> For the complete documentation index, see [llms.txt](https://oracledocs.magpie.gg/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://oracledocs.magpie.gg/the-oracle/accuracy.md).

# Accuracy — we grade ourselves

**`GET /oracle/v1/accuracy`** — public, keyless, recomputed nightly.

## The methodology (walk-forward, out-of-sample)

Every priced, catalog-mapped realized sale is scored against what the oracle claimed **strictly before that sale existed**:

* **Published track**: the sale vs the last *published* value (max 14 days old) from the anchored daily history — including whether the sale landed inside the published `low..high` band (the band's honesty rate).
* **Walk-forward track**: the sale vs what the production estimator would have said using **only earlier sales** (minimum 2 priors; thinner series are *skipped and counted*, never guessed).

No sale is ever compared against a value it influenced. The anchored history (see [Verifiability](/the-oracle/verifiability.md)) makes the metric impossible to retro-fit: the values being graded were committed on-chain before the sales happened.

## How to read the numbers

```json
{
  "oracle_published":     {"n": …, "medape_pct": …, "p90_ape_pct": …,
                           "bias_pct": …, "band": {"n": …, "hit_pct": …}},
  "realized_walkforward": {"n": …, "medape_pct": …, "p90_ape_pct": …,
                           "bias_pct": …, "skipped_thin": …, "by_game": {…}}
}
```

* **medAPE** (median absolute % error) is the headline — robust to a single mis-mapped slab. **p90** shows the tail. The mean (MAPE) is reported but never the headline, by design.
* **bias** — signed error: are we systematically high or low?
* **band hit rate** — how often reality landed inside our published band.
* **skipped\_thin** — how many sales we *refused* to score for lack of prior data. Publishing this number is the point: coverage can't hide behind a rate.

## Why publish this at all?

Because nobody else does, and because a reference price without a published error is marketing, not data. Integrators (marketplaces, lending protocols) can quote our error instead of asking users to trust a black box — and the metric improves mechanically as realized coverage grows, in public.


---

# 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://oracledocs.magpie.gg/the-oracle/accuracy.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.