Operator brief · 431

Profitable and edge-bearing are two different claims.

The key idea

The confusion

One is an outcome. The other is a property.

Profit is a description of what a particular sequence of trades did. Edge is a claim about what the strategy tends to do, of which that sequence is one draw. The distinction sounds academic until capital is scaled against it, at which point it becomes the most expensive thing in the account: scaling a real edge compounds, and scaling a favourable sample amplifies variance that was always going to mean-revert. Every evaluation in MARS is built to keep these two claims separate, because the equity curve refuses to.

Why the curve cannot answer it

A rising line is consistent with several different underlying truths.

An upward curve is produced by a genuine edge, and also by a marginal edge that got a good run, and also by no edge at all with a favourable draw, and also by a negative edge whose losses have not arrived yet. The line does not distinguish between them because it has no access to the distribution it came from. This is why the evaluation pipeline never terminates at cumulative return. It decomposes the result into the structure that produced it, and asks whether that structure is the kind of thing that repeats.

FigureSame result, four different underlying claims

Schematic. Identical sample profit, entirely different reasons to believe it.

What the claim requires

An edge claim has to survive being attacked.

The evaluation asks the strategy to withstand tests a lucky sample cannot. Branch decomposition checks whether the profit came from the structures the plan intended or from one branch carrying everything. Rolling stability windows check whether the effect holds across the sample or lives inside one stretch of it. Outlier-robustness removes the best trades and looks at what is left. A profitable sample passes none of these reliably. An edge tends to survive all of them with its shape intact, which is the operational definition the system uses. Note that surviving these tests is not the same as passing them cleanly — a strategy whose robustness degrades but does not collapse is telling you something useful about how much of its result was structural.

The scaling asymmetry

The cost of confusing them is not symmetric.

Treating a real edge as a lucky sample costs opportunity — the strategy is under-deployed and earns less than it could. Treating a lucky sample as a real edge costs capital, and it does so precisely when the position size has been raised on the strength of the belief. The two errors are not equally priced, which is why MARS makes the edge claim carry the burden of proof rather than granting it by default when the account happens to be up.

Where it enters the machine

Deployment is justified by the claim, not by the balance.

The distinction is not philosophical housekeeping. Downstream, the throttle sizes against a model, the benchmark compares live behaviour to a simulated envelope built from that model, and the Weekly Scorecard grades whether the model still holds. All of that machinery is calibrated to an edge claim. If what was actually deployed was a favourable sample wearing an edge's clothes, every instrument below is now measuring live results against an envelope that was never real — and the system will report healthy readings for as long as the luck lasts.

The honest position

Most candidates are undecided, and undecided is a real answer.

The uncomfortable output of a rigorous evaluation is that a great many strategies land in neither camp: the evidence is genuinely insufficient to call it either way. The correct response is not to round toward the interpretation that permits trading. It is to keep the strategy at evaluation scale, keep accumulating evidence, and let the verdict arrive when it is actually earned. A system that only ever returns yes or no on demand is not measuring anything — it is manufacturing permission on a schedule. Holding a candidate at evaluation scale is cheap; the only thing it costs is the return that would have been earned had the edge been real, which is exactly the amount that should be risked on a claim that has not yet been established.

Connected inside MARS

Forex Strategy Evaluation →

Advanced EV Analytics Lab →

Weekly Trading Scorecard →

Monte Carlo Lab →

Every brief documents the same shipped system.

The complete MARS package — eleven workbooks, three TradingView indicators, the full manual library — $497.

Get MARS Pro — $497 Explore the Platform