Do not judge the machine by one output.

The sentence

Judge the operating condition, not the last result.

The doctrine is compact: do not judge the machine by one output — judge it by the full operating condition of the machine. As a slogan it is agreeable and easy to nod at. As a rule it is demanding, because a trading account produces outputs constantly and every one of them arrives with an implicit invitation to draw a conclusion. Taking the sentence literally means declining almost all of those invitations, which is a far stranger discipline than it first appears. The difficulty is not intellectual agreement, which is immediate and universal; it is that the invitations arrive several times a day and each one individually seems too small to matter.

What it forbids

Most of the ordinary conversation about trading.

Applied strictly, the doctrine rules out a long list of familiar behaviours — not because they are undisciplined, but because each infers machine condition from a single output. The list is worth stating explicitly, because traders who agree with the principle routinely do all of these within the same week.

• Concluding a setup is broken after three losses.
• Increasing size because the last five worked.
• Judging a week by whether it was green.
• Retiring a branch on a bad month with no drift evidence.
• Trusting a strategy because a recent trade paid handsomely.

What counts as operating condition

Five readings, taken together, over a window that can support them.

The alternative to output-judging is not vagueness — it is a specific panel. Operating condition in MARS means expectancy and its stability, drawdown depth and duration, risk-adjusted profit quality, deployment efficiency, and the direction of equity acceleration, all read over a window with enough trades to mean something. The reason five numbers appear rather than one is structural rather than thorough-for-its-own-sake: each is capable of looking healthy while the machine deteriorates, and each was selected for what it stops the others from concealing.

Why one output is so persuasive

Recent results feel like evidence because they are vivid, not because they are informative.

A single result carries almost no information about a distribution and enormous emotional weight, and those two facts move in opposite directions. The most recent trade is the one the operator watched, felt, and remembers in detail — which makes it feel like the strongest available evidence when it is close to the weakest. This is why the doctrine has to be structural rather than attitudinal: no amount of knowing about recency bias prevents the most recent output from feeling decisive while it is happening.

The cost of the discipline

You will hold a system through stretches that feel wrong.

The doctrine has a genuine price and it should be stated rather than glossed. Judging by operating condition means continuing to deploy a system during losing runs where every output says stop, and it means declining to size up during winning runs where every output says press. Both feel like errors while they are being made. What makes it bearable is not conviction but evidence — the panel says the machine is intact, the benchmark says this drawdown is inside the expected band, and the drift readings say nothing structural has changed. It is worth naming that this cost is front-loaded: the discipline is hardest in the first year, before the operator has personally watched the panel be right about a stretch their instincts read wrongly.

The one output that does matter

A rule violation is a machine reading, not a result.

There is a single exception worth naming. An override taken without documentation, a position sized outside its tier, a stop moved after entry — these are not outputs of the machine, they are changes to it, and they warrant immediate attention regardless of how the trade resolved. The doctrine says do not judge the machine by one result; it says nothing about ignoring one act of dismantling it. Confusing the two is how the principle gets misused as a reason to overlook the thing that actually needed the reaction.