NFL Board · Buckeye vs consensus

Coral BetBuilder

Coral BetBuilder prices, the ones placement would bet at, not the Buckeye board below. 98 sides priced against named books; showing the 3 at or above 0% vs Pinnacle or FanDuel. Prices 33m old. Coral copies DraftKings on most props, so vs DK sits near −5% by design: the signal is FanDuel disagreeing. vs Pin is Pinnacle named through FanDuel (context unless switched on; Pinnacle's prop ladders are short and low-limit). Page only: no cards, no placement. “unproven” = a market type no real ticket has used yet.

Coral BetBuilder game lines

Line shopping, not a play list. Coral BetBuilder moneyline / spread / total / team-total ladders (every alt rung, halves included) priced against Pinnacle and FanDuel from BetUnfair. 844 rungs priced; showing the 5 at or above 0%. “~” = the book does not quote that exact line, so its ladder is interpolated; the EV shown is the worst case after that family’s measured error. A book whose own limit is under $1,000 never counts. Snapshot 34m old — the Coral block is probably not running. Other leagues (NCAAF, CFL, MLB, NHL, WNBA) are in debug/heritage_game_plays.json.

Buckeye prop board

No props posted. Game-linked props only go up around gameday.

Where Buckeye’s number differs from the sharp cluster. Line-shopping and middling — not +EV.

No number disagreements.

Where the NFL data lives, and how to ask it a question. Nothing here is a price — it is documentation of files on the box.

What is captured

The tape is the asset. Every priced Buckeye side, every cycle, with the full per-book consensus beside it — append-only, never deduped. The page and the workbook are views of right now and are thrown away each cycle; the tape is the only record that Buckeye was on a number at a time. Buckeye's own price is irreplaceable — it exists only at scrape time. Results are backfillable from ESPN at any time, so a gap there costs nothing.

Ask it a question

How to read nflscore

It answers "is FanDuel sharper than Pinnacle in this market?" — not "is this profitable". Every book quoting a market is making a probability forecast of the same game, so the books are scored with log loss, a proper scoring rule: a book cannot score better by shading its number, so a lower score is a real claim about its forecast.

Why not ROI: measured on this repo's own log, ROI runs t=0.92 where CLV runs t=5.39 — it needs roughly 88× the sample, because it multiplies a probability question by a payoff. The comparison here is paired: same game, same market, same line, same outcome, so everything the two books share cancels and only the forecast gap is left. And it needs no bets placed.

Before believing a number

⚠ One NFL week is one slate. 14 games is not a finding. Give it ~4 gamedays.
⚠ Reading ten buckets and keeping the winner is how noise becomes a rule. The ELO Settings Optimizer produced confident chips over 3,725 bets that all evaporated under an honest bar. With ~10 buckets here, use |t| > 2.8, not 2.
⚠ Sharper means a better anchor, not a beatable book. Whether a Buckeye price is +EV against it is the separate question this page's Plays tab already answers.