"testbot_expand4 — six new mechanisms vs frozen v1"

Same universe/split/costs/fill/liq-gate as expand2/3. Judged by testbot_baseline.judge.

#judge: T1 formation 189d vs frozen baseline v1 (overlap 5030d)

| leg | base Sharpe | cand Sharpe | base maxDD | cand maxDD | | IS | 0.88 | 0.81 | -54.7% | -55.6% | | OOS | 1.74 | 1.80 | -28.5% | -25.4% | IS bootstrap p=0.764 | IS p<0.05: NO | OOS Sharpe >= base: OK | OOS DD within +5pp: OK verdict: FAIL — move on

| T2 formation 12-1 | 0.78 | -55.6% | 1.81 | -29.5% | +265.4% | 6.4 |

#judge: T2 formation 12-1 vs frozen baseline v1 (overlap 5030d)

| leg | base Sharpe | cand Sharpe | base maxDD | cand maxDD | | IS | 0.88 | 0.78 | -54.7% | -55.6% | | OOS | 1.74 | 1.81 | -28.5% | -29.5% | IS bootstrap p=0.651 | IS p<0.05: NO | OOS Sharpe >= base: OK | OOS DD within +5pp: OK verdict: FAIL — move on

| T3 industry momentum | SKIPPED (sector-map coverage < 200) | | | | | | | T4 inv-vol weighting | 0.82 | -53.5% | 1.51 | -25.3% | +138.7% | 10.0 |

#judge: T4 inv-vol weighting vs frozen baseline v1 (overlap 5030d)

| leg | base Sharpe | cand Sharpe | base maxDD | cand maxDD | | IS | 0.88 | 0.82 | -54.7% | -53.5% | | OOS | 1.74 | 1.51 | -28.5% | -25.3% | IS bootstrap p=0.999 | IS p<0.05: NO | OOS Sharpe >= base: NO | OOS DD within +5pp: OK verdict: FAIL — move on

| T5 H63 | 0.87 | -60.9% | 1.56 | -30.8% | +199.2% | 4.7 |

#judge: T5 H63 vs frozen baseline v1 (overlap 5030d)

| leg | base Sharpe | cand Sharpe | base maxDD | cand maxDD | | IS | 0.88 | 0.87 | -54.7% | -60.9% | | OOS | 1.74 | 1.56 | -28.5% | -30.8% | IS bootstrap p=0.379 | IS p<0.05: NO | OOS Sharpe >= base: NO | OOS DD within +5pp: OK verdict: FAIL — move on

| T6 breadth gate | 0.71 | -24.1% | 1.74 | -7.8% | +36.8% | 2.2 |

#judge: T6 breadth gate vs frozen baseline v1 (overlap 5030d)

| leg | base Sharpe | cand Sharpe | base maxDD | cand maxDD | | IS | 0.88 | 0.71 | -54.7% | -24.1% | | OOS | 1.74 | 1.74 | -28.5% | -7.8% | IS bootstrap p=0.999 | IS p<0.05: NO | OOS Sharpe >= base: OK | OOS DD within +5pp: OK verdict: FAIL — move on

| T7 long-short+borrow | -0.67 | -95.0% | 0.25 | -31.1% | +7.0% | 8.5 |

#judge: T7 long-short+borrow vs frozen baseline v1 (overlap 5030d)

| leg | base Sharpe | cand Sharpe | base maxDD | cand maxDD | | IS | 0.88 | -0.67 | -54.7% | -95.0% | | OOS | 1.74 | 0.25 | -28.5% | -31.1% | IS bootstrap p=1.000 | IS p<0.05: NO | OOS Sharpe >= base: NO | OOS DD within +5pp: OK verdict: FAIL — move on

#Verdicts (2026-09-09)

ADOPT: nothing. Six ideas, six rejections (one skip). Baseline now unbeaten across 22 judged variants.

• T1/T2 formations — FAIL. 9-month and 12-1 both lose IS (p=0.76/0.65) with equal-or-worse OOS. The 6-month formation stands.
• T3 industry momentum — SKIPPED. sectors.json covers 101 names; the broad universe needs 200+. Untested, not rejected — needs a real GICS map.
• T4 inv-vol weighting — FAIL decisively (IS p=0.999 against, OOS 1.51 vs 1.74). Equal-weight stands for momentum picks.
• T5 H63 — FAIL. H21 stands; H42 remains the furthest tested fallback.
• T6 breadth gate — FAIL on the bar, notable insurance profile. OOS Sharpe tied (1.74) with DD crushed to -7.8% and turnover 2.2/yr — but IS pays (0.71, p=0.999 against). Same family as vol-target/credit-gate: protection that costs full-cycle Sharpe. Unadopted; threshold stays pre-registered (tuning it now would be threshold-gridding).
• T7 long-short — catastrophic FAIL (IS -0.67, maxDD -95%, OOS 0.25). Directional equity shorts destroy in this setup even with modeled borrow. Strong evidence for the long-only + beta-hedge-proxy-only stance. (This does NOT touch the C7 SPY beta-hedge, a different mechanism.)

#Methodology note

First T6 print was a vacuous PASS: the timing scale was applied before the fully-invested renormalization, which erased it (identical series, PASS on degenerate p=0.000). Caught by inspection, fixed (scale after renorm), re-ran: real T6 is the insurance profile above. Lesson re-learned: any PASS on an identical-returns print is guilty until proven innocent.