Adaptive Trend Finder, honestly backtested — trends in noise, and an edge the payout eats.
I rebuilt it (best-Pearson-R² channel) and backtested properly: causal, out-of-sample, permutation null.
Synthetics: ~50%, below the ~52% break-even → negative expectancy on every horizon.
Baskets: a genuinely REAL mean-reversion (perm p→0.000, 52–54%) — but the payout (55–62% BE) eats it.
Structure ≠ profit. Method + code: github.com/nexusfinancial-dev/deriv-house-edge. Not financial advice.