Free open-source strategy auditor for trading backtests.

Public preview: usable now as a local CLI and Streamlit app. Scoring thresholds and the hosted Overlay Preview may still change before v1.0.

Try without installing · Run Overlay Preview · Full audit report · Security

Upload a return curve, equity curve, or trade log. Get a fast QSX Score from 0 to 100, plus the checks that usually decide whether a backtest is worth more research.

• QSX Score and grade
• Overfit and too-good-to-be-true checks
• Buy-and-hold comparison
• Random timing test
• Monte Carlo stress test
• Shareable PNG scorecard
• Localized CLI, PNG scorecard, and web output (en, zh, ja, ko, es, pt-BR)
• Optional QSX Overlay Preview

Many strategies with a positive edge still fail because of poor risk sizing and exposure control. The optional Overlay Preview lets you test whether dynamic risk sizing may improve your own strategy.

QSX Strategy Score is not a replacement for QuantStats, pyfolio, or a full research notebook.

Use QuantStats when you want a detailed performance tear sheet. Use QSX Strategy Score when you want a fast screening answer:

Is this backtest worth deeper due diligence, or does it look fragile, lucky, overfit, or mostly beta?

The output is intentionally compact: one score, the main failure modes, a shareable scorecard, and an optional QSX Overlay Preview.

Scorecards link to a full audit-report workflow at quantscopex.com/report for users who want deeper due diligence after screening.

QSX Strategy Score can preview an external risk-sizing layer:

QSX Crypto Universal Position Engine 1.0

It is not an entry signal, exit signal, or coin selector. It is designed as an overlay that sits outside your original strategy:

original strategy returns x QSX dynamic exposure = overlay-adjusted curve

Research audit example:

Your result may differ. The purpose is to test whether the overlay improves risk-adjusted performance on your own strategy.

Overlay Preview rejects trade logs with overlapping per-position trades. Upload an equity curve or daily return series so the preview uses the aggregate strategy path.

From this repository:

git clone https://github.com/jianweiweng05/qsx-strategy-score.git
cd qsx-strategy-score
python -m pip install -e ".[app,excel]"

Score a strategy:

qsx-score examples/strategy_alpha.csv --asset BTC --lang en

Supported languages:

qsx-score examples/strategy_alpha.csv --lang zh
qsx-score examples/strategy_alpha.csv --lang ja
qsx-score examples/strategy_alpha.csv --lang ko
qsx-score examples/strategy_alpha.csv --lang es
qsx-score examples/strategy_alpha.csv --lang pt-BR

Export a PNG scorecard and JSON report:

qsx-score examples/strategy_alpha.csv --asset BTC --out card.png --json report.json

Run the web app:

streamlit run app/streamlit_app.py

QSX Score: 59 / 100
Grade: NEEDS WORK

Headline:
Indistinguishable from random timing (p=0.32)
No proven timing edge.

Key problems:
- Max Drawdown: -83%
- Random timing test failed
- High dependency to buy-and-hold: corr +0.90, beta +0.82

This does not mean the strategy is useless. It means the uploaded return path looks more like asset beta plus risk exposure than proven timing edge.

Returns:

date,return
2021-01-01,0.012
2021-01-02,-0.004

Equity curve:

date,equity
2021-01-01,10000
2021-01-02,10120

Trade log:

entry_time,exit_time,pnl_pct,side,symbol
2021-01-01,2021-02-01,3.2,LONG,DOGE

CSV, TSV, Excel, TradingView-style exports, return series, equity curves, and closed-trade logs are supported.

• Scoring model
• QSX Overlay Preview
• Privacy boundary
• Security policy
• Methodology notes

Free = screening. Pro = due diligence.

Free = screening, not proof. QSX Strategy Score is computed from the uploaded performance path; it does not inspect strategy code, raw market data, execution simulation, exchange fills, or the full parameter-search process.

It cannot prove the absence of code-level look-ahead, survivorship bias, unrealistic fills or slippage, hidden leverage, capacity constraints, manual selection, or train/validation contamination.

A high score is not investment advice. A flagged score does not prove a strategy is fake; it means the backtest method should be checked before trusting the result.

MIT.