jdl22
05-24-2004, 11:20 PM
I lost a bunch of my hand histories about a month ago and as a result don't have a very big sample of hands. I wouldn't have had too many anyway having only been playing since January. Anyway, before I get too many hands for this type of thing I decided to use the export feature in PT and establish a system to improve on the "session notes" that are currently included in PT.
I multitable NL/PL100 at Party. I feel that PT is still not setup optimally for multitabling in terms of sessions. The problem is that it considers each table a session. While this has improved somewhat in that you can find your "true win rate" I would like to combine my sessions to get my "true" standard deviation. So I exported the session detail notes to Excel and separated my sessions. I used as a rule that any period of 15 minutes or more where I wasn't playing would constitute a break and hence a separate session (so for example if I was 3 tabling and the three tables ended at 22:37, 22:39: 22:34 and my next tables started at 23:00 and 23:02 the same day those would be broken up into 2 sessions but not if the second set started at 22:45). I also considered any time in between playing that was less than 15 minutes as time played. The reason for this is that it is likely time spent searching tables or on waiting lists and is part of the overall playing time.
To the point of the post. Here is what I did:
After separating the sessions I got this data from each session: minutes played (had to manually calculate), $ won, hands played, big bets (2 big blinds, later I will always use bb to mean big bets) won/hr, bb won/100 (these last two don't get used later I just realised). Using this thread (http://forumserver.twoplustwo.com/showflat.php?Cat=&Board=probability&Number=612037& Forum=f11&Words=standard%20deviation&Searchpage=0& Limit=25&Main=612037&Search=true&where=sub&Name=&d aterange=1&newerval=1&newertype=y&olderval=&oldert ype=&bodyprev=#Post612037) (Thanks BruceZ!) I calculated my SD both for bb/hr and bb/100. I also summed each of the used columns and used them to find the average $/hr and $/100.
As a result of that I got a mean of $43.53/hr or $39.82/100 (10.88 bb/hr 9.96 bb/100) and a SD of $124/hr or $121.84/100 (31 bb/hr 30.46 bb/100).
As I expected my SD was lower than PT's calculations using 1 table as a session. PT gives me SD/hr of $133 (33.47 bb) and sd/100 of $173 (43 bb).
Next I wanted to find confidence intervals. Because I had 38 sessions I assumed the distribution was normal. I got both 95% and 99% CI. I'll go through the steps I used for the /100 99% calculations, I used a similar method for the /hr and 95% calculations. For the CI I took 9.96+-2.576*30.46/sqrt(6563/100). 6563 is the number of hands I've played. This interval is (.27,19.64). The 99% interval for $/hr is (.58,21.19).
OK so here are my questions/issues:
- Have I made any mistakes in the methods mentioned above?
- Can I conclude that I am a winning player just based on this sample? If so, why do people often say that one cannot make this conclusion even after many more hands than this? If not, what more information is necesary to make this conclusion?
- I believe I should use this SD when calculating my ROR and so forth. Is this correct?
- My sessions are very heterogeneous. I have 3 of less than 10 minutes and 3 of greater than 200 minutes. How does that affect this if at all? Can I still use the normal distribution assumption given that this is the case?
Sorry this is so long. I hope that others will find it helpful.
I multitable NL/PL100 at Party. I feel that PT is still not setup optimally for multitabling in terms of sessions. The problem is that it considers each table a session. While this has improved somewhat in that you can find your "true win rate" I would like to combine my sessions to get my "true" standard deviation. So I exported the session detail notes to Excel and separated my sessions. I used as a rule that any period of 15 minutes or more where I wasn't playing would constitute a break and hence a separate session (so for example if I was 3 tabling and the three tables ended at 22:37, 22:39: 22:34 and my next tables started at 23:00 and 23:02 the same day those would be broken up into 2 sessions but not if the second set started at 22:45). I also considered any time in between playing that was less than 15 minutes as time played. The reason for this is that it is likely time spent searching tables or on waiting lists and is part of the overall playing time.
To the point of the post. Here is what I did:
After separating the sessions I got this data from each session: minutes played (had to manually calculate), $ won, hands played, big bets (2 big blinds, later I will always use bb to mean big bets) won/hr, bb won/100 (these last two don't get used later I just realised). Using this thread (http://forumserver.twoplustwo.com/showflat.php?Cat=&Board=probability&Number=612037& Forum=f11&Words=standard%20deviation&Searchpage=0& Limit=25&Main=612037&Search=true&where=sub&Name=&d aterange=1&newerval=1&newertype=y&olderval=&oldert ype=&bodyprev=#Post612037) (Thanks BruceZ!) I calculated my SD both for bb/hr and bb/100. I also summed each of the used columns and used them to find the average $/hr and $/100.
As a result of that I got a mean of $43.53/hr or $39.82/100 (10.88 bb/hr 9.96 bb/100) and a SD of $124/hr or $121.84/100 (31 bb/hr 30.46 bb/100).
As I expected my SD was lower than PT's calculations using 1 table as a session. PT gives me SD/hr of $133 (33.47 bb) and sd/100 of $173 (43 bb).
Next I wanted to find confidence intervals. Because I had 38 sessions I assumed the distribution was normal. I got both 95% and 99% CI. I'll go through the steps I used for the /100 99% calculations, I used a similar method for the /hr and 95% calculations. For the CI I took 9.96+-2.576*30.46/sqrt(6563/100). 6563 is the number of hands I've played. This interval is (.27,19.64). The 99% interval for $/hr is (.58,21.19).
OK so here are my questions/issues:
- Have I made any mistakes in the methods mentioned above?
- Can I conclude that I am a winning player just based on this sample? If so, why do people often say that one cannot make this conclusion even after many more hands than this? If not, what more information is necesary to make this conclusion?
- I believe I should use this SD when calculating my ROR and so forth. Is this correct?
- My sessions are very heterogeneous. I have 3 of less than 10 minutes and 3 of greater than 200 minutes. How does that affect this if at all? Can I still use the normal distribution assumption given that this is the case?
Sorry this is so long. I hope that others will find it helpful.