#21
|
|||
|
|||
Re: Predicting BB/100 based on other stats
I just want to add a couple of things:
Being a game of incomplete information, modeling a winrate is darn near statistically impossible without an insane amount of samples. There is also a huge amount of multicollinearity inherent in this model. While your BB/100 predictions would still be valid, you can not pull out the effect of individual regression components. Third, as everyone has said, you would need to either run a regression for each case (limit, # of players) or include a bunch of simultaneous equations. We always throw around "sample size insufficient" on these boards, but this is something that needs stressed. To draw any meaningful conclusions, you would need a HUGE amount of samples. |
#22
|
|||
|
|||
Re: Predicting BB/100 based on other stats
I'm sorry Nate, I'm being a bit dense.
I don't really get what you've shown, other than people that win a lot have high win rates [img]/images/graemlins/confused.gif[/img] (I'm being facetious on purpose, but that's basically what I don't get) |
#23
|
|||
|
|||
Re: Predicting BB/100 based on other stats
[ QUOTE ]
I'm at almost exactly 40% for full-table hands since I started tracking at the beginning of this year. [/ QUOTE ] Would you think that would be indicative of running well, playing well or a combination thereof? Obviously I would think combination, but is it weighted by one or the other moreso? Seems like many of the stat posts that I've seen that 30-35 seems about avg. Or is that heavily dependant on too many variables to be able to pin down? Or am I looking at it the wrong way? (most likely) |
#24
|
|||
|
|||
Re: Predicting BB/100 based on other stats
[ QUOTE ]
I'm sorry Nate, I'm being a bit dense. I don't really get what you've shown, other than people that win a lot have high win rates [img]/images/graemlins/confused.gif[/img] (I'm being facetious on purpose, but that's basically what I don't get) [/ QUOTE ] Yes, it is correlational. I think this is just a predictive model and isn't really supposed to prove anything. |
#25
|
|||
|
|||
Re: Predicting BB/100 based on other stats
[ QUOTE ]
[ QUOTE ] I'm sorry Nate, I'm being a bit dense. I don't really get what you've shown, other than people that win a lot have high win rates [img]/images/graemlins/confused.gif[/img] (I'm being facetious on purpose, but that's basically what I don't get) [/ QUOTE ] Yes, it is correlational. I think this is just a predictive model and isn't really supposed to prove anything. [/ QUOTE ] I think its biggest use (if its accuracy is nailed down more), will be to let people know with more certainty whether they are running well, bad, really well, terribly, or just ok etc. etc. -SmileyEH |
#26
|
|||
|
|||
Re: Predicting BB/100 based on other stats
[ QUOTE ]
[ QUOTE ] I'm at almost exactly 40% for full-table hands since I started tracking at the beginning of this year. [/ QUOTE ] Would you think that would be indicative of running well, playing well or a combination thereof? Obviously I would think combination, but is it weighted by one or the other moreso? Seems like many of the stat posts that I've seen that 30-35 seems about avg. Or is that heavily dependant on too many variables to be able to pin down? Or am I looking at it the wrong way? (most likely) [/ QUOTE ] w$sf is highly dependent on average number of people that see a flop when you see a flop. 40% is not atypical at 15/30 for a TAG. |
#27
|
|||
|
|||
Re: Predicting BB/100 based on other stats
Outstanding post. I think there a lot of ways we can tweak this model to make it much more accurate. I'm a statistics major with a lot of regression and data analysis experience. If you want, I'd be interested in helping you with this. I sent you a PM.
I tried to determine my optimal VPIP given my W$WSF and W$SD using your model (simple max/min calculus problem). It came out to be 36.91% with a maximum BB/100 value of 7.86. These numbers don't really mean much as the model is still in works but I think it means the coefficients for VPIP are off. How did you go about modeling VPIP as a quadratic? |
#28
|
|||
|
|||
Re: Predicting BB/100 based on other stats
But why doesn't your winrate alone tell you that?
|
#29
|
|||
|
|||
wtf !@#%
total hands : 72301 5/10 party (april/04- sept/04)
vip: 19.11 wsf: 35.74 W$SD: 49.68 (ouch! this is embarrassing) formula says im -0.33/100 according to ptrack, this is off by over 4.00/100bb does this mean that im a fish (with good table selection) that is running well? good thing i havent played for like 5 months. although im thinking about playing 15/30. |
#30
|
|||
|
|||
Re: wtf !@#%
I find it hard to believe you're winning that much in that game with those numbers. I'm not doubting you. I'm just a little taken aback. I can't imagine anyone could sustain over 3 bb/100 in the hell hole that is 5/10
|
|
|