Cash-Game Rake and Win Rates: Count Costs Once
Keep units and cost definitions consistent
A decision EV expressed in big blinds is not a win rate in bb/100 hands. To calculate an observed win rate, divide net winnings in big blinds by the number of hands and multiply by 100. For example, a hypothetical net result of +50bb over 1,000 hands is +5bb/100 hands, not evidence that the long-run rate is +5.
Do not subtract rake twice
Check the hand-history or tracker definition. If winnings already reflect rake removed from pots, do not subtract that rake again. If starting with a genuinely gross result, deduct only costs not yet included. Keep rakeback, bonuses and time-based seat fees separately identified so comparisons use the same accounting. Do not subtract a guessed “leak loss” from observed net winnings: the played mistakes are already reflected in those results.
Rake percentages, caps, no-flop-no-drop rules and time charges differ by game. A solver model must match the applicable structure as closely as its controls allow. A simplified percentage-and-cap setting does not reproduce every room's rules.
Interpret the sample
No universal 20,000-hand threshold proves a strategy profitable. Total hands, relevant decision opportunities, variance, opponent selection and uncertainty all matter. Compare like-for-like stakes and formats and separate observed winnings, all-in adjusted results and solver estimates; none is interchangeable with a measured long-run edge.