What 144 real live cash sessions can and cannot show
By Tom Howcroft, live cash player and Overshove builder
One Overshove player has logged 144 genuine live cash sessions and 572 hours. The player confirmed these are real play records. We publish only the session count and hours, then show what a sample of that size can support.
The observed record
On 11 October 2026, the cash-game filter in the player's signed-in Overshove Stats screen showed 144 sessions and 572 hours. The player confirmed that the cash sessions were genuine play records, not tests or example entries. That is one account, not a survey or a representative sample of live players. The underlying session rows were not exported for this note.
We have not published the player's identity, venues, stakes, dates, winnings, losses or notes. The player approved publication of counts and hours only. The observed average is 572 divided by 144, or about 3 hours 58 minutes per session. That average describes time logged in this one account; it says nothing about skill or profitability.
Hours are not hands
A live cash log usually records hours rather than an exact hand count. If a table deals an illustrative 25 to 30 hands per hour, 572 hours would correspond to roughly 14,300 to 17,160 hands. At an illustrative 27 hands per hour, it would be 15,444 hands. None of those hand counts was observed in this account. Table speed, breaks, seat changes and game format can move the true figure substantially.
This distinction matters when comparing a live player's history with an online graph. Hundreds of live hours are a serious habit and still represent a modest hand sample for estimating a long-run win rate. Logging completeness is worth celebrating; statistical certainty is a different question.
An illustrative uncertainty check
For a simple independent-hand model, take an assumed standard deviation of 90 big blinds per 100 hands and the illustrative 15,444-hand midpoint above. That is 154.44 blocks of 100 hands. The model standard error of an estimated win rate is 90 divided by the square root of 154.44, about 7.24 BB/100. Multiplying by 1.96 gives an approximate 95 percent half-width of 14.2 BB/100 around an observed rate.
This is a worked model, not a measured confidence interval for this player. We have not used or published their win rate, and we have not measured their hand count or standard deviation. Real live results can be dependent, games can change and costs affect the cash result. An honest tracker should separate an observed record from assumptions used to interpret it.
What this example proves
It proves that one player kept a 144-session, 572-hour cash record in Overshove, and it demonstrates a repeatable way to discuss sample size without exposing their financial results. It does not establish a winning player, a typical session length for other players, or the performance of Overshove against competing apps.
If you want to use your own log for a stake or bankroll decision, first confirm that losing sessions and costs are present. Then inspect your actual hours and game mix. Treat hand count and volatility assumptions as assumptions unless you have measured them. The free variance and bankroll calculators on this site can illustrate scenarios; they cannot infer your true edge from an incomplete log.