Studies run against this site's own archive of captured sportsbook and prediction-market prices. Nothing here is a tip sheet. It is what the stored prices say when you go and count them.
It is written for bettors deciding where to shop, and for anyone who wants to check the claim rather than take it. Each study states its question, its method, its sample and its caveats, and links the exact data it used. That data is free: the whole price archive is downloadable as CSV from the historical odds data page, so every figure in this section can be rebuilt from source by someone who does not trust it.
The live example of the last rule: a finished study of closing line value on player props is deliberately not in this list. The prop prices it rests on were invalidated by the pricing failure described in the exchange price autopsy below, and it is being re-run on clean data before it goes anywhere.
Being far from the market and being repeatably wrong are close to unrelated: across 128 published cells the correlation is r = +0.03.
30 days · 3,292 games · 82,704 scored quotes · 31 play dates
Read the study →Nineteen percent of quoted exchange prop sides were derived from the opposite side of the book, which is not a cheaper seat but one already occupied.
Found 2026-08-12 · killed the same day · signal v1 retired publicly
Read the study →The claim that prompted it did not hold, and measuring a book against its own closing price turns out to be close to useless.
30 days · 520 games with a usable close · MLB, Tennis, UFC, WNBA, MLS
Read the study →Why the site publishes this at all. Everything here is measured from the same price history that drives the line shop, so the studies are a check on the site as much as on the books. When that check comes back badly, as it did in August 2026, the account of it goes in this section too.
