Strip everything off. No stops, no targets, no exits. Just the raw event measured over a few fixed horizons.
If it can’t beat a baseline on its own, walk away. Not because the strategy can’t be made profitable, but because whatever edge you find will be coming from your risk management rather than your idea.
The trap here is common. Someone announces a new signal, you look at the backtest, and they’ve accidentally recreated trend following.
When the filter becomes the signal
Here’s the part I find genuinely interesting.
Say your raw signal is noise, you run it through an optimizer, and it spits out filters that make the curve look great. At what point is the filter the signal?
If relative volume is what’s actually doing the work, maybe relative volume should be the starting point, and finding the best entry inside that universe is the real job.
You can test this. Apply the same filter to a few adjacent entries. If they all perform about the same, your entry was never the edge.
The claim I don’t buy
You’ll hear well-known traders say you can make any signal profitable with good risk management. Take a random entry, manage it well, print money.
Maybe, at enormous scale, gross of costs. But why would you? You’d be six months into a drawdown with no idea whether anything was ever there.