Having less than 10k hands will lead to an error margin thats large enough where we shouldn't be to critical. For any sample size we can calculate the probability of your true winrate existing within a certain range, but with small samples it can be really wide.

As an example, lets use either 1k or 10k hand samples and a Standard Deviation of 20 BB/100

(Standard deviation is a measure of dispersion. You can find it in HM2 but as of now its expressed as 1/10th its actual value in HM2. I think they're going to fix that).

With 1K hands the 95% confidence interval:

= winrate +/- 1.96*Standard Error of the Mean

= winrate +/- 19.6*STDEV/SQRT(SampleSize)

= winrate +/- 1.96*20/SQRT(1000)

= winrate +/- 1.96*0.63

So, with 1k hands we can estimate with 95% confidence that your true winrate is within 1.23 Big Bets of the output value.

With a 10k sample we get our winrate +/- 0.39 BB/100 with 95% confidence.

This narrow range (10k sample) is much smaller and so while both will have a 95% accuracy as defined by our confidence interval of 95%, the larger sample is much more precise.

Technically its going to be even a little wider than this... but i've simplified by using z-scores (the 1.96 thing).