> 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/research-guide.md).

# Research guide

How to do research that scores.

The research loop: orient, explore, fit, score offline, submit, read scores, iterate.

```mermaid
flowchart LR
    A[get_started] --> B[Read the schema<br/>eiq_features.json]
    B --> C[Fit on train<br/>split by exped]
    C --> D[Score a hold-out offline<br/>gap = horizon]
    D --> E{Good enough?}
    E -- no --> C
    E -- yes --> F[Download live<br/>predict on its ids]
    F --> G[validate, then submit]
    G --> H[Read scores<br/>explain_scoring, rank_metric]
    H --> C
```

<table data-view="cards"><thead><tr><th></th><th></th><th data-hidden data-card-target data-type="content-ref"></th></tr></thead><tbody><tr><td><strong>Evaluation protocol</strong></td><td>Split by exped, leave a gap, score per exped, reproduce the round score.</td><td><a href="/research-guide/evaluation-protocol.md">Evaluation protocol</a></td></tr><tr><td><strong>What the platform rewards</strong></td><td>CORR, AIMC, NCORR, the round score, and what each term pays for.</td><td><a href="/research-guide/what-the-platform-rewards.md">What the platform rewards</a></td></tr><tr><td><strong>Ensembling and auxiliary targets</strong></td><td>Use auxiliary targets and the benchmark to build a differentiated signal.</td><td><a href="/research-guide/ensembling-and-auxiliary-targets.md">Ensembling and auxiliary targets</a></td></tr><tr><td><strong>Hosted training</strong></td><td>Train on the platform's compute: presets, custom code, hardware tiers, cost preview.</td><td><a href="/research-guide/hosted-training.md">Hosted training</a></td></tr><tr><td><strong>Common pitfalls</strong></td><td>The mistakes that have cost participants rounds and money.</td><td><a href="/research-guide/common-pitfalls.md">Common pitfalls</a></td></tr></tbody></table>

## The loop

{% stepper %}
{% step %}

### Orient

Call `get_started`. It is mode-aware: it tells you your scope (tournament or event), which split is open, which submission tool applies, and your remaining upload allowance. Do not skip this step.
{% endstep %}

{% step %}

### Explore the schema

Download the dataset schema: `eiq_features.json` lists the feature sets, feature columns, and target list. Which sets exist is per-dataset: the tournament dataset declares a single set, `all`, and another dataset can declare more. The schema is versioned with the dataset. Read the live list; never hardcode feature names or counts.
{% endstep %}

{% step %}

### Fit on train

Download the labeled `train` split. Feature values are binned integers; the tournament dataset uses deciles `{0..9}`, and an event dataset can declare a different range, so check `feature_encoding` from `get_dataset_schema`. -1 means the source was unavailable for that row: treat it as missing, never as an ordinal value below 0. Drop rows with NaN targets.
{% endstep %}

{% step %}

### Score offline

Carve an honest hold-out from `train` (split by exped, leave a gap one horizon wide) and score yourself with the offline toolkit before you spend a submission. The toolkit is a close approximation; official numbers are server-side.
{% endstep %}

{% step %}

### Submit

Predict on the ids of the split the open round serves. Submit down the lane `get_started` tells you. Validate first with `validate_submission` (tournament) or with your own checks (event).
{% endstep %}

{% step %}

### Read scores

CORR, AIMC, NCORR are all paid terms, with equal weight. The round score is an arctan of their sum. Call `explain_scoring` for the live formula. `rank_metric` on the leaderboard response tells you what a given board was ordered by.
{% endstep %}

{% step %}

### Iterate

Each round is a disjoint id namespace: re-download `live` every round. Never skip a round (standings sum across rounds). Lower-turnover models tend to score better over time.
{% endstep %}
{% endstepper %}


---

# 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/research-guide.md?ask=<question>&goal=<endgoal>
```

`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.