Last night is the day before this morning. A day with no NBA games is skipped back to the last day that had some, at most three days; past that the section says when the last games were. The night's box scores are read at 5 AM Eastern.
Each side's totals from the men it started that day: ESPN's box scores and ESPN's lineup for the night. A bench man's game does not count, as on ESPN. A percentage is made over attempted; in turnovers, fewer wins.
Box Out's week report's projected categories, every category counted once, as of yesterday morning and as of this morning: this morning's is the figure at the top of the page. The move is what the night did and anything else that changed overnight.
Each side's categories as posted: ESPN's live score for the matchup on a live day, with tonight's games in it as they are played; on a replayed one, each side's started lines on this period's days before it, from the stored box scores.
Box Out's seating measure: each count over this league's weekly spread for its category, turnovers against, percentages left out. It is the measure that decides who sits on a full day. His usual is the same measure on his per-game line this season, five games at least, else last season's. The bar runs from nothing to 1.25; its tick is his usual. The nights that counted come first, best first; then the bench; last, men added this morning, whose night was not the team's.
A big night is at least 1.5 times his usual and at least 0.3 above it; a quiet night at most half his usual and at least 0.3 below it; neither without a usual above zero. On 2025-26's 18,716 nights, big fired on 11.8% and quiet on 7.6%. They are the only two labels.
Free agents who played last night and can play. The ten best nights among them are each priced by Box Out's week search for your team with the wire narrowed to him, as the news alerts ask it: his best add, its drop, and what it does this week. At most five are listed, those worth a look first; a man under your line is listed with his figure all the same.
last-night: matchup (the night, the week before and after, the score so far), men and best_counted, wire's listed. Built when it is asked, and not kept.
This week, by category
Select a category to see what moves it. The tick on each bar is 50%.
Each category's chance this week, head to head against this period's opponent, from the stored week report: first its quick read of the chances, then the full report. It is built again after every read of ESPN's data. The other side is our estimate of their side, on the roster they hold. The bar is the chance, with a tick at half: a figure at the tick is a coin flip.
Likely yours is a chance at or over 65%, likely theirs at or under 35%, and in the balance is the rest: where the page splits the league's categories for the eye, not a claim the model makes. The Matchup page draws the same split out of the same code, and opens each category's totals.
Select one and the panel beside it shows its chance, the score so far and the projected finish, and the moves the week report considered that change it, most helpful first. Set aside this week marks a category you are not chasing: the week's moves are put in order again with that category's change this week counting nothing, and its changes are drawn grey on every move. Chase counts it half again in that order and Ease off counts it half. A mark changes the order of the moves and which are worth a look. It changes no number: every figure counts each category once, and the chances never change. Kept for this team and this matchup period, so the digest and the connector order the same way; next week starts with nothing set aside. The season-long build is the Build page's.
Preview move redraws the rows with the move's own figures from the week report: where each category stands now, where the move would take it, and the difference in whole points. A move that follows another in the report's plan is previewed with the one before it, because that is how the report worked it out. Reset puts the rows back. Nothing is worked out here, and nothing touches ESPN.
Beside each category, the most one add from the wire raises its chance this week, in whole points, and how many free agents raise it 5 points or more ("1 at +5"). Box Out's week search, every category counted once: each free agent with a legal add is judged once, by his own best add, so each carries his own drop. A best add of +3 or more is named; below that the row says nothing reaches above it ("nothing above +2"). A category at 97% or more, or 3% or less, is decided and not searched. On a bye, or with no adds left this period, the week is not read.
Select a category and its panel puts that best add in full: the add and the drop, the category's chance before and after, what it does to every category counted once, each category it moves by a whole point, and Preview move, which redraws the rows as any move's preview does. Where nothing reaches +3 and nobody reaches +5, the panel says the wire can't move the category this week: that is for you to weigh against Set aside, and it sets nothing.
For each category, the wire's best and fifth-best free agent by that category's part of Box Out's per-game measure, each with where he would rank among the men the league's teams hold, injured reserve included. Field-goal and free-throw percentage are not ranked, since the measure leaves a percentage out, and nor are turnovers, where the men who barely play would rank best.
The chances: pickups/glance, answered from the stored report. The totals and every move's change: pickups/stream. The wire beside each category and in its panel: wire-supply's week and season, kept for the day as the week report is.
