Research
Findings, with
the script attached.
Every question we wrote down before testing it, the verdict it came back with, and the committed script that produces it. The ones that found nothing are published with the rest.
How far our forecasts miss, minus how far the market's do (Brier score) · above zero, ours miss by more
+0.0142
The bar is the range the true figure plausibly sits in, given how much luck could move it (interval).
The interval excludes zero, only just: its low end is +0.0007.
95% interval 0.0007 to 0.0263.
Append-only
4 entries, dated 2026-09-10.
Nothing is edited or removed. A decision that is later reversed stays here, because why something was tried is worth as much as how it turned out.
What the terms in these entries mean
- pp
- percentage points.
- edge
- a real, repeatable advantage over the betting price.
- shrinkage
- pulling each player's figure toward the average of similar players, more so when there is little data.
Retention keys on ingestion time
quotes.tsis when a price existed; a 14-day window on it deleted an 806-credit backfill on arrival. Every time-based policy now keys oningest_ts, stamped by the writer.The 13-18pp edge was positive by construction
gross_edge = |model - market|because the side is chosen after the fact. A noise model scores over 10pp on it. The ledger now records signed edge and reports the unselected population alongside.Shrinkage compression is correct, not a bug
Model means sit 5-9% below each player's own prior mean and within 1% of the held-out actual. The player's own mean is the optimistic number.
Polling tiers key on kickoff, not
close_tsKalshi closes a prop at game end and Polymarket at kickoff, so one field meant two instants and mid-game Kalshi ran at 60s while the same claim ran at 10s.
Decisions logged since the newest entry here: coming once the decision log kept beside the figures is exported.
How this is computed
Listed newest first. A reversal is nested under what it reverses only where the entry names it; inferring the link from the wording would be a guess presented as a record. 0 of 4 entries name an entry they reverse.