How we try to prove ourselves wrong.

01 Where ideas come from

We source hypotheses from academic literature, market structure, and our own trading. Every idea enters the same pipeline; none get a pass. A paper with a thousand citations gets the same treatment as a hunch from the desk — the pipeline doesn't care about pedigree.

02 How we test

Every candidate runs the same gauntlet: clean historical data, realistic cost and slippage modeling, strict in-sample / out-of-sample separation, and validation checks designed to catch the classic self-deceptions — lookahead, overfitting, survivorship, regime luck. The out-of-sample data stays sealed until a strategy has earned the right to touch it.

The framing that governs all of it: we are not trying to prove a strategy works. We are trying to prove it doesn't, as cheaply as possible.

03 How strategies die

Most of them. Kill criteria are set before testing starts — degraded out-of-sample behavior, sensitivity to parameters that shouldn't matter, costs eating the edge, results that can't be reproduced from the logs. When a strategy dies, we write it up with the same care as a survivor: what we expected, what happened, and what the autopsy showed.

04 How we report

Every published backtest carries its assumptions — data, costs, slippage, window — and the required hypothetical-performance disclaimer, at full prominence, every time. Simulated results are never blended with live results. When live results exist, they're labeled live and reconciled against the backtest publicly, drawdowns included.

This is the disclaimer that travels with every backtest we publish:

05 What we do with survivors

We trade them with our own capital. That's the whole exit path right now. Down the road, some systems that survive testing and run live may become standardized, licensed strategy software — the honest version of that story lives here — but nothing goes on sale until it has survived the full gauntlet and months of disclosed live trading.

Trading futures, and trading in general, involves substantial risk of loss and is not suitable for everyone. Full risk disclosures.