Halfway into a losing run, two explanations fit the evidence equally well. The method is fine and this is the sample it produces from time to time, or something in the market or in your execution has changed and the method no longer works. Both produce the same equity curve. Waiting for the statistics to resolve the ambiguity means waiting a long time, and quite possibly through a great deal more loss.
Statistics Cannot Answer This Quickly
Distinguishing bad luck from a genuine drop in edge on the numbers alone requires a sample large enough to be confident that recent results are inconsistent with the old distribution. For a strategy with a modest edge, that sample is far larger than the drawdown you are currently sitting in.
This is worth stating plainly because it removes a false hope. The equity curve will not tell you. Any threshold you pick, some number of consecutive losses or some percentage decline, will fire regularly on healthy strategies and slowly on broken ones. The information you need is not in the results, so no amount of staring at them will produce it.
Check the Trades, Not the Curve
It is in the trades. Pull the losing ones and ask, one at a time, whether each was the trade the method is supposed to take. Was the range within the width the rules allow. Was the entry taken at the specified trigger rather than a few moments early. Was the stop where the plan put it. Was the trade exited by the rule or by hand.
A run of losses on trades that were all correctly identified and correctly executed is a normal losing run, whatever it feels like from the inside. A run of losses where a third of the entries were early, or the stop was moved on several, or two of the setups would not have qualified on a careful reading, is not evidence about the strategy at all. It is evidence about the execution, and the strategy has simply not been tested during that period.
Look for a Changed Condition, Not a Changed Number
If the trades were clean, the next question is whether the environment has shifted in a way the method depends on. Breakout strategies are exposed to one specific condition, namely that movement tends to continue after a level gives way. When markets enter a period of persistent mean reversion, that condition is absent and every clean break fails on its merits.
This is checkable independently of your own results, which is what makes it valuable. Look at the sessions themselves. Are ranges resolving and extending, or is price breaking and returning inside repeatedly, across many instruments and many days. A market wide change visible in the price action is a much better explanation than a strategy that mysteriously stopped working, and it implies a different response: wait rather than rebuild.
Execution Drift Masquerading as Strategy Failure
The most common cause of a strategy that has apparently stopped working is that it is no longer the strategy being traded. Drift is gradual and nearly invisible from the inside. A slightly earlier entry to avoid slippage, a slightly wider stop to avoid the noise, a session skipped because the setup looked poor, and after a few months the thing being executed is not the thing that was tested.
The defence is a written specification detailed enough that another person could grade your trades against it without asking you questions. Without one, the question of whether the system is broken cannot even be posed properly, because there is no fixed object to compare the recent behaviour against.
A Procedure Rather Than a Verdict
A reasonable process runs in a fixed order. First confirm the trades were taken to specification. If they were not, the problem is execution and the verdict on the strategy is suspended until a clean sample exists. If they were, look for a market condition that removes the method's premise, and if one is present, reduce size or pause and wait rather than modify anything. Only when the trades are clean and the environment looks normal does genuine edge decay deserve serious weight, and even then it deserves a larger sample before a rebuild.
What this procedure mostly buys is time spent on the right question. Most drawdowns investigated this way turn out to be execution problems or ordinary variance, and both of those have better responses available than redesigning a method in the middle of a losing run, which is the one response almost everybody reaches for first.