[Mendacious]

[ QUOTE ]
The point is, Mendacious, that a 12BB/100 winrate over 10K hands is more meaningful than a 5BB/100 winrate over 50K hands, in terms of deciding if you're a winning player.


[/ QUOTE ]

I am not sure I agree with this.

Here's why:

I think for this math to be accurate it requires SOME assumption about distribution of cards dealt/situations that are occurring-- especially in a PL/NL game where you can easily have hands with 100bb swings.

The math behind confidence intervals and standard deviations in cards is tremendously incomplete and inadequate as applied to NL and PL games and the almost infinite possible variations of how cards can be distributed between players and on the board.

It also does not take into account how many hands of 10k or 50k actually are significant.

As an example I offer the following of how far off this reasoning can be even assuming a completely even distribution.

Two individuals and a robot are tested to measure psychic ability. One indivicual has no psychic ability whatsoever, the other individual can predict accurately 54% of the time. The robot picks Heads everytime.

All three predict 10000 coin tosses, and the psychic predicts 5400 correctly and both the non-psychic and the robot predict 5000 +/- 50 correctly. You can probably measure with a pretty high rate of confidence that the psychic has a statistically significant edge. Let's assume, however that every 1000 tosses, one toss is weighted as equivilant of 200. (which is about the equivilant swing of getting stacked). Of those 10 tosses, 7 come out heads, and the psychic misses 6, the non-psychic gets 6 correct. What does this do to the confidence levels? Now it is necessary to toss the coin many more times to start to get meaningful information. And this assumes the type of normal distribution that you get flipping coins-- which is not remotely the same as the complex distributions you get playing poker.



My point is that over a run of 10,000 hands does not necessarily include much more than 100 truly significant hands. The nature of the distribution of luck in terms of both board and opponent cards is such that 100 significant hands doesn't tell you a whole lot.

MY POINT: When you have weighting, and you don't even take into account the number of hands it takes to get an even distribution of luck, the math is not necessarily that much more precise than experience in arriving at how many hands it takes to determine someone's win rate with a high degree of confidence.

Feel free to set me straight about this. It is always enlightening.