> 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/the-judge-and-the-ruling.md).

# The judge and the ruling

## The judge and the ruling

The judge is the impartial reader. It never argues a side. It takes the full debate transcript, the shared corpus, and any verified evidence, and returns a probability distribution over the outcomes that sums to one. For a binary market the first-outcome probability is the headline number.

### Reading the debate

The judge weighs each advocate's case against the evidence, credits the arguments that hold up, and discounts the ones that do not. Its written reasoning is kept alongside the number, so the ruling is legible: you can see which claims carried weight and which were set aside.

### After the ruling

Three light post-processing steps run on the verdict, in order:

1. **Recalibration.** An optional temperature can sharpen or soften the distribution. The default leaves the judge's calibration as-is.
2. **Identity attachment.** Each probability is paired back to its advocate and outcome for the trace.
3. **Deviation cap.** If a scenario author set a maximum deviation band, the verdict is clamped into it. This is an author guardrail, not a default constraint.

### The model

The judge model is chosen by per-purpose routing rather than hardcoded, and independently of the advocate model, so the deployment can pin a strong model for the ruling while running lighter models for other steps. See [Cost and reproducibility](/nfa-docs/the-forecasting-engine/cost-and-reproducibility.md).


---

# 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/the-judge-and-the-ruling.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.