Why not just backtest it yourself?

You can, and you should

Nothing about running a backtest is hard any more. The libraries are free, the data is cheap, and a few hundred lines of code will show you exactly what your idea would have done. We're not going to pretend that step is a moat, or that you need us to draw a cumulative return chart.

But notice what that backtest actually told you: what one set of rules did on one slice of history that you already know the shape of. That is all it told you. The difficulty was never producing the chart. The difficulty is trusting it.

The hard part is the search

Anyone can optimise a strategy until the historical chart looks good. The hard part is knowing whether you found an edge — or simply searched until you found something that looked like one.

Every variation you try is another ticket in a lottery. Change the stop, move the entry time, add a filter, widen the target — twenty tries in and one of them will look brilliant by luck alone, on data that has already been seen. You cannot feel the difference from inside, because the lucky one arrives with the same confident equity curve a real edge would have. The search that produced the result is invisible in the result.

This isn't a discipline problem you can solve by being careful. Statisticians have a name for it — multiple testing — and the correction is arithmetic, not willpower: the more things you tried, the higher the bar a result has to clear before it means anything. Most backtesting setups don't count the tries at all.

Two paths, one of which knows what it did

The sequence is the point. Nothing can skip a stage, and the stage that separates Mithos from a good backtest script is The Mithos rigor harness — five checks, run automatically on every strategy, with no setting to turn them off.

What the harness actually does

• 01 Out-of-sample testing — part of the history is held back and never used to build or tune the strategy, so there is always data it has not seen.
• 02 Walk-forward validation — the strategy is rebuilt and retested across successive windows, checking the result survives forward in time rather than only in the window it was fitted to.
• 03 Placebo controls — compared against random entries with the same risk, costs, and exits, to check your entry signal is actually doing the work.
• 04 Multiple-testing penalty — the bar you need to clear rises the more variations have been tried, so trying enough things until one looks good doesn't count as finding an edge.
• 05 Robustness analysis — concentration and dispersion checks make sure the result isn't being carried by one or two lucky trades.

And because honesty cuts both ways, the harness can reach a third outcome: undecided. When there isn't enough evidence to say either way, it says so, rather than dressing up uncertainty as a recommendation.

When you genuinely don't need Mithos

If you write down one idea, once, with fixed rules — no tweaks, no variations, no second pass — and you only want to know what it would have done, a script is fine. You haven't searched, so there's nothing to correct for.

Almost nobody works that way. The moment you iterate — and iterating is how strategy development actually happens — every pass quietly raises the bar, and only a harness that counts the passes knows where it stands. That is the job: not running backtests, but keeping score honestly while you search.