Sample sizes for stats

[bay]

I don't know if this has been done before-if so please can someone link me to the thread- but I thought it might be useful for there to be a thread where the statistically trained among us could set out sensible sample sizes for HUD stats. My plan is to use the 'minimum samples' feature in the HUD settings more than I do at present so I can avoid being distracted or misled by statistics which are too small to be reliable.

Of course we can and all guess at what is a decent sample for a given stat and we all all know, for example, that VPIP/PFR are going to be more meaningful after a small number of hands than say fold to river raise . But it would be good if someone with real statistical knowledge could say at what point we could start to give credence to the various stats.This does not mean 100% statistical reliability but just enough to say that a sample is enough for a stat to be meaningfully indicative.

To start off, I am interested in the following:
WTSD
W$WSF
Fold to 3 bet
4 bet range
Fold to 4 bet
Flop cbet fold
Raise flop cbet

Many Thanks

[tob1as1]

I'd like to see this one too.

[Patvs]

For any stat, I use the low min sample size of just "3".


As an example: Fold BB to Steal, when this stat shows "100%", or "0%" (based on 1 to 2 hands) or 50% based on two hands, it's going to cost me money to experiment myself to find out how often this opponent folds his BB to a Steal.
(by stealing with any two cards if he has a 100% Fold BB to Steal)

But if it's still 100%, (or 66%) based on 3 hands I am just going to attempt to steal all his big blinds.
When this fails (and the percentage drops to 50%) I can reevaluate (and stop stealing :)) .

If it stays 100% (or increases to 75%) I continue to steal with any two cards.


---

You can basically use this tactic for any stat, based on only a min sample of 3.