I use sample size in my HUD. But I use an estimation based on my feel and the math to decide how reliable the stat is.

For example, if a person 3bets 80% of the time, but the sample is only 5 hands. Compare this to another who 3bet 30% of the time but the sample is over 10 hands. Who is more likely to be 3betting a looser range?

Then consider what would happen if, in this spot, the average 3bet is only 10% over your 200k hand sample. And you know from experience that no one ever has a 3bet over 30% anyway.

Is there a nice convenient formula we can plug all these variables into in order to give us the 3bet range our opponent likely has with 70% (or some other %) confidence?

I feel like bayes' theorem is only part of the equation, because it assumes that the probability value of A and B to have 100% reliability, with no variance. Hopefully someone more mathematically knowledgeable than me can help?