When do you shoot a strategy? (and when do you just sit still)

Every system owner meets this night eventually: the thing has been losing for six weeks, and your finger hovers over the off switch. Here is the uncomfortable truth that decides it: a healthy strategy in a normal rough patch and a dead strategy look identical on the screen. The equity curve cannot tell you which one you own.

So the decision cannot come from the curve. It comes from two questions asked in order.

Question one - is anything actually outside the envelope? Before going live you should know your system's expected worst depth, expected drawdown length, and expected performance band. A losing streak inside all three is not information; it is Tuesday. Touch nothing.

Question two - only if something breached - does the reason this system ever made money still exist? Every strategy worth running has a one-paragraph explanation of its edge: a session habit, a structural behaviour, a cost someone else must pay. If that explanation still holds, you are inside variance, however ugly; reduce size if you must, but stay. If the mechanism itself is gone - the behaviour vanished, the costs changed, the crowd arrived - then the system is not "in a drawdown". It is over, and no parameter massage brings back a dead reason.

Between the two extremes, one asymmetry keeps accounts alive: when unsure, trade smaller - never revenge-tune, never panic-flatten. Retiring a system by procedure is not defeat; edges wear out the way tyres do, and changing a worn tyre is called maintenance.

One practical note, because question one has a prerequisite people skip: you cannot ask whether something is outside the envelope if you never measured the envelope. Your own closed history already contains it - every drawdown episode you have lived through, how deep each went, how long each took to recover, and how long the worst one kept you underwater. That is a measurement, not an opinion, and it takes a minute to produce. I published a small script for it in CodeBase this week; run it on your account before the next bad month, not during it.