Rigour. Reproducibility. Realism.
Every study begins as a hypothesis. We formalise it into a set of unambiguous trading rules, code them into our backtesting pipeline, and run them against tick-level historical data. The output — equity curves, trade logs, and performance metrics — is analysed statistically, interpreted candidly, and published in full.
No black boxes. No cherry-picked results. No skipped steps.
Every study is held to the same five standards — regardless of outcome.
| Standard | What It Means |
|---|---|
| IS/OOS Split | 70% of data for development, 30% held out for validation. We never test on data used during development. |
| Monte Carlo Simulation | 1,000+ randomised trade sequences to estimate the distribution of possible outcomes — not just the one path that happened. |
| Parameter Stability | Parameters are varied across a defined grid. If performance collapses outside a narrow optimal zone, that is a red flag for data-mining, not a valid edge. |
| Transaction Costs | Realistic slippage and commissions per instrument. ES: 1 tick slippage, NQ: 2 ticks. Commissions: $2.50 per side per contract. Every result is net of costs. |
| Multiple Timeframes | A strategy must hold up across resolutions. If it works on 1-minute but fails on 5-minute and 1-hour, the edge is not structural. |
| Parameter | Detail |
|---|---|
| Tick data source | Databento MDP3 |
| Instruments | ES (E-mini S&P 500), NQ (Nasdaq-100), CL (Crude Oil), 6E (Euro FX) — others on request |
| Historical depth | 2020–present, growing with each new data ingestion cycle |
| Resolution | Tick → 1m / 5m / 15m / 1h / Daily |
| Default slippage | ES: 1 tick · NQ: 2 ticks (configurable per request) |
| Default commissions | $2.50 per side per contract (configurable per request) |
| Execution model | Bar-close execution with configurable order delay |
No. We test yours. We never sell strategies, indicators, signals, or subscriptions. Our revenue comes from backtesting services — not from the performance of any strategy.
We use tick-level data, realistic slippage and commission models, Monte Carlo simulation, and strict IS/OOS validation. TradingView bar-replay runs on OHLC data with no slippage model. That difference matters — especially for high-frequency or intraday strategies where tick dynamics change outcomes.
Typically 3–10 business days depending on complexity, data volume, and our current queue. Simple single-instrument tests are faster; multi-parameter or cross-instrument studies take longer.
Not in MVP. Our current data infrastructure covers futures: ES, NQ, CL, 6E. Crypto may be added in a future phase.
That is the most valuable outcome. A clear, data-backed "this does not produce an edge" saves you time, capital, and the cognitive bias of believing otherwise. We publish negative results — they are not failures, they are successful falsifications.
Yes. All our published research is available on the Research page. You will see the exact format, rigour standards, and candid language before you submit anything.
Have a question we did not answer?
Contact Us →