NBA Player Projections
What the model expects from every rostered player over the week ahead, and the full distributions for tonight's games
Updated: 13 Jun 2026Pick your league scoring
Every number below is scored the way your league scores it. The choice is remembered in this browser.
Use it with your own squad
Save your roster once and the workspace turns this page into your week: the five to field tonight, the Captain call by a rule with no model in it, and the players in your squad with the thinnest week ahead. It is saved to your account, so the same squad answers you here, on your phone and inside Claude.
This week
Every player ranked by what the model expects him to score over the days ahead.
| # | Player | Team | Court | GamesG | Week totalTotal | Per gameAvg |
|---|
How the week numbers are made
Every scheduled game in the days ahead goes through the same thirteen models as tonight. The player's role in each of them comes from his own recent record rather than from a published lineup, since there is none yet, and the days are added together. That sum is exact for the average whatever the stats do together, which is why we publish it as an expectation and never as a range or a curve for the week.
Compare projected totals and scheduled games over the next seven days. Our historical study used some information that arrived during each week, so it does not establish how much a pickup chosen at the start of the week would improve your score. Read the methodology and validation status.
Tonight's games
Full distributions as JSON or CSV
| Player | Team | Opponent | H/A | Salary Cap | PTS | REB | AST | 3PM |
|---|
How Player Projections Work
We generate full probability distributions for twelve stat categories (points, rebounds, assists, 3-pointers made, steals, blocks, turnovers, field goals made and attempted, free throws made and attempted, and NBA Salary Cap points) for every rostered player in tonight's games, not only the expected starters. Each stat has its own dedicated model trained on thousands of player-game observations.
The models learn from a combination of player-level and game-level features: recent game logs, the player's own role history such as rolling minutes and how often he has started, opponent defensive context, pace of play, home and away splits, and how opportunity is redistributed when a team's usual rotation changes.
Bench players are trained on every roster row, including the games a player did not play, which are read as zero rather than skipped. Training only on the nights a reserve happened to appear would select on the outcome and flatter every bench projection.
The week ahead runs the same models over every game a player has scheduled in the days to come. There is no published lineup that far out, so his role comes from his own recent record rather than from a team sheet, and the days are added together. That sum is exact for the average, so we publish a weekly number as an expectation and never as a range or a curve.
The result is not a single number: it is a complete picture of what is likely, what is possible, and how confident the model actually is. The PDF chart shows the probability of each specific outcome. The CDF chart answers "what is the probability of going over X?" at any threshold.
Read the full methodology breakdown
Predixsport model outputs, built from published box scores and published lineups. Not affiliated with or endorsed by the NBA. We use no injury, fitness or availability data and we publish no probability that a player appears. A player or a representative can object to a named-player surface, and we can remove that player from these pages, our API and connected AI assistants while we review the request. Privacy policy.
Frequently Asked Questions
What stats do you project for NBA players?
Twelve of them: points, rebounds, assists, three-pointers made, steals, blocks, turnovers, field goals made and attempted, free throws made and attempted, and NBA Salary Cap points. Each projection includes a full probability distribution, not just a single expected value.
Why do some cells show a range instead of a number?
Because that is the honest way to publish them. Every one of the twelve models beats a naive rolling average on the shape of the outcome, which is what a distribution is scored on. Only five of them also beat it on the single-number error: points, rebounds, field goals made, field goals attempted and Salary Cap points. For the other seven we show the 10th to 90th percentile range and the full curve, and we do not print a bare average we cannot stand behind.
Which fantasy scoring systems are covered?
Four: ESPN points, Yahoo 9-category, Sleeper points and NBA Salary Cap. Pick one and every player gets that league's projected score. The score is the scoring formula applied to our component projections, which is exact for the average whatever the stats do together. Only NBA Salary Cap has a model of its own for the total, so it is the only preset where we also publish a range and a curve for the score itself.
Do you store my roster?
No. It lives in the browser you typed it into. Nothing about it reaches our database, so clearing your browser data clears the roster too. The roster tools sit in the NBA Fantasy workspace behind a free sign-in, which is what keeps them off this public page, and the sign-in tells us nothing about your squad.
How far ahead do the projections go?
Tonight's table is the day's slate, with the published starting fives in it. The week section covers every game a player has scheduled in the days ahead, projected without a lineup because none exists yet, and adds them up. A weekly number is an expectation: it has no range and no curve, because the sum of several games is not the distribution of one of them. When a batch carries no look-ahead rows yet, the section says so rather than showing an empty table.
Do you tell me who to captain, or which five to field?
The NBA Fantasy workspace adds a saved squad and the NBA Salary Cap Gameday card. It compares legal fives by projected total and tells you when only one fits. A separate captain rule backs your best form player of the ones with a game left in the Gameweek, on the first day he is in his team's published five, and it spends the chip on the last day your squad can still use it. That rule beat a week-start form pick by 1 to 5 official points a Gameweek over the 2025-26 season, measured with the official 2025-26 Gameweek calendar. Where we hold no official calendar for the season being played, the workspace works in the ISO week of the day's slate and says so, which is an approximation of the boundary and not an official Gameweek. Form is the average of a player's season-to-date and last-five realised fantasy points, counting recorded games he did not play as zero.
What does the availability label mean?
It is the published lineup status of the day and nothing else. In today's published five means the team has published its starting five and the player is in it. Not in the published five means the lineup is out and he is not in it. No lineup yet means the team has not published one at the time of the run. We do not publish injury news, a chance of playing, or any other availability claim.
How are the player projection models built?
Each stat has its own dedicated model trained on historical player game logs combined with contextual features like opponent defensive efficiency, game pace, home and away splits, and the player's own role history: his rolling minutes, how often he has started lately and how long since he last appeared. We ingest no injury news of any kind. The models output a complete probability distribution for each stat category.
How do I read the PDF chart?
The PDF (Probability Density Function) chart shows the probability of each specific outcome as a bar. The tallest bar is the most likely single outcome. The dashed E[X] line marks the expected value. A narrow distribution means high confidence; a wide one indicates more uncertainty. Full guide to probability charts.
How do I use the CDF chart for over/under analysis?
The CDF chart shows the probability of a player going over any given value. Find a threshold on the x-axis and read the corresponding percentage. For example, if the chart shows 62% at 20 points, there is a 62% chance the player scores more than 20. The complementary 38% is the under probability.
How often are player projections updated?
Once a day, before the games begin, using the published lineups of the day and the latest game logs, for every rostered player rather than the expected five. The look-ahead projections for the days after that are written by the same run.
Which players are included?
Every player on the projected roster of each team playing tonight, which is roughly 19 names per side rather than the starting five. The roster is the union of the players who appeared in the team's last ten box scores, the season roster sheet, and the published starting five when one exists. A same-day signing or call-up outside all three gets no projection, which is the honest answer rather than a guess.
Get NBA Player Projections Daily
Player projections updated before every game day. Join our Telegram channel and receive daily AI-driven predictions for NBA and all other sports, directly on your phone.
Join Telegram Channel