Let's say you flopped 0 sets over 1,000 hands with pocket pairs. A serious statistical anomaly. Then over the next 9,000 hands, you flopped the normal number over every 1,000 hands: exactly 117.
Here's your data set in format (X, Y) where X = pairs, Y = sets
(1000, 0) ==> 0.00% sets
(2000, 117) ==> 5.85% sets
(3000, 234) ==> 7.80% sets
(4000, 351) ==> 8.78% sets
(5000, 468) ==> 9.36% sets
(6000, 585) ==> 9.75% sets
(7000, 702) ==> 10.03% sets
(8000, 819) ==> 10.24% sets
(9000, 936) ==> 10.40% sets
(10000, 1053) ==> 10.53% sets
This data set is "regressing to the mean." By percentage, you are getting closer and closer to 11.7 percent of sets flopped. But by number of sets flopped, you are remaining exactly 117 sets lower than would have been expected. There is no time period over which you (can expect to) flop more sets than expected.
In short, "regression to the mean" simply does not mean that you ever get to flop more sets than expected, over any time period.Statistics: Posted by k3nt — Mon Oct 09, 2006 11:05 am
]]>