← all research
STUDY

"Long-Only Momentum — Realistic (Net) Backtest"

2026-08-26

Full delisting-inclusive _ADJ universe. Long-only top 10 names by trailing return, monthly rebalance, NO shorting. Realistic execution: next-open fills + turnover costs (flat grid 5/10/25/50 bps AND liquidity-tiered). IS/OOS 2016-01-01. Gross column uses next-open fills too (lower than the earlier same-day gross in momentum_spectrum).

Top 10 names (absolute; ~2% of 500-universe)

L(d)H(d)GrossNet@5Net@10Net@25Net@50TieredOOS Net@10 pOOS Tiered p
63211.401.371.341.271.141.270.0000.000
126211.401.381.371.311.221.310.0000.000
189211.321.301.291.251.171.250.0000.000
252211.251.241.231.191.131.190.0000.000
126631.301.291.281.251.201.250.0000.000
252631.121.121.111.091.061.090.0000.000

Production config detail (L=126, H=21, top 10)

Net equity curve (10bps, full universe)

Concentration check — true top-decile (top_k=50, ~10% of 501)

Verdict

If net Sharpe stays > 0 with OOS p < 0.05 at realistic (tiered / 25bps) costs, the edge is deployment-grade. The production sleeve runs on the 101-name liquid universe (lighter drag than this full-500 stress).