"F7 Fear&Greed contrarian skip -- placebo battery"

ETH gate60 window 2022-12-02..2026-09-18 (382 long days). OOS >= 2025-01-01. Placebo = random skip of the SAME count, N=3000. Rule: REJECT if OOS placebo p>=0.05.

#Placebo skip (does the F&G>=75 threshold beat random skipping?)

Segment	observed SR	n_skip	null mean	null p95	empirical p
OOS	1.39	9	2.45	3.26	0.988
FULL	1.31	104	1.74	2.60	0.794

#Threshold sensitivity (OOS Sharpe by F&G threshold)

threshold	skips (OOS)	OOS Sharpe
fng>=60	98	1.24
fng>=65	77	1.33
fng>=70	57	1.65
fng>=75 <- PRIMARY	9	1.39
fng>=80	0	1.12
fng>=85	0	1.12
fng>=90	0	1.12

#Dose-response (Spearman fng vs forward ret, OK long days)

rho = -0.039 (n=378); a real greed-veto needs a negative, monotone relation.

#Cross-coin (same gate, OOS placebo p)

coin	observed SR	n_skip	null mean	empirical p
BTC	0.79	17	1.01	0.722
SOL	0.64	11	-0.11	0.117

#Verdict: REJECT

ETH OOS placebo p=0.988 (observed 1.39 vs null 2.45); FULL p=0.794.
REJECT -> the F&G>=75 skip is indistinguishable from skipping random days; sentiment-skip stays parked (F8 DVOL untouched).