Knowing When a Setup Has Stopped Working
Setups do stop working. Conditions change, an inefficiency gets crowded, the instrument that suited the rule behaves differently than it used to. Retirement is a normal outcome and not an admission of anything. The difficulty is that the early stages of a setup genuinely decaying and the middle of an ordinary losing run present in almost exactly the same way, and there is no test that separates them cleanly.
Why the Evidence Arrives Late
Deciding whether a strategy has stopped working means comparing its recent behaviour to its historical behaviour and asking whether the difference is larger than chance would produce. For an approach taking roughly one trade a session, the number of observations accumulates slowly, and the variation between stretches is wide.
The uncomfortable consequence is that by the time the evidence is strong enough to be convincing, a substantial period has already passed. There is no way around this that does not involve reacting to noise. Anyone who identifies decay quickly is also identifying it during runs that were nothing of the sort, and they will not be able to tell which was which.
Signs That Are Worth More Than Results
Because results are slow, the more useful indicators are structural. If the setup depended on a condition and that condition has visibly changed, the argument does not require a performance record. A rule built around a particular instrument's usual range behaving one way is aimed at something that is no longer there if the instrument's behaviour has shifted.
Changes in how the trades fail are also informative in a way the count of failures is not. A setup that used to lose small and occasionally lose large but now loses in a different pattern is behaving differently even if the totals happen to be similar. That is a change in the shape of the thing rather than in its outcome, and shape changes faster than averages do.
What Does Not Count
A run of losses of any length is compatible with an edge that still exists. So is a stretch where the setup fires much less often, which usually reflects conditions rather than the rule. So does the fact that some other approach has been performing better recently.
The most misleading signal is the feeling that the setup has stopped working. That feeling is produced by a losing run whether or not anything has changed, arrives at the same point in both cases, and is the reason so many setups get abandoned just before conditions turn. It is not evidence and it is very hard to treat as though it is not.
Deciding in Advance
The most practical protection is to define the retirement condition when the setup is adopted rather than when it is struggling. A stated threshold, whether in the number of sessions, the depth of a drawdown, or the disappearance of the condition the setup relies on, is a decision made without a position and without a recent loss attached to it.
Written down at the outset, it does two jobs. It makes stopping possible without an argument, and it makes continuing possible too, because a bad stretch that has not reached the threshold is explicitly within what was expected. Half the difficulty of a drawdown is not knowing whether it is normal, and a threshold written in advance answers that question at the point where it is being asked.
Stopping Without a Verdict
Retiring a setup does not require concluding that it never worked, and that framing causes real harm. It invites the interpretation that the time spent was wasted, which makes stopping feel like a loss to be avoided rather than a decision to be made.
A setup that worked and then stopped is the ordinary case. Conditions were suitable, then they were not. Recording the retirement with the evidence that prompted it leaves something usable, because conditions come round again and a setup shelved with a clear note about why can be reconsidered later on its merits. A setup abandoned in frustration leaves nothing behind but the frustration.