How OFC is calculated

How OFC Crusher evaluates placements, and what has changed in the calculation.

Core idea: evaluate on a shared sample

In OFC you never know the opponents' hands, so evaluations run over a shared, fixed set of dealt hands (a sampled world). A large set of boards is generated by a fixed procedure, and each placement candidate for your 13 cards is scored over that whole sample to produce an expected value (EV).

The dealt set is identical and fixed for every user. The same input yields the same numbers in any environment, and results do not waver from run to run.

Evaluation flow

1. Enter the hand and the seat (each of HU / 3-way uses its matching sampled world) → 2. Evaluate candidates (candidates are enumerated according to the cards and the street, and each is scored over the whole sample) → 3. Show EV and the candidate ranking → 4. Check the supporting information for the placement you are considering.

When Fantasyland (FL) continuation matters, the lines that keep staying in FL are evaluated over the same sampled world.

Calculation update of 2026-10-17

For the ranking of candidates, we updated the correction that keeps top candidates from being overestimated in some situations. The evaluation bias for Fantasyland continuation was updated as well. Both changes are already live in production numbers.

Nature of the numbers

EVs are model estimates and contain error. They neither indicate nor guarantee real-money profit. The details of the algorithm are not published, but the premises used for scoring (the shared sample, the same rules, the same model) are always the same.