ForbiddenPicks

Methodology

How every number on Forbidden Picks is made, where the data comes from, and how results are graded. No proprietary code, but no hand-waving either.

1. Start from the market

Sportsbooks set sharper prices than pick'em apps and move first when news breaks. For each player and stat, the site collects the prices from up to 7 sportsbooks and exchanges and removes each book's margin (the vig) to estimate the market's fair chance. A price only counts when at least two books agree; a single outlier is treated as a likely stale line, not an edge.

2. Add a projection where we have the stats

For NFL players, a stats model (currently v2.1.0) projects the stat from recent form, opportunity, matchup, lineup and game environment, and blends that with the market's view, weighted by how solid the inputs are (data confidence). Other sports use the market alone, plus recent-game hit rates where available.

3. Compare with what the pick'em app pays

A pick'em app pays the same multiplier for every standard pick, so each pick has a break-even chance (about 54% at PrizePicks and Underdog for most entries). A line where the fair chance is above break-even has positive expected value. Entry payouts follow each app's published rules, including Flex tiers, DNPs, ties and reversions.

4. Rate it and size it

  • Strong Value / Value: a clear edge over break-even with enough market agreement.
  • Lean: a small edge that may not justify a bet.
  • Pass: no edge, or negative EV.

Stakes are a capped fraction of your bankroll (about a seventh to a third of Kelly, set by your risk profile), inside your daily and loss limits. Sizes never increase after a loss.

5. Record and grade every prediction

Model predictions are saved before kickoff with their price, projection, probability, data confidence and model version. Before the game, only the closing price is updated; at kickoff they're frozen. After the game, the official stat grades it once. Nothing is deleted, including losses. See the track record.

Closing line value (CLV) compares the saved price with the last price before the game at the same line. Calibration checks whether picks rated, say, 60% win about 60% of the time.

Data sources

  • Sportsbook odds and pick'em lines: The Odds API, a licensed odds aggregator (DraftKings, FanDuel, BetMGM, Caesars, theScore Bet, Kalshi, Novig, PrizePicks, Underdog, DraftKings Pick6). Not directly from the sportsbooks.
  • Live scores and box scores: ESPN's public scoreboards and box scores, with The Odds API's scores as a backup.
  • Historical NFL stats: the open nflverse dataset.
  • NBA player game logs and photos: ESPN's public player data.
  • Injuries: not a separate feed yet. Injury news reaches the site through the sportsbooks' line moves.

Lines are refreshed on a schedule (more often as games get close), not streamed, and every page shows how old its data is. Sources can be delayed or wrong; always confirm the line in the app before you play.

Limits

Probabilities are estimates, not guarantees. Small samples swing a lot, so the track record labels them. Sports betting and fantasy sports involve financial risk: only play where it's legal and with money you can afford to lose.