Predicting BB/100 based on other stats

[Jeff W]

This formula predicts my BB/100 to be 2.57.

I am very skeptical that my true winrate is actually that high.

[astroglide]

punched wrong first time but it's still WAY off for me on a my current 212k hand db after several re-checks

[Senor Choppy]

One of the best posts I've seen on 2+2 in a while, Nate.

After 275k hands, my predicted is .49 below my actual, FWIW.

[cwl]

i tried the formula on 3 sets of hands i have.

paradise 20-40 - was a bit off but not worlds away, was about half my win rate.

paradise 40-80 - was way off, about 3 times my actual win rate.

higher limit games ranging from 50-100 through 300-600, mostly 100-200 through 200-400 - was way off, about 4.5 times my actual win rate.

whether any of this is interesting or not i dont know. i imagine of my 3 sets of data the 20-40 is probably the most like the party 15-30 and thats the one that i got the closest result from the formula. both the 40-80 and the higher limit games are tighter than the 20-40 and i imagine the party 15-30 and both gave me a way inflated win rate. if im understanding things right it seems like this makes some sense. i filtered for at least 7 players for the stats i used but for both the 40-80 and higher limit games a higher percentage of the hands have specifically 7-8 players while more of the 20-40 are actually full games. this might have an impact as well.

fwiw here are the saw flop %, and avg players from the summary tab for each game.
20-40 30.95/8.15
40-80 28.86/7.04
high limit 29.33/8.13

it seems reasonable to me that equivalent stats in a game with less people seeing the flop and smaller pots would correspond to a smaller win rate than your formula predicts.

[astroglide]

[ QUOTE ]
After 275k hands, my predicted is .49 below my actual, FWIW

[/ QUOTE ]

wasn't it .49 above for yours? predicted was > actual.

[Senor Choppy]

Oops...you're right.

[CourtesyFlush]

Should I assume that this can't be used for loose passive games where your W$SD and W$WSF will be lower, but on average the pots you win will be much larger? According to this calculation my winrate is -.88 for .5/1 and 1.08 for 1/2 and this is very low compared to my results so far. My VPIP is a rather rocky 14% so this could be playing a factor as well.

[AlexM]

Wayyyyyyy off for me:

15/30
TH - 112,198
VPIP - 16.89
W$WSF - 37.17
W$SD - 50.64

Predicted BB/100 - 0.17
Actual BB/100 - 2.01

One the obvious problems with the formula is that it doesn't accurately measure the fact that calling down too much (something I do) is WAY better in *limit* (not in NL) than folding too much. It treats them equally, and that's definitely very flawed. I suspect a large part of this is that the data you received includes far more players that fold too much but are otherwise decent than it does those who call too much but are otherwise decent. Considering the number of weak/tight players out there, this isn't very surprising.

[JasonP530]

The point is to prove that tight agressive play is a good thing, by finding that the regression supports that. It should be noted that Nate did not use a particularly large sample size both in hands and in people, so the results may be a little off. Also, the regression formula he gave is a line, and each person is a point around that line, so if you are far away from the line, that doesnt mean that it is wrong per se, since there are outliers. I am surprised that agression was not a significant factor in the equation, as more agression(up to a point), is usually beneficial. I would think that perhaps agression on the river would be correlated with winning at showdown as people are more likely to be winning when they are betting. Also, many of the times you lose at showdown(if youre a TAG), occur when the river goes check/check and your AK/AQ is no good. I would think this would somehow be included.

[AceHigh]

[ QUOTE ]
-VPIP
-W$WSD (Won $ when seeing showdown)
-W$WSF (Won $ when seeing flop)


[/ QUOTE ]

Hmmm...it occurs to me that one thing that helps all these numbers is defending your Big Blind in 1-on-1 situations.