I automated my leadership-rotation swing method (buy leaders coiled at support, 4–14% structural stops). The honest backtest numbers, caveats first.

Ran this on paper for a while, went live a few weeks ago. I'd rather have it torn apart now than find the holes with real money.

The idea: rank every name in the S&P 500 and Nasdaq-100 by relative strength, buy the leaders when they pull back to support, size by conviction, go to cash when nothing scores. Long only. No shorts, no inverse ETFs. The engine trades the scanner's published score and nothing else. No second model, no override.

Numbers, unflattering read first: the 2017 to mid-2026 backtest is +638% vs +282% for SPY, but through the end of 2025 it was basically even with SPY. The whole edge lives in leadership regimes; the rest of the time it just loses a little less. It did carry about a third less max drawdown. Survivorship-free (point-in-time S&P membership, delisted names included), next-open fills, costs modeled. No live track record yet and I'm not going to pretend otherwise.

What I actually want picked apart:

Letting one score be the only trigger. Fragile, or clean?

Long-only, giving up the entire short side.

The outperformance living in so few regimes.

Happy to share the writeup if the rules allow it, didn't want to link-drop into a design thread.