01 / IMPLEMENTATION & CONTRIBUTION
What the work involves
A standalone local study implements transcript-corpus ingestion, settlement-label processing, mention-time approximations, and hierarchical broadcaster/team effects, separate from the upstream Kalshi bot.
Technical depth
Beta-smoothed word baselines, hierarchical crew/team effects, empirical-Bayes shrinkage, knockout-regime changes, bounded corpus-logit tilts and leave-one-out evaluation with corpus-leakage adjustments.
The project family
prediction_market02 / RESULTS
What came out of it
The documented study found crew-settlement effects more informative than corpus tilts for words with settlement history. The result is a dated research finding, not an established trading edge.
03 / SUPPORTING EVIDENCE
Follow the source
Implementation notes, project records, and supporting artifacts.
Source context & project scope
Treat this as a dated July 2026 study. Captions and highlights have unequal coverage; first-mention timing is an approximation from price paths and excludes unsuitable contexts. No trading edge or out-of-sample deployment claim.
Separate original study, source coverage hierarchy, probabilistic model and ablation discussion.
SOURCE · 2026-09-17