One tab per stat. Points, rebounds, assists, threes, points+rebounds+assists, a Double-double tab and a First basket tab. The count on each tab is how many players the books price tonight — at a posted line, or (first basket, double-double) at a Yes price.

Each row. The posted line with the best price and book; ours — our calibrated chance to go over it — with the model's raw number beside it; the likely range (10th to 90th percentile); and the tier.

Tiers. strong, lean and pass follow how far the model sits from the book at the posted line; held means the play is withheld on the sheet (a 15-point gap, a threes over at 2.5 or higher where our tail runs hot, or a dated hold); not priced is a player under the 18-minute pricing gate or without enough history.

Teammate out tonight. The chip filters to players whose rotation teammate is ruled out, and each row then shows who is out, how much the model moved him (minutes and usage), and his raw split without that teammate. Sort by "Bump" to rank them.

Search and sort. Search by player or team; sort by posted line, our chance, edge, minutes or bump.

The headline tile. Our calibrated chance to clear the book's line, the best price, and beside it what the book says and what the model says. "We differ by" is the calibrated gap — the one that counts. Calibrated means shrunk toward the market by a factor fitted on the graded record (one part model to four parts book; assists about half and half), because the raw model runs hot at both ends.

Likely range. 10th to 90th percentile, the middle half and the median — the shape of the distribution. We never publish a projected stat line.

Minutes. Projected minutes with their spread, and the usage multiplier when a teammate is out (the redistribution from how the team has played without him).

Tabs on the card. One per stat plus Double-double; every panel below re-judges for the stat you pick.

First basket tab. Our chance he scores the game's first field goal (FanDuel and Caesars instead settle on the first points, where a free throw counts), the best Yes price with the implied chance and our edge, his team's chance of scoring first (from the two jumpers' tip-win rates), his share of the team's first baskets and first shots, and every start judged: ✓ scored first, ○ took the first shot, ✗ neither. Starters only. For fun — the market is sharp.

Ladder. Our chance over each line around the posted one, with the best price we found at that rung. Where a book prices the rung the number is calibrated against it, the same way the headline is; where none does, the rung is greyed and shows our raw model number. Books post these as alternate lines and only quote the over, so the market number behind a rung is recovered from the posted line's margin — good to a few points, not exact.

Check any line. Step the line, type one, or tap a rung: the hit tiles (last 5 / 10 / 15, this season, last season) and the game-log bars re-judge against it. "Back to the posted line" restores the book's number.

Game log. The last 15 games as bars against your line, each with the line that was posted that night where we have it. Chips switch to home, away, or vs tonight's opponent.

With and without teammates. For each rotation teammate: the player's average in games that teammate played vs games he missed (this season + last, inside the teammate's tenure), the number of games, and what the model does when he is out — a shrunk estimate, so it moves less than the raw split. OUT TONIGHT flags tonight's absences; greyed rows are former teammates and can never be out tonight.

Home and away. Hit rate at the chosen line, average and minutes at each venue; tonight's venue is outlined. A raw split — the minutes model already carries the home term.

What the opponent allows. What tonight's opponent gives up to the player's position group per stat over its last 20 games against the league; rank 1 allows the most.

Drivers and the shot chart. The inputs behind the number — minutes, start share, usage, pace, the line, rest — and the last 300 attempts by zone with league zone percentages.

Double-double. Our chance of any two of points, rebounds and assists reaching 10, from the three distributions given the minutes node, next to the book's Yes price when one is posted and the player's actual rate this season and last.

The tab. Hit tiles and the game log judged on any-two-at-10+ (10+ stats highlighted), a tick/cross strip with pts / reb / ast under each game, and the home/away and with/without splits as rates. No ladder and no line stepper.

Edge. Percentage points between our calibrated chance and the de-vigged market.

De-vigged market. The consensus two-way price with the book's margin removed.

Conditional on playing. Every chance assumes the player plays; books void and refund when he does not.

Pricing gate. At least 18 projected minutes and three games of history; below that the card shows the ranges but no chance against a line.

Usage. Field-goal attempts, free-throw trips and turnovers per minute — the share of the offence a player uses.