Hoop DossierHow it works
Plain answers about what you see.
Data first
Every page starts with facts: lines from the sportsbooks, the official injury report, confirmed lineups, beat-writer and insider news, team and player numbers. Nothing on the default pages is our opinion.
Where it comes from
- Official NBA injury report Every report the league publishes on game days.
- Official NBA lineups Expected and confirmed starters and the inactive list.
- Beat writers and insiders Posts on X and Bluesky from national insiders and every team's beat reporters.
- ESPN and RotoWire Injury notes, transactions, scores, schedules and national TV.
- Sportsbooks Spreads, totals, moneylines and player props, from the sportsbook named on each line.
- Team podcasts We listen to every Locked On team show to track the hosts' calls (shown on the analyst page).
Brand new, and in testing
Hoop Dossier is brand new. The whole site is in testing, so I am changing things often: layouts, wording and features can move from one day to the next. The facts come from real box scores, lines and reports, but a page can still get something wrong or look off. If you see something wrong, let me know; it comes straight to me, and fixes show up on What's new.
Prediction tools are opt-in
Our predictions (projected scores, minutes and player stats, model leans) and the research tools marked EXPERIMENTAL (situational splits on player pages, the prop board) are works in progress. They can be wrong and have no proven record yet, so do not rely on them. They stay hidden until you switch on Prediction tools in the top bar. More
Splits and hit rates
A split compares a player's games in one situation (a teammate out, the second night of a back-to-back, a certain opponent) with all his other games. We label each one by how much to read into it: "tiny sample" (under 5 games), "only a few games" (5-9), "could be noise", "clear gap" (10+) or "big gap, solid sample" (15+): a description of past games, not a forecast. Hit rate is how often he went over tonight's line.
News speed
On game days we check most sources every few minutes (see source status). The games page, game previews, the Rundown and your watchlist update on their own while open; player and screener pages show when they were built. Each news item shows when it was posted.
Our track record
Behind the site runs a private lab that tests our models with simulated bets at real sportsbook prices, on games that had not happened yet. Nothing here is real money, and only the lab's chosen bets count (not exploratory ones). Every model output is Experimental. Nothing can be labelled Proven until it clears a pre-registered statistical test on its own exact market and model version: enough bets over enough separate game days, measured closing prices, closing-line value that stays positive after accounting for luck and repeated checking, and a profit after the sportsbook's cut, confirmed again on later games. Winning 50% is not enough: at −110 prices it takes about 52.4% to break even.
| Market | Stage | Settled bets | Game days | Closes measured | Won | Closing-line value | Return |
|---|---|---|---|---|---|---|---|
| Game lines (spread, total, moneyline) | EXPERIMENTAL | 0 | 0 | — | — | — | — |
| Player props (over/under) | EXPERIMENTAL | 0 | 0 | — | — | — | — |
| Milestones and specials | EXPERIMENTAL | 0 | 0 | — | — | — | — |
Game projections have a record of their own, not bets: every finished game with a pregame projection is scored (0 finished games had none). So far: 29 games scored, winners picked 20 of 29, average margin miss 10.2 pts (closing line 9.2 pts), win chance closer to the result than the closing odds in 14 of 29 · all versions: 45 games scored, winners picked 29 of 45, average margin miss 12.2 pts. Full record
Read more: how projections work · inside the Lab.
What the numbers mean
- Average, last 5, last 10
- His per-game numbers in games he played this season only. Before his first game the page says when they appear; last season never feeds this season's numbers.
- Hit rate / over rate
- Share of those games he finished over the line shown. A game landing exactly on a whole-number line is a push: it counts as neither over nor under and is listed separately.
- Split
- His average in games with one condition (a teammate out, a back-to-back) versus all his other games. A fact about past games, not a prediction. "Big gap, solid sample" (15+ games), "clear gap" (10+), "could be noise", "only a few games" (5-9) and "tiny sample" (under 5) describe how much past evidence there is, not how likely it is to repeat. Situations often overlap (road games and back-to-backs share nights), so several agreeing splits are not separate confirmations.
- Per 36 min
- His production scaled to 36 minutes, to compare players with different playing time.
- Win chance from sportsbook odds
- The moneyline converted to a probability with the bookmaker's margin removed. It is the market's view, not ours.
- Minutes (season averages)
- Minutes he actually played. Projected minutes are a model output and only appear with Prediction tools on, labelled Experimental.
- Line and quote age
- Lines are the sportsbook's current numbers with the book and when we last saw them. A line with an age or "no longer offered" may have moved.
What it is not
Nothing here is betting advice, and we never take bets.