From payouts to a decision.

The central question

Can Jev infer whether a bet is worthwhile from its payout table, and do recent wins or losses sway that choice? The live lab supplies structured payouts, fair-die rules and six controlled recent outcomes. Calculated EV, bankroll and stake fields are not sent.

One click, one decision

Roll sends one Bet/Skip question to Jev. The decision is saved before the die outcome is generated. The rolled payout row is highlighted and the next offer appears. Autoplay stops after 30 rolls. Inspect the payout heading or expand a completed log entry for exact request details.

Recorded experiments

Each history/magnitude combination contains 1,000 recorded offers. All combines the four conditions into 4,000 samples at the selected magnitude. The six-roll histories are controlled inputs, not a rolling feed of new simulation outcomes.

Expected profit and the benchmark

Bets have fixed size. The experiment curve adds the offered EV when Jev bets and zero when it skips. Optimal EV adds only positive offered EV. Actual outcomes do not enter either curve. A win does not make a negative-EV bet correct.

The supplied-EV control

Adding calculated EV led to 100% correct actions in the recorded sample. Findings & discussion preserves that result and its limits. The live lab now focuses on the unresolved task of calculating EV from payouts.