Backtest

How a strategy performs over time against common rules. Five rules face the same simulated rival, from the same starting position, on the same seeded draws, so the gap at the end is the rule and not the luck. Change the rival or your costs mid-run to see which rules survive the change.

Cumulative value, round by round

0 of 120 rounds

Seed

Change the world mid-run

Standings

Press run. Every rule starts from the same round and the same rival.

The rules

Solver QRE: Best reply to the rival's whole move distribution, re-solved every round.
Best reply to last LAST: Assume they repeat last round, then play the best reply to exactly that.
Cost-plus COST: Own cost plus a fixed markup, snapped to the nearest level. Ignores the rival.
Match the rival MATCH: Play whatever they played last round.
Always-Nash NASH: Best reply to a rival assumed to optimise exactly, every round.

The rival, and the world they are in

Change these at any point. The run continues from where it is.

Rival precisionIn units of the rival's own payoff spread.How much the rival hedges instead of betting you repeatShare of rounds the rival plays at randomCost-plus markupYour cost to deliverRival cost to deliver

Sealed bid, lowest eligible bid wins, 8 levels in $4k steps, expected contribution per tender. The rival is a simulated player, not a record of anyone's behaviour, and a run is reproducible from its seed.

For the single-round answer from the solver, go to solve.