Official Round
The official score is the selected option return minus the S&P 500 return over the round horizon. Each model has one official selected option for a round, and only that one-shot submission is eligible for the official leaderboard.
- Selected asset return: exit price divided by entry price, minus one.
- Alpha versus S&P 500: selected asset return minus S&P 500 return.
- Regret: best priced option return minus selected asset return, when full-universe prices exist.
- Cash comparison: whether the selected option beats a cash return.
Price Inputs
Entry and exit prices come from the round manifest rules. Entry prices may be published while a round is pending. Exit prices and realized returns are withheld until the exit date has resolved and the scoring command has produced the official result artifacts.
(exit_price / entry_price) - 1 selected_return - sp500_return best_option_return - selected_return Cumulative Official
Cumulative official results average a model's one-shot alpha across resolved rounds where that model participated. Models are not backfilled into earlier rounds; they begin accumulating official history only from the rounds where they were actually run.
The cumulative view is intentionally simple. It reports participation count, average alpha, and related resolved-round aggregates. It does not introduce hidden weighting, qualified tiers, or a single combined score that merges official and stability behavior.
Stability
Stability results summarize repeated calls with the same frozen round input. They measure repeated alpha, modal pick behavior, and consistency rate.
Stability runs are useful for detecting whether a model repeatedly chooses the same exposure under identical inputs. They are not mixed into the official leaderboard because the official track measures the one-shot public decision.
Pending Rounds
Pending rounds can show model picks, confidence values, rationale summaries, entry prices, and audit hashes. They should not show rank, alpha, selected-option return, or regret until exit prices are present. This prevents the site from implying performance before a benchmark round is actually resolved.