How we build this board
No model, no picks. Here is exactly what the number on the board is, where the market-implied probability comes from, and how we count a player's history against it.
The line is the book's, not ours
Every line on this board is a real number a sportsbook posted for tonight's game. We don't adjust it, model it, or rank it. It shows up here exactly as the book has it, alongside the best price we found across the books we track.
The Filters panel's Best line at selector scopes the board to rows where one specific book is the one currently posting that best price — useful for answering “where is DraftKings best tonight” without opening every row. It is arithmetic on the books' own numbers (whichever book pays out the most on a given side), not a ranking of the books against each other, and it never reorders rows or implies which side to take.
Your book is a different control: pick one sportsbook and the whole board re-prices to that book's own line and price on every row it offers, instead of whichever book happens to be sharpest per row. Hit rates keep measuring against the board's own line history — they are not recomputed per book — so a row where your book's line differs from that history is marked rather than shown as if nothing changed.
De-vig: from a price to a probability
A posted price carries the book's built-in margin, the vig. A two-sided market like Strikeouts Over/Under prices both sides above 50% once you convert the odds to probability. That extra slice is the hold. To strip it out, we convert each side to an implied probability, then scale the pair so they sum to 100%.
We do this across every book quoting the line and take the median as the market-implied probability shown on the board. It's the market's own read on the outcome, with the vig removed. Not our estimate, not a prediction. That's why we label it plainly as “Mkt implied.”
Two ways we count hit-rate
The board tracks how often a player has actually cleared a line, in two different ways. They answer slightly different questions, so we keep them as separate columns rather than blending them into one number.
- Vs tonight's line. Take tonight's posted number and apply it retroactively across the player's last 10 games. Shows how the player has done against the exact threshold being offered right now.
- Vs the line that night. For each past game, use the line that was actually posted that night, not tonight's number. Lines move as a player's role or matchup changes, so this is the truer read on how a player has performed against whatever the market asked of him at the time.
Both cuts come from graded box scores, not projections. A game where a player didn't play, or where the market voided the line, is excluded rather than counted as a loss. Same rule for a push: a tie is left out of both counts rather than counted as a win or a loss, and it never uses up one of the 10 games in the window.
What this board isn't
There's no tier here, no expected-value percentage, no pick. We used to run a graded prediction model against these markets. The honest result: the market is efficient and we don't have a repeatable edge to sell you. So we dropped the prediction framing and kept the part that's actually useful, the line and the history behind it.