Most player projection tools give you one number: "LeBron will score 25.3 points tonight." That's it. No context, no confidence level, no sense of how likely other outcomes are.
We do it differently. For every projected starter in tonight's NBA games, we generate a complete probability distribution across four stat categories. You get the expected value, sure — but you also get the full picture of what's likely, what's possible, and how confident the model actually is.
Four Stats, Full Distributions
We project four categories for each player. Each one gets its own probability distribution chart.
Points (PTS)
The core scoring projection. Accounts for the player's recent scoring output, opponent perimeter and interior defense, and expected game pace.
Rebounds (REB)
Total rebounds — offensive and defensive combined. Sensitive to opponent rebounding rate, pace, and whether key teammates are in the lineup.
Assists (AST)
Reflects the player's passing role. Heavily influenced by usage rate, teammates' ability to convert, and whether the team plays through the ball handler.
Three-Pointers Made (3PM)
Expected made threes. Depends on the player's recent shooting volume and accuracy, plus how many open looks the opponent typically allows from deep.
What Goes Into the Models
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 — rolling performance windows that capture current form, not just season averages
- Opponent context — defensive efficiency, pace of play, how they defend the specific position
- Home / away splits — some players perform notably different on the road
- Playing time and role — expected minutes, starter status, and usage patterns within the team's system
- Injury context — when key teammates are out, usage and opportunity redistribute across the remaining players
Why distributions, not just averages?
An expected value of 23.5 points tells you the weighted average. It says nothing about how spread out the outcomes are. Our models output the full probability distribution — the likelihood of every possible outcome — so you can judge the uncertainty yourself rather than trusting a single number.
Reading the PDF Chart
The PDF (Probability Density Function) is a bar chart. Each bar shows the probability of a specific outcome — for example, the probability of scoring exactly 22 points.
The tallest bar is the single most likely outcome. The dashed vertical line labeled E[X] marks the expected value (the weighted average of all outcomes).
Pay attention to the shape. A narrow, concentrated distribution means the model is relatively confident. A wide, flat distribution means there's real uncertainty — the player could reasonably land anywhere across a broad range.
For a deeper walkthrough on reading probability charts, see our guide to probability distributions.
Reading the CDF Chart
The CDF (Cumulative Distribution Function) answers a different question: "What's the probability of going over X?"
It's a line that starts near 100% on the left (almost certain to score more than 0) and drops toward 0% on the right. The steeper the drop, the more concentrated the distribution is around that range.
Practical example: say a sportsbook sets a player's points line at 22.5. Find 22.5 on the x-axis of the CDF chart and read the corresponding percentage. If it says 58%, the model gives a 58% chance of going over and 42% of going under. Simple as that.
The tooltip on each data point shows both the over and under probability, so you don't need to do any math.
Why Distributions Beat Single Numbers
Two players can both have an expected 23.5 points. But one might have a tight distribution concentrated between 20 and 27, while the other has meaningful probability mass stretching from 12 to 38. Same expected value, completely different risk profiles.
The distribution tells you things the expected value can't: how much variance there is, where the probability mass sits relative to a given threshold, and whether a line is tight or has clear lean to one side. That's the information you need to make better decisions.
Models are analytical tools, not crystal balls. NBA games have real randomness — a hot shooting night, foul trouble, a blowout leading to reduced minutes. The distributions reflect this uncertainty honestly. Use them as one input alongside your own judgment.
See Tonight's Player Projections
Updated daily before tip-off. Every projected starter gets full expandable PDF and CDF charts across all four stat categories.