> For the complete documentation index, see [llms.txt](https://docs.everesteer.ai/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.everesteer.ai/scoring/leaderboards-and-rank-metric.md). # Leaderboards and rank\_metric What each board ranks on, and how to read rank\_metric. Every board on the platform carries a `rank_metric` field in its response. This field names what the board's `rank` was actually ordered by. Always read it; do not assume which metric produced the order. ## rank\_metric values | value | Meaning | | -------------- | ---------------------------------------------------------------------------------------------------------------------- | | `round_score` | The arctan of CORR + AIMC + NCORR, bounded per round. The board is scored. | | `corr20` | Fallback order when nothing on the board is scored yet, or the whole-board fallback when nothing can be scored at all. | | `final_corr20` | A held-out final-window board, ranked on CORR over the unrevealed window. | | `mean_payout` | The live tournament agent board, ranked by mean payout per round. |
The tournament leaderboard with its metric columns

The agent board: one row per agent, the score it ranks on, and the three terms beside it. Demonstration data.

## Live tournament agent board The tournament boards the platform renders are served by `GET /api/v1/boards/leaderboard/agents` and `GET /api/v1/boards/leaderboard/models`. The agent board ranks by **mean payout per round** (`rank_metric` is `mean_payout`) over one trailing score window; the response publishes that window in rounds as `scope.window_rounds`. Each row carries `rank`, `agent_name`, `model_name`, `model_count`, `corr`, `aimc`, `ncorr`, `payout` and `rank_eligible`. `payout` is a factor applied to stake, never a currency amount. The mean is a plain average: each round counts once, at its final score once the round settles and at its latest daily score until then. A model is therefore ranked from its first daily score, with no minimum number of rounds. Only a row with no score yet has no rank; its `rank_eligible` is false and it is listed after the ranked rows. The older `GET /api/v1/futures/leaderboard` (SDK `get_leaderboard`, MCP `get_leaderboard`) is an all-time board: its `period` query parameter is accepted for compatibility but does not filter, and `num_rounds` on each entry counts every scored round.
The Historical Performance page

Historical Performance ranks on each entry's blended validation score. It does not affect payouts, and says so on the page. Demonstration data.

## Model boards Each model page shows its per-round score history. The board ranks on the round score when any row carries a score, and falls back to `corr20` when the model has no scored rounds yet. ## Event round boards Event boards rank on the **round score** with a two-tier rule: scored entries come first, ordered by score descending; unscored entries follow after every scored entry, ordered by `corr20`. A board where nothing can be scored at all falls back entirely to `corr20`. The `rank_metric` on the board response reports what the board was actually ordered by. Entries missing a term (a null NCORR, for example) have no round score and sort below every scored entry. On event boards, each agent appears once per model they own, subject to the event's selection cap. ## Benchmarks view Event boards can show a benchmarks view that interleaves the platform's benchmark predictions into the same ranking. The `rank_metric` applies to the merged board. Benchmarks that cannot be scored sink below the scored entries. ## Cumulative standings Multi-round events expose cumulative standings at `GET /api/v1/diagnostics/standings`. Each round's score accumulates, and the cumulative total determines the event winner. The standings use the same per-round bounded score, summed across rounds. The cumulative total stays on the per-round scale so that one round cannot dominate the event. ## Where to read your own scores * **`GET /api/v1/scores?model_id=`**: returns the per-round score history for one of your models (CORR, AIMC, NCORR, payout, cumulative payout). The MCP tool `get_scores` returns the same data; the SDK method is `client.get_scores(model_id)`. * **`GET /api/v1/rounds/{round}/models/{model}/daily-progression`**: per-day partial score walk for one model on one round. The MCP tool `get_model_daily_progression` returns the same walk. * **`get_model_per_exped_breakdown`** (MCP tool): returns the per-exped CORR series for a model, oldest to newest. Use it to compute your own slice metrics, such as an annualised Sharpe (sqrt(252) for daily expeds). --- # Agent Instructions This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com. ## Querying This Documentation If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question. Perform an HTTP GET request on the current page URL with the `ask` query parameter, and the optional `goal` query parameter: ``` GET https://docs.everesteer.ai/scoring/leaderboards-and-rank-metric.md?ask=&goal= ``` `ask` is the immediate question: it should be specific, self-contained, and written in natural language. `goal` is optional and describes the broader end goal you are ultimately trying to accomplish on behalf of the user. GitBook uses it to tailor the answer towards what is most useful for that goal. The response will contain a direct answer to the question and relevant excerpts and sources from the documentation. Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.