[rikz]

So, what would be the some of the concepts? Obviously hand samples won't help much since if you always make good decisions (in theory) then your negative variance would be the result of "bad beats" where the less likely outcome from the time you got your money in turned out to actually occur.

What about these.

Mean win rate and standard deviation in terms of BB/100 and BB/hour. Some interesting observations might come from identifying if super-multi-tabling yields higher or lower variance between players with similar styles and win rates.

Risk of ruin calculations relative to blinds structure, max buy-in, a player's personal variance over some sample size of hands, and total bankroll.

I think some interesting stuff could come from getting 100K hands from a PT database of LAGs with an + win rate and a TAGs with a postitive win rate for the same sample size and compare variance between styles. One would expect a larger variance for the LAG, but maybe a few samples might find no significant difference in variance.

Perhaps some calculations on how many hands at a given level is long enough to conclude with some degree of certainty (say 90%) that a positive or negative BB/100 win rate is due to something other than random variance. Is it 10K? 100K? 1000K? How does the degree of certainty increase as sample size increases with the "game" constant being same on-line site and same blind/buy-in structure.

I'll try to think up some more.