Bayesian winrates

[jogsxyz]

Don't trust the 15bb/100 s.d. number. The poker utilities use an arcane method for calculating s.d. It uses the formula for live play, session-by-session. Online it's possible to compute the s.d. on a hand-by-hand basis. The old method biases the number lower by about 15%. A truer estimate of s.d. may be 18bb/100.

[jason1990]

That's interesting. I've known for some time that PT computes SD session-by-session. In the past, I've exported data from PT to MATLAB and computed SD hand-by-hand. Now that I think about, I seem to recall the MATLAB number was frequently, if not always, larger than the PT number, but I don't recall such a large discrepancy like you are describing. Off the top of my head, the only thing I can imagine which might cause such a bias is if the hands within a single session were negatively correlated. The rotating blinds seem to offer a natural negative correlation, but I can't imagine that's enough to make a noticeable difference.

[jogsxyz]

It's huge. Compute some data yourself.
Also there's more than one method to compute it.

When you bet and no one calls, is that bet part of the pot?