The moves this day's week report found worth a look, in its own order. The first leads: the add, the drop, and what the pair changes this week, category by category. When the report finds none, the page says so and lists the closest, with their numbers.
The large figure is the change in categories you are expected to win this week. Longer term has the rest: the change in an ordinary week over the weeks left, and the two together, which is the figure the report ranks by. On a build, or with a mark for the week, the order and the label are theirs and every figure still counts each category once.
Games left is the man's games in the days left in the period; a status is ESPN's word for him that morning; the bid is the ladder's 75th percentile for a man like him in this league. Each category's change is in whole points of win chance, the two figures a row shows subtracted.
Opens each move's working in the drawer, the bid and its ladder among it. Every move the report considered is on the Matchup page, under More. Nothing here touches ESPN: you make the move there.
pickups/stream's plan (the moves it lists as worth a look), then moves, in the route's own order; adds_left, adds_budget, faab_remaining for the budget.
It describes the lineup as set on ESPN, because that is the one that counts: whether it fills every starting place that can be filled tonight, and if not, how many more it could. The two figures under it are the report's own: the places ESPN's lineup fills with a man who plays tonight, of the places this roster can fill; and the places nobody on the roster can fill.
A line under them is a fact that asks for a decision today, drawn only when a report already read here says so: a place in the lineup as set on ESPN tonight that is empty, or holds a man with no game, while the bench could fill it; a man ESPN has out or in doubt tonight; and the days left in the period with a place going empty that the wire could fill, as one line. Before the draft it is the draft.
Check lineup opens the lineup as Box Out would set it: the day's own report's seating, the most this roster can fill, and a proposal, not what is set. Where the lineup as set on ESPN has another man in a place, or nobody, the place carries a grey DIFFERS tag and names who ESPN has there, and the caption says once whether the two agree. A difference that fills as many places is not raised outside the table: there is nothing to gain by changing it. Nothing here sets a lineup on ESPN.
A man with a game and no place in the lineup as Box Out would set it is a row under the places, tagged BENCH; the Matchup page says why he is left out. The rest of the bench, men who cannot start this day and injured reserve, is folded under the table.
What he did that day, once the box score is read from ESPN: minutes, field goals and free throws made/attempted, then three-pointers, points, rebounds, assists, steals, blocks and turnovers. A dash in every column is a game not played yet, or not read yet; a zero is kept, because 0 for 6 is not a night off. Tip-off is in your own time. On a period that is over it is the day's lines as played.
A grey tag beside a man is ESPN's word for him that morning (Q questionable, D doubtful, DTD day to day, OUT, IR injured reserve); a man with no tag is active. OUT · PLAYED is a man ESPN had out that morning whose box score shows him playing.
The lineup the other team set for the day, as last read from ESPN: a fact about what they set, not a projection, in the league's order of places. It is folded because nothing in it can be acted on.
Yours: today's lineup and starts (as Box Out would set it), actual, actual_starts and actual_known (as set on ESPN), lined up by slot in pages.js (lineupPairs); benched, idle and injured_reserve for the bench; fix for the places ESPN gives away. Theirs: lineups. A place going empty: pickups/stream's empty_days. The table is pages.js's, the Matchup page's own.
In the league
The league's newest published recap, when it has one: its headline and first paragraph, as the Newsroom page prints them. Written by a language model from the league's stored facts, checked by code against them, and approved by the league's owner.
The league's last adds, claims, drops, trades and status changes, and posts on the league's board, over the seven days to the day on the page, newest first. Each sentence is the one the morning email sends; a line about your team carries the orange edge, one about your opponent a grey one, and a post that mentions you an orange dot. Times are in your own zone, named in the foot, and each line sits under its date there. Your own moves, graded, are on Move history.
changes: each sentence is the API's own. The recap: newsroom's published.
Your place
The league's unit is categories, and its standings' order is the category record, as ESPN's is: the share of categories won (a tie counts half), then categories won, then fewest lost. The rows are put in that order here, from the league's records as last read from ESPN; three of them: yours, the one above it and the one below. The whole table is the Standings page's.
The category record as it stands, as last read from ESPN: a past day asked for still shows the latest. The projected finish (the projection's mean simulated record) and the playoff odds are from the season's projection, the one the Standings page's Projected view draws; that projection orders its seasons and seeds their playoffs by the same category rule as the standings, so the two places are one order.
The records: standings, the latest stored whatever ?today= asks. The finish and the odds: projected.