Position-weighted top-trader longs do not clear validation

Verdict: NO_EDGE · Asset: BTCUSDT sum_toptrader_long_short_ratio · Horizon: next 8h · Sample: chronological 60/20/20

Leftover after ttratio killed count top-trader (corr 0.99 vs global). This cycle uses the notional / position-weighted top-trader ratio. Frozen discovery P90_sumtop=1.43883.

What happened

Q1 val FAIL: HIGH−MID = −3.0 bp, t=−0.95 (|t|<2). HIGH-rate drifted badly: disc 10% → val 68% / OOS 54% — the frozen tail is not a stable event stream.

Q2: corr(sumtop, global)=0.157 — not the same costume as ttratio. Incremental γ looks interesting on paper (t_γ=−2.04) but is irrelevant: Q1 failed, so Q3 was not opened.

Trading

Do not replace global account-count crowded-long with position-weighted top-trader. Distinct series, null gate.

Final Verdict

NO_EDGE

The demoted WATCH object remains global count_long_short_ratio. Neither top-trader count nor top-trader position unlocks a cleaner book.