hope you can use this.


What I did was to filter out some marginal starting hands in each of the positions: UTG, HJ and CO. I left out Button - partly because I am lazy and partly because I dont think that rigid preflop guidelines are very useful OTB. When I choose to play a hand UTG its basically from my preflop chart. When I play a marginal hand OTB it depends soooo much on the blinds. So if I could show that 32o was profitable OTB it would simply be because I only play it when the blinds fold way too much. But this doesnt mean that 32o is +EV OTB regardless.

The hands I have chosen in each position I consider at the bottom om my range and slightly outside. Based on the last year or so of debating I would guess that most of you guys consider these hands to be anywhere from marginal to very marginal.

A few words on the samples.

- The db is my entire LHE career. 600k hands dating back from last summer. Limits are 0.25/0.5 through 50/100. The vast majority of the hands are played at 1/2 thorugh 5/10. 1/2 and 5/10 are the "biggest" limits.


The precise filter I used was:
Exactly 1, 2 or 3 of the button
Not a blind
Raised first in



Ok, so lets start by each position

What #Hands , BB/hand and BB mean should be pretty clear.

The Z value is the uncertainty as MrWookie describes here*

The probability that the play was +EV is then found by typing in the Z value here . So this number is the probability that raising the hands was a +EV play and not just variance.






These results all indicate that at least I can start opening up a little more. I am really surprised that the hands did so well.

However it is important to mention that there is selection bias issues. A lot of these hands I only play when I think the conditions are right. So the fact that they are winning could just be because I do a great job estimating when its +EV.

Another issue is that these results bundle up hands that dont play like each other at all. Like ATo, 98s and 44.


So to find out whether different types of hands did well or poor I bundled the same kind of hands in various positions.

These are the results










A few things come to mind

- The good results in the Axo hands is mostly because A4o does well in the CO. Judging from this I am not sure I need to open op a lot more with these hands. Altho its prolly not -EV.

- Axs looks to do better in general. I will definetely be popping these more often in the HJ. A7s used to be at the bottom of my range UTG. I think I can start looking for more spots to raise A6s and A5s.

- KJo UTG and KTo HJ looks to be slightly +EV. The good results with Kxo is mostly because K9o in the CO is doing really well. So one should be carefull concluding that KJo is +EV based on that. Altho it most likely is.

- There is definetely sample size issues with Kxs. Again the good results is from 1 single hand doing well.

- The Qxo and Jxo hands did surprisingly well. I have never been sure about QJo HJ and JTo in the CO. But this has certainly changed know. Maybe these hands does well because they are less dominated than say A9o when 3-bet?

- With Qxs and Jxs there is the same issue as with some of the other results. J9s accounts for more than half of the BB's won but only a little more than 10% of all observations. I tried removing it from the sample and this of course changed results quite a bit.

- As before, if we remove J9s from the "Medium sc" sample the results change. These hands would barely be +EV had it not been for J9s.

- Interesting results with pp's. 33 and 22 does really well in the CO but outside the CO it looks close to neutral EV.


The overall results from all this rambling is that I am gonna open up a bit more. These were all hands that I fold quite often. The fact that they look to be +EV in general could be either because a) I do a good job identifying the spots where I can profitably open them or b) they do better than what at least I expected.



*) To calculate it one needs a standard deviation for the observations. Since I couldnt retrieve that I had to estimate it/assume a value. I have assumed its 30. And just to assure you that I didnt took that number out of my ass I exported the hands that showed up under the "UTG" filter and imported them into a new database. Actually the observed standard deviation was 28 or so. But to be on the safe side I just assumed it was 30 overall. This is of course problematic from a scientific point of view. But on the other hand; i am not publishing this in some fancy scientific journal.

BTW. If you disagree with me on the standard deviation being 30, its pretty easy to do the calculations using another standard deviation