[ QUOTE ]
When you have an abnormally tight player over a small sample some regression to the mean is likely to occur and this should be taken into account in range assignment.
You could call it an application of Bayesian decision making- given the prior knowledge of the actual distribution of tight players, the probability that he he is tighter than he appears is small and the probability that he is looser is large.
[/ QUOTE ]
yep exactly. OP is on the money IMO. when we see weird stats in small samples we have to remember that it is likely that these players are not truely as weird as they appear from a small sample, and we should skew our ranges slightly more in the direction of them being an average player. this might appear to contradict
a post i made a while ago, but i still think some stats have more validity in a small sample than people realise.
AF on the other hand seems to require a big sample, so is particularly appropriate for OP to comment on.
if someone statistical wants an idea for a great theory post, run some analysis on the common stats like vpip/pfr/af/wtsd etc and give us the confidence intervals for these at different values and sample sizes. it would be so useful!
for people who like graphs, here's a graph i made earlier summarising OP's point:
