The most expensive thing in systematic trading is not a bad strategy. Bad strategies announce themselves fast and you cut them. The expensive thing is a good strategy that has quietly stopped being good while you are still trusting the test that approved it.
A backtest is a photograph. It tells you a strategy had an edge in a market that already happened. Markets do not hold still for the photograph. Volatility changes and the crowd changes. The inefficiency your strategy harvests gets arbitraged away or disappears outright. None of that sends you a notification. The numbers just start drifting, slowly enough that a human watching a P&L will explain it away for months. The manager who notices last pays the most.
So at my fund, no strategy gets to live on its resume. Before a strategy trades a dollar, we write down what healthy looks like for it. Not a feeling, an actual statistical range built from the strategy's own history: how its results should distribute, how deep a normal losing stretch can run, where normal ends and broken begins. Those boundaries get fixed before the strategy goes live and they do not move once it is running. A boundary you can renegotiate under pressure is not a boundary, it is a suggestion.
Then live results are checked against that range every single day. As long as the strategy stays inside its lane, it keeps its capital. The moment it crosses the line, it is suspended automatically. No meeting, no "let's give it one more week." And the switch only works in one direction. The system can shut a strategy off on its own, but only a human can turn it back on, and only after we understand what actually changed. Sometimes the diagnosis is that the environment shifted and the strategy was right to go quiet. Sometimes the edge is just gone. Either way, you find out by reading evidence, not by donating to the market while you wait for the answer.
The reason this has to be automatic is that humans are terrible at this exact job. When a strategy you built starts bleeding, everything in you wants to give it room. You remember the good years. You have a story for every losing week. The market does not care about your story, and the only defense I have found in twenty years of doing this is to make the decision before the emotion exists. Write the line in calm conditions, then let the line do the talking.
Verification goes further than watching results. On a regular cycle we rebuild each strategy's numbers from raw data, independently, hunting for the silent errors that creep into any live system. And when a result looks too clean, that is when we get the most suspicious. The review's whole job is to attack the result and see what survives. Results that can defend themselves get capital. Results that need defending from me do not.
People hear all this and assume it comes from doubt. It is the opposite. This is what conviction looks like once it has to be earned. I can size a strategy with real confidence because I know exactly what it has survived, and I know that the day it stops working, my system will tell me before my account statement does. The backtest gets a strategy hired. Re-earning its place every day is how it keeps the job.