> For the complete documentation index, see [llms.txt](https://nfa.gitbook.io/nfa-docs/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://nfa.gitbook.io/nfa-docs/the-forecasting-engine/advocates.md).

# Advocates

## Advocates

The debate is carried by advocates. Each one is assigned exactly one outcome and argues only for it. This is what forces the strongest possible case for every side into the open, instead of a single model quietly settling on a middle number.

{% hint style="success" %}
**This is measured, not asserted.** Across 137 resolved markets, the full engine scored a mean Brier of 0.289 against 0.404 for the same judge reading the same evidence with the debate withheld. See Accuracy and skill.
{% endhint %}

### One advocate per outcome

After the outcomes are framed, the engine casts one advocate per outcome. Each advocate is given:

* the **outcome** it must argue for,
* a generic **name** and **role** (an archetype, not a real person),
* a one-line **thesis**, and
* a short list of **research queries** to pursue.

A binary market produces two advocates. A large field produces one per named candidate plus one for the aggregated Other.

{% hint style="warning" %}
**Advocates are never real, identifiable people.** The caster is instructed to use a generic authority archetype or title, never a named individual, even when the research corpus surfaces one. The frontend renders the role and the thesis, never a real name. This is a hard product and legal rule, enforced in the prompt and guaranteed by a generic fallback advocate if casting ever fails.
{% endhint %}

### Information asymmetry

Each advocate then runs its own targeted research, separate from the shared corpus: a small number of queries, recency-biased, filtered to the subject, and de-duplicated against what the corpus already contains. The result is a per-advocate dossier.

This is deliberate. The judge reads the shared corpus, but each advocate also brings evidence the judge did not start with. The debate is therefore a genuine exchange of new information, not a rephrasing of one shared brief. It is also the mechanism that lets a well-supported minority outcome overturn a lazy consensus.

### The exchange

Advocates deliver opening statements, then exchange rounds of rebuttals. A rebuttal is framed as defense of its own outcome and offense against the others, and is steered toward the crux, the single decisive question the judge has flagged. The goal is not to win a debate on style, but to surface the evidence that actually moves the probability.


---

# 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://nfa.gitbook.io/nfa-docs/the-forecasting-engine/advocates.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.