stat person help

In the last week, I played roughly 32 hours of poker and netted $2000. This was in a combination of 5/10 holdem and 5/10 omaha hi lo on paradise, along with a 4/8 game with a 6/12 kill in a casino. Im just wondering what are the odds of results like that with an expectation of 1 BB an hour, as well as at 2 BB an hour. That seemed like a ridiculous result to me over such a big sample, so i just wanted to know the liklihood of that. Thx much

One other question, I have a log of playing sessions that includes number of hours and the result...Is there any way to extract and hourly SD from this?

I started to try to prove this mathematically but it is not worth the effort. You cannot tell anything about win rate from 32 hours of play. Period. You had a good run of cards. You may be a winning player, you may not. Your standard deviation for 32 hrs could easily be $1000 given the game types, etc... 32 hours is not in any way a big sample. It is tiny.


1BB/hr is very small compared to the standard deviation which, in my opinion, could be as high as 20BB/hr in these games. The difference between +1 BB/hr and -1 BB/hr cannot even be seen with any level of certainty until you have hundreds of hours of play under your belt. Standard deviation increases proportional to the square root of hrs played, while expected win is linear, so eventually it is possible to get a win rate with some accuracy.


I would also guess that a terrible player is just as likely to hit a run like this as a winning player. Even though the terrible player loses say 5BB/hr, his standard deviation is enormous because he is playing more hands. So even though he will be a huge loser in the long run, he is very likely to have a few big rushes. And thank God for that, because that's what keeps him coming back.


Good run, congrats, nice to hear something positive around here. When you hit your run of losing 200BB in a week, you can make the requisite whiny post, and I won't even be mad.

To answer your question, my educated guess is you'll hit a (positive) run like this about 1-2% of 32hr sessions whether you win 1BB/hr or lose 1BB/hr.


Godspeed,


Glenn

oh i have logs going back much farther than that, and i want to get an hourly SD, sorry to clear up any confusion...(that was part of the reason why I wanted to get the hourly SD from my logs, in order to estimate the liklihood of this event.) ive been beating the 5/10 game for about 1.5 BB an hour, but i have no clue how to get the hourly SD without starting a new log on an hour by hour basis. thank you very much.

The formula's and how to calculate this are in Gambling Theory and Other Topics, but I use StatKing to calculate this stuff. It's VERY tedious to calculate for any serious number of sessions. /images/smile.gif Trust me!


StatKing gives you a lot of other statistics too, and if you're beating the game for that much, the cost of StatKing pays for itself in maybe one session. /images/wink.gif


- Tony


P.S. The book is still good even if you buy StatKing for many reasons.

where do i get statking ?

It becomes more likely the worse you play. A great player will suffer smaller swings in both directions.


What you describe is nothing special. You'll probably post again when you're stuck two dimes in a week.

Average your hourly Results.


res[avg] = total win/hours played


Take the sum of the squares of the differences between each hourly result and the average.


sum ((res[i] - res[avg])^2)


Divide this by the number of hours.


sum/n (actually, for a small data set you divide this by n-1, but I am assuming that n will be large)


This gives you the variance.


Take the square root of this to get SD.


Now everyone write that down. You can get probabilites from this by finding the area under the standard normal curve between two points, or between a point and infinity...but that's another lesson.

Yes, sorry for confusion again...My log has my sessions recorded by the session as opposed to by the hour, i was wondering if there is any way to reasonably extrapolate and HOURLY SD from a PER SESSION log. hehehe, i know im not that greatest stat guy so i know i caused all the confusion /images/wink.gif --Ricky

thank you for the advice. ive played a lot of online poker, have won, and have never had a streak like that good or bad. this week pocket aces seem the norm and they have held up EVERY time. i know this won't go on forever, I was just simply wondering what the odds of a streak like that happening was for a tight agressive player.

If you know the length of your sessions then it is possible to extrapolate this information (it might require some fuzzy math)...however, it is really a pain to do. If you have some kind of math background then look up the equations on the internet and get to work. If not, this is going to take you a long time. It might be a worthwhile excercise since that analysis you will go will give you a better understanding of relevant aspects of probability. If you can program at all, it might be a good idea to write a program that does the work for you (VBA would be ok if it's in Excel, or if it is in a txt file just parse it with C++ or something). Doing this by hand would be very tedious. I will look into writing a program that does this (for and Excel file) tonight because I keep records the same way. If I get it I'll let you know.


-Glenn

p.bb_std = ((sum((p.bb(rng).^2)./p.t_ses(rng))-p.bb_hn^2*p.sum_t)/n2)^.5 ;


p.bb is an array of net big blinds won/lost in a session

p.t_ses "" "" session times

p.bb_hn is your hourly win rate

n2 is the number of sessions.


any questions? (sum is an obvious intrinsic function)


best,


zooey