Edge Audit
Send me your trading strategy, backtest, or bot. I run it through the same falsification gauntlet I built for my own research — and give you an honest written verdict: is the edge real, or is it an overfit that will die live?
What you get
- A clear verdict — real edge / overfit / inconclusive, stated plainly.
- Walk-forward out-of-sample results — optimized on train, tested on the unseen window, rolled across regimes. The test single backtests never show.
- Cost reality — performance net of realistic fees + slippage, and at 2× stress.
- Robustness — does it beat a random baseline, and does it hold when you nudge the parameters?
- What would kill it — the regime, cost, or crowding that ends the edge.
- A written report (PDF/Markdown) with the numbers, so you can act on evidence.
How it works
- You send the rules (or code/backtest) + the market + timeframe.
- I reproduce it no-lookahead and run the gauntlet: economic rationale → random baseline → walk-forward OOS → parameter-plateau → fee/slippage stress → capacity.
- You get the report. If it's overfit, I show you exactly where. If it survives, you'll know how thin and how regime-dependent.
The engine behind this is the same one that tested 8 edge families across 3 live feeds and 3 timeframes and rejected all of them — read the write-up. Methodology is open; the judgment is the service.
Packages
Reality Check
- One strategy, one market
- Walk-forward + random baseline
- Net-of-fees + 2× cost stress
- 1-page verdict, 3–5 day turnaround
Best for a quick "is this real or am I fooling myself?"
Full Edge Audit
- Deep multi-regime walk-forward
- Parameter-plateau + capacity/scaling
- Multi-market / multi-timeframe checks
- Full written report + a call to walk through it
Best before you deploy real capital or buy someone else's bot.
Custom Research / Build
- Bespoke edge research or a paper-first bot
- Built with the same falsification rigor
- Honest go / no-go before anything goes live
Best if you want a system built — and validated — properly.
Prices are starting points; scope sets the final quote. Pay only when the report is delivered.
Who this is for
- You have a strategy that backtests well and want to know if it's real before risking money.
- You're about to buy a bot / signals / "system" and want an independent, honest second opinion.
- You want a paper-first system built by someone who optimizes for truth, not a sales narrative.
Who it's not for: anyone wanting a guaranteed winner or a number talked up. I report what the data says, including "this doesn't work" — that honesty is the product.
Tell me about your strategy — market, timeframe, and what you've seen so far.
Start an Edge Audit →DM on X, or open an issue / discussion on GitHub.