Volume mostly tags the vol regime, not the next move

Verdict: WATCH · Asset/TF: BTCUSDT 1h (primary) · Sample: chronological 60/20/20

Hypothesis: joint extreme volume + price shock (bucket C) carries more next-bar information than a pure price shock (A), a pure volume spike (B), or neither (D).

What happened

On validation, C beat A on forward magnitude (C/A ≈ 1.13–1.20). B looked like A. D was well above unconditional |fwd r| (volume extremes are not pure noise for size). Sign: no increment — C ≈ A on direction.

On OOS 1h, raw C/A collapsed to ~1.03–1.07. After controlling for the event’s own |r|, the residual flipped null/negative. So the val “edge” was mostly selection: bigger contemporaneous shocks, not independent volume information.

Trading

Q3 mild fade on C events dies at 5–10 bps. Not tradeable.

Final Verdict

WATCH

Volume is mostly a vol-regime proxy. The joint extreme looks stronger on val via selection plus a weak residual that fails the primary OOS. Not promoted; not a clean NO_EDGE either (val magnitude gate opened). Revisit only with a cleaner volume definition (e.g. residualized vs recent |r|) or a different venue microstructure feed — not this kline volume alone.