Social attention adds no predictive power for intraday move size beyond price volatility

Instrumetriq Research Note. Using 380,932 coin-sessions from the Instrumetriq Tier 3 archive (213 days, 2025-12-15 to 2026-07-15, 278 crypto assets), we test whether social attention predicts intraday move size beyond what a plain price volatility reading already gives. It does not, and that holds however attention is measured: distinct posting authors, engagement, reach, and the Engagement Coefficient the Instrumetriq extension shows as a coin's chatter level. Price volatility alone ranks large movers at AUC 0.77-0.83, every attention measure ranks them at 0.50-0.54, and adding any of them to volatility changes the out-of-sample AUC by 0.000. A positive control that peeks at the answer moves that number by +0.01 to +0.14, so the test detects added value when it exists. Attention precedes larger moves only because attention and volatility rise together. Refines the attention-volatility note, whose lift result is reproduced here and stands. Reproduction code and free weekly samples are mirrored across GitHub, Hugging Face, Kaggle, and Zenodo. Full archive and current-month updates: instrumetriq.com.