How To Read a No-Go Backtest

Abstract

A no-go backtest is not wasted work. It is a saved future mistake. The important question is why the candidate failed: bad signal, bad execution, too little support, too much concentration, poor robustness, or no incremental portfolio value.

Developers who record those reasons build a better research map over time.

No-Go Is A Diagnosis

The worst no-go report says only "failed." A useful report separates launch integrity, data coverage, trade count, PnL, drawdown, robustness, and portfolio contribution.

If a run completed all folds, passed artifact validation, and still failed, that is a clean research rejection. If the run had missing data or broken routing, it is not a strategy conclusion yet. It is an infrastructure issue.

Virtual Edges Can Disappear

One common failure is the virtual anti-signal. A developer flips the sign of historical PnL and sees a possible edge. Then the executable version fails because reversing direction changes contract selection, quote path, and fill behavior.

That is a useful lesson. PnL transformations are not always tradable transformations. A strategy has to survive as an executable decision stream.

Concentration Matters

A candidate can make money and still fail. If most of its value comes from one day or one narrow market pocket, promotion may be unsafe. Concentration is not automatically fatal, but it changes the evidence standard.

For options sleeves, also inspect whether the candidate adds independent days to the book or simply increases exposure when the active basket is already working.

Takeaway

Read no-go reports as research assets. They tell you which branches are closed, which infrastructure needs repair, and which attractive shortcuts do not translate into executable strategy logic.