DERB

[Glenn]

"Your assertions would also seem to imply that DERB can not run a confidence interval on his own stats."

He can, but he would be using incomplete information to come to his conclusion. That's the problem. If he had no other knowledge of the situation other than his own win rate and SD, this would be his best bet. We are working with additional information. If you win 6-9 BB/hr with 99% confidence according to a Z-table, and no one else in the world wins more than 3 BB/hr, there is clearly greater than a 1% chance you win less than 6.

"but does this really mean that a sample of 100K hands are rendered meaningless just because we went looking for a LAG that is a winning player?"

They are not rendered meaningless. They are mitigated to a significant (but arguable in magnitude) extent.

[rigoletto]

Thanks! I appreciate it!

[UprightCreature]

If we really want to learn something about DERB what we need to do is look at his future hands from this point on. The data up to this point is what we used to select him, that data is biased. Future data is not.

If we flip 1000 coins 1000 times and have a coin that is 4-sigma away from the mean it may be an unfair coin, but because of selection bias we don't know (its in fact not shocking to have a coin 4-sigma out in this situation). To find out if the coin is fair we flip that coin another 1000 (or whatever) times and we can then apply a confidence interval to those results because they aren't biased.

So lets wait a few months and then look at DERBs data from today (or maybe the op date) and draw conclusions from that sample. That data won't suffer from selection bias.

[elindauer]

[ QUOTE ]
If we really want to learn something about DERB what we need to do is look at his future hands from this point on. The data up to this point is what we used to select him, that data is biased. Future data is not.

[/ QUOTE ]


Dead on. I hope someone who knows who this guy is actually does this!