---
name: easyquant-audit
version: 0.1.0
description: >-
  Audit a quantitative trading strategy for credibility — overfitting risk,
  data quality, in-sample vs out-of-sample decay, and walk-forward evidence.
  Outputs a structured, evidence-based audit report. Never predicts returns.
metadata:
  api_base: https://api.easyquanttrading.com
  auth:
    type: header
    key: X-API-Key
    demo_token: easyquant-demo-2026
  endpoint: POST /api/audit/strategy
  billing: per-request credit (demo token: free)
  trusted_engineering: true
  tags: [quant, backtest, overfitting, audit, credibility]
---

# EasyQuant Audit Skill

Audit **someone else's quantitative strategy** and get a structured
**credibility report** — overfitting risk, data quality checks, in-sample vs
out-of-sample decay, walk-forward evidence, and a clear verdict.

> ⚠️ **Compliance red line (mandatory).** Every response must carry:
> *"Past performance does not guarantee future results. This is not investment
> advice."* The audit evaluates **trustworthiness / overfitting / data
> quality** — it never predicts returns, never promises profits, and never
> guarantees capital preservation.

## What this is for

This fills a gap most quant platforms leave open: **trustworthy backtesting
and anti-overfitting audit**. Instead of claiming "this strategy makes money",
it answers *"can you trust this backtest at all?"*

Use it to:
- Vet a strategy shared by a vendor, a community, or a past self.
- Check whether in-sample metrics survive out-of-sample scrutiny.
- Detect selection bias from parameter tuning.
- Get evidence-graded recommendations before relying on any backtest.

## Authentication

Send your API key (or the demo token during evaluation) as a header:

```
X-API-Key: easyquant-demo-2026
```

The demo token is free and rate-limited. Production keys are issued per credit
balance (see [billing](#billing)).

## Endpoint

```
POST {api_base}/api/audit/strategy
Content-Type: application/json
X-API-Key: <your-key>
```

## Request schema (all fields optional)

Provide as much as you can — the more complete the input, the more
authoritative the report. Unknown fields are reported as `unknown`, never
fabricated.

| Field | Type | Description |
|-------|------|-------------|
| `name` | string | Strategy name |
| `description` | string | Strategy idea / logic |
| `code` | string | Source code (optional context) |
| `dataset` | string | Dataset / symbol, e.g. `XAUUSD_H1` |
| `instrument` | string | Instrument |
| `timeframe` | string | e.g. `H1` |
| `parameter_count` | int | Number of tunable parameters |
| `trials` | int | Number of candidate configs tested (N) |
| `sample_size_bars` | int | Total bars in sample |
| `is_walk_forward` | bool | Whether walk-forward was done |
| `walk_forward` | object | `{method, folds, passed, oos_sharpe, note}` |
| `in_sample` | object | `MetricSet` for in-sample window |
| `out_of_sample` | object | `MetricSet` for out-of-sample window |
| `metrics` | object | Full-sample `MetricSet` if not split |
| `extra` | object | Any additional context |

**`MetricSet`** fields: `total_return`, `annual_return`, `sharpe`, `sortino`,
`max_drawdown`, `win_rate`, `profit_factor`, `trade_count`, `bars`, `start`,
`end`.

## Example request

```bash
curl -X POST https://api.easyquanttrading.com/api/audit/strategy \
  -H "Content-Type: application/json" \
  -H "X-API-Key: easyquant-demo-2026" \
  -d '{
    "name": "momentum_breakout_v3",
    "dataset": "XAUUSD_H1",
    "instrument": "XAUUSD",
    "timeframe": "H1",
    "parameter_count": 6,
    "trials": 500,
    "in_sample":   {"sharpe": 2.8, "max_drawdown": 0.12, "trade_count": 400, "bars": 3000, "start": "2020-01-01", "end": "2023-01-01"},
    "out_of_sample":{"sharpe": 0.4, "max_drawdown": 0.30, "trade_count": 120, "bars": 1000, "start": "2023-01-01", "end": "2024-01-01"},
    "is_walk_forward": false
  }'
```

## Response (abridged)

```json
{
  "schema_version": "easyquant-audit/0.1.0",
  "overall_verdict": "medium_risk",
  "overall_verdict_cn": "可信度一般——需补充时间留出/walk-forward 后再评估",
  "combined_risk_score": 44.5,
  "overfit_risk": {
    "score": 60.0, "level": "medium",
    "reasons": [
      "调参候选数 N=500，存在选择偏差风险（N 越大越需时间留出/去偏夏普检验）",
      "样本外夏普较样本内衰减超过 50%——表现可能无法外推（此处为同一拟合窗口内的时间留出检验，非真正冻结的样本外）"
    ]
  },
  "data_quality": {
    "score": 0,
    "checks": [
      {"check": "样本量", "status": "pass", "detail": "3000 根K线，样本充足"},
      {"check": "样本泄露", "status": "pass", "detail": "样本内外区间无重叠"}
    ]
  },
  "in_vs_out_of_sample": {
    "available": true, "gap_pct": 85.7, "verdict": "high",
    "label": "时间留出段衰减≥50%（高）"
  },
  "walk_forward": {
    "performed": false,
    "hint": "建议执行 walk-forward（滚动时间留出）检验，避免仅凭单一回测窗口下结论。所有检验均为历史回测/模拟，非实盘。"
  },
  "recommendation": "优先降低过拟合…；历史表现不代表未来，非投资建议。",
  "disclaimer_cn": "历史表现不代表未来，非投资建议。本报告为可信度筛查，不构成收益承诺或保本承诺。",
  "disclaimer_en": "Past performance does not guarantee future results. This is not investment advice.",
  "not_investment_advice": true,
  "no_return_guarantee": true,
  "no_capital_guarantee": true
}
```

## Interpret the verdict

| `overall_verdict` | Meaning |
|-------------------|---------|
| `low_risk` | Metrics look externally consistent, but still historical — no guarantee. |
| `medium_risk` | Needs out-of-sample / walk-forward evidence before reliance. |
| `high_risk` | Low credibility — fix data & out-of-sample issues first. |

## Billing

- Demo token: free, rate-limited.
- Production: each audit consumes 1 credit (checked **before** the audit runs).
  On insufficient balance the API returns `402 Payment Required` and does **not**
  charge or run the audit.

## Honesty contract

- We are a **credibility screening**, not a licensed audit and not an
  endorsement.
- Missing input → `unknown` fields, never fabricated conclusions.
- Every report carries the disclaimers above — there is no "guaranteed profit"
  output anywhere in this system.