NM1/STOP/DM1 on the broad universe (judged vs frozen v1). IND99 on the 99-verse (judged vs recomputed B99; v1 shown for context).
| Variant | IS Sharpe | IS maxDD | OOS Sharpe | OOS maxDD | OOS ret | turn/yr |
|---|---|---|---|---|---|---|
| NM1 intermediate 12-7 | 0.89 | -53.6% | 1.79 | -23.4% | +203.8% | 9.3 |
judge: NM1 intermediate 12-7 vs frozen baseline v1 (overlap 5030d)
| leg | base Sharpe | cand Sharpe | base maxDD | cand maxDD | | IS | 0.88 | 0.89 | -54.7% | -53.6% | | OOS | 1.74 | 1.79 | -28.5% | -23.4% | IS bootstrap p=0.244 | IS p<0.05: NO | OOS Sharpe >= base: OK | OOS DD within +5pp: OK verdict: FAIL — move on
| DM1 crash-hedge | 0.91 | -44.6% | 1.74 | -28.5% | +221.9% | 8.2 |
judge: DM1 crash-hedge vs frozen baseline v1 (overlap 5030d)
| leg | base Sharpe | cand Sharpe | base maxDD | cand maxDD | | IS | 0.88 | 0.91 | -54.7% | -44.6% | | OOS | 1.74 | 1.74 | -28.5% | -28.5% | IS bootstrap p=0.590 | IS p<0.05: NO | OOS Sharpe >= base: OK | OOS DD within +5pp: OK verdict: FAIL — move on
| STOP -15% engine | 0.90 | -48.4% | 1.73 | -27.1% | +196.9% | 9.3 |
judge: STOP -15% engine vs frozen baseline v1 (overlap 5030d)
| leg | base Sharpe | cand Sharpe | base maxDD | cand maxDD | | IS | 0.88 | 0.90 | -54.7% | -48.4% | | OOS | 1.74 | 1.73 | -28.5% | -27.1% | IS bootstrap p=0.799 | IS p<0.05: NO | OOS Sharpe >= base: NO | OOS DD within +5pp: OK verdict: FAIL — move on
| B99 BASE (99-verse, ref) | 1.05 | -48.8% | 1.64 | -18.4% | +170.0% | 8.3 | | IND99 industry momentum | 0.83 | -51.5% | 1.10 | -17.6% | +66.6% | 10.0 |
IND99 vs B99 head-to-head (overlap 5030d)
IS p=0.972 | B99 OOS 1.64/-18.4% | IND99 OOS 1.10/-17.6% | verdict: FAIL
Verdicts (2026-09-09)
ADOPT: nothing. Scoreboard: 33 judged variants, 0 adopted.
- NM1 intermediate 12-7 — FAIL, nearest miss of the program. Beats BASE on all three numbers (IS 0.89, OOS 1.79, DD -23.4%) but p=0.244. The Novy-Marx direction is right; the magnitude isn't significant. Watchlist: the single best challenger for any future fresh-OOS round.
- DM1 crash-hedge — FAIL, insurance again. IS 0.91 with DD -44.6% (10pp cut), OOS identical (never fired post-2022 — no 2yr-bear+rebound occurred). Same family as S3/credit-gate/T6: protection unneeded in this window. p=0.590.
- STOP -15% — FAIL. Whipsaw eats the protection (turnover +0.8/yr, OOS tied, IS n.s.). Stops don't help momentum here; the rebalance IS the risk control.
- IND99 — FAIL decisively (p=0.972 against, OOS 1.10 vs 1.64). Moskowitz-Grinblatt does not replicate long-only on this universe/era. Holding whole winning sectors dilutes into mediocrity.
Literature mapping (why the papers don't transfer)
- Barroso-Santa-Clara vol-scaling doubles Sharpe with leverage (scale UP in calm). Under this book's 1.0x cap only the cut survives → insurance, never alpha. Our V7/V8/T6 confirm exactly that half.
- Blitz residual momentum uses 36m FF betas; our 126d EW-market beta sits near 1 (V2==V1 twice), so there is nothing to orthogonalize here.
- M&G industry momentum is long-short over 20 industries; its long leg alone, on 11 GICS sectors, loses to stock selection here.
- Skipped with reason: profitability tilt (fundamentals start 2021-10), 36m-beta resid (warmup eats the GFC), factor momentum (heavy build).