What is a Good Win Rate in Poker (bb/100)?

Why this stat has no band

Every other figure on a stats report is a frequency you control: how often you raise first in, how often you continue on the flop, how often you get to showdown. Those settle in hundreds or a few thousand opportunities, which is why bands for them are useful. bb/100 is not a frequency, it is an outcome — of all those frequencies, of your opponents, of the rake, and of variance that dwarfs the signal. The registry behind our leak report records it as "Big blinds won per 100 hands — needs 10k+ hands to be meaningful", and gives it no band at all. Publishing one would mean grading you against a number your sample cannot possibly resolve.

How long a win rate takes to appear

The arithmetic is the same one the win rate calculator runs in reverse: hands = 100 x (1.96 x standard deviation / win rate)², with the standard deviation in bb/100. At the 100 bb/100 used by our own downswing simulation for 6-max online cash:

• •10 bb/100 — a crushing rate at micro stakes — needs about 38,000 hands before a 95% interval clears zero.
• •5 bb/100 needs about 154,000 hands.
• •3 bb/100 needs about 427,000, and 2 bb/100 about 960,000.
• •1 bb/100 — a real edge, and enough to beat many games — needs roughly 3,842,000 hands, which is more than most players see in a career.

A genuine winner spends a lot of time losing

Our downswing dataset simulates 100,000 careers of 100,000 hands at each win rate, at a standard deviation of 100 bb/100, from a fixed seed. The table above is the part that matters for reading your own graph. A true 5 bb/100 winner is still below zero after 10,000 hands 30.7% of the time, and after 50,000 hands 13.2% of the time. A true 2 bb/100 winner — a solid player in a rake-heavy game — is still losing after a full 100,000 hands 26.4% of the time. Even a 10 bb/100 crusher shows a loss after 10,000 hands 15.9% of the time. So a losing month is evidence of very little, in either direction.

Live win rates, and why they are unknowable

Live players ask this question more than online players, and they have the weakest answer available. A dealt table runs 25 to 30 hands an hour — an estimate, published with its range — against several times that on a single online table. Confirming a 5 bb/100 live win rate at that speed would take 5,100 to 6,100 hours in the chair; even a 10 bb/100 rate takes 1,300 to 1,500 hours. What live players can compute instead is the money: hourly = bb/100 x big blind x hands per hour / 100. A 10 bb/100 win rate in a $1/$2 game is $5.00 to $6.00 an hour before tips and travel, and a 20 bb/100 rate — the sort of figure a soft live pool can support — is $10.00 to $12.00. One caveat on all of it: the 100 bb/100 standard deviation above is an online 6-max figure, and a slower, more passive live game is not guaranteed to match it.

What to measure instead

If bb/100 needs six figures of hands and your frequencies need three or four, measure the frequencies. Every one of them has a band, each band is the one our leak report applies, and each settles fast enough to act on this month: PFR in about 300 hands, flop c-bet in about 400, WTSD in about 1,900. None of them tells you your win rate. All of them tell you whether the decisions that produce a win rate are sound, which is the question you can actually answer.

Check your own numbers

The win rate calculator takes the hands you have, returns a confidence interval, and says how many more you would need for the margin you want. The leak report grades the frequencies instead, which is the faster route to a decision. Both are free, and neither will tell you that a good win rate is a particular number, because nobody honestly can.