People often assume sports betting AI models are built from simple box score stats things like points per game, three point percentage or field goal attempts. In reality our NBA models process thousands of opponent and situation adjusted data points every night.
We are not modeling numbers in isolation.
We are modeling basketball in context.
The inputs below are just a sample of the more advanced metrics that feed into our NBA projections. They sit on top of a much larger engine that includes coverage level tracking, lineup specific unit ratings and market calibration layers but these are the building blocks users can see, understand and in some cases tweak.
That means you can see exactly what is driving your edge instead of betting blind.
If you want to test these same metrics on tonight’s slate, you can do it in the Rithmm app with a free trial.
1. Opponent Adjusted Three Point Tendency and Efficiency
We do not just look at how many threes a player or team attempts. We adjust for who they are shooting against and when they are shooting.
Opponent Adjusted Three Point Tendency & Efficiency measures how often a player shoots from beyond the arc and how efficiently they make those shots, after accounting for
Opponent defensive strength at the three point line
Quality of closeouts and switching schemes
Game state context such as lead, close game or blowout
This helps separate real three point skill from volume against weak defenses or garbage time situations.
Inside Rithmm you can see how this metric shifts your implied lines when you emphasize matchup‑sensitive shooting over raw volume.
2. Opponent Adjusted Two Point Tendency and Efficiency
Mid range and interior scoring are even more sensitive to matchup and scheme.
Opponent Adjusted Two Point Tendency & Efficiency tracks how often a player or team attacks inside the arc and how efficient they are, recalibrated for
Rim protecting ability and help defense
Defensive drop coverage versus aggressive closeouts
Score and time context
This keeps the model from treating a high percentage two point night against a soft interior as a sustainable trend.
If you experiment with this metric in the app, you can isolate teams that consistently beat the spread when they attack specific defensive flaws rather than just shooting volume.
3. Opponent Adjusted Free Throw Generation and Efficiency
Getting to the line is a skill but it is also shaped by how aggressive the defense is and how often fouls are called.
Opponent Adjusted Free Throw Generation & Efficiency combines
A player’s ability to draw contact
Opponent foul tendency and defensive intensity
Game state context such as foul trouble or late game tight scores
That way we avoid overrating players who only get to the line against soft defenses or in runs of loose officiating.
In the Rithmm model builder you can see how ramping up this metric changes your probabilities on close‑line totals and late‑game props.
4. Defense and Game Situation Adjusted Tempo Rating
Not all fast paced games are created equal.
Defense and Game Situation Adjusted Tempo Rating measures how fast a team plays and how often they transition, after adjusting for
Opponent defensive tendencies push it or slow it down
Score and time of possession
Personnel and lineup sets
This helps distinguish between teams that want to play fast and thrive versus those forced into a higher pace against a run and gun opponent.
If you build a model that leans into tempo‑sensitive metrics you can spot overpriced totals and underpriced unders more consistently.
5. Opponent Adjusted Player Efficiency Offense EPA
Expected points added EPA is a core lens for how we value offensive possessions.
Opponent Adjusted Player Efficiency Offense EPA estimates how many points a player adds per possession, conditional on
Lineup context and role within the offense
Opponent defensive strength and scheme
Game situation early, close or late
This metric surfaces players who consistently create value in tougher matchups not just against weaker defenses.
In the app you can see how shifting EPA‑heavy weights changes your implied lines on moneylines and player props compared to raw scoring stats.
6. Opponent Adjusted Turnover and Assist Tendency
Ball security and playmaking matter as much as scoring.
Opponent Adjusted Turnover & Assist Tendency captures how often a player holds onto the ball and moves it for teammates, after adjusting for
Defensive pressure intensity
Opponent ball hawking tendencies
Game state constraints such as clock, score and defensive focus
This stops us from treating a high turnover night against an elite defense as a permanent flaw in decision making.
If you experiment with this metric you can build models that react to defensive pressure and matchup‑driven turnover spikes instead of chasing noise.
7. Opponent Adjusted Defensive EPA Threes
Raw opponent three point percentage is noisy and unreliable on its own.
Opponent Adjusted Defensive EPA Threes estimates how well a defense actually prevents efficient three point scoring, factoring in
Quality of shooters they face
Shot type breakdown catch and shoot versus pull ups
Game state and defensive game plan changes
This isolates true defensive ability at the arc from random variance.
Inside Rithmm you can see how weighting this metric affects your over under and team‑total probabilities especially in marquee three‑point matchups.
8. Opponent Adjusted Defensive EPA Twos
Interior defense is just as matchup driven.
Opponent Adjusted Defensive EPA Twos measures how well a defense limits efficient scoring inside the arc, conditional on
Rim protecting ability and help rotations
Defensive scheme and pick and roll coverage
Offensive personnel and pace
By contextualizing this metric we avoid overrating defenses that look good only because they face softer interior attacks.
If you build a model that leans into two‑point defense you can spot mispriced team totals and prop lines that ignore true rim‑protection quality.
9. Injury Adjusted Minute Projections
Raw minutes are a poor predictor when rotations are in flux.
Injury Adjusted Minute Projections forecast how many minutes a player is likely to log based on
Injury reports and return timelines
Rest patterns and load management signals
Coach driven rotation changes and lineup adjustments
This lets the model adapt fast when a starter misses a game or a role player steps into a bigger role.
In the app you can see how shifting minute weights changes your probabilities on player props and team totals after roster news drops.
10. Travel and Rest Adjusted Fatigue Index
Back to backs, late traveling and short rest windows all impact performance.
Travel and Rest Adjusted Fatigue Index bakes physical load into every projection, factoring in
Time between games
Travel distance and time zone changes
Recent workload and minutes accumulation
Research shows both travel related and cumulative fatigue can meaningfully reduce shooting efficiency and defensive impact so building this into the model helps us find value before the market adjusts.
If you emphasize fatigue‑indexed metrics you can build models that systematically lean into home‑under rested teams or road‑team‑fade situations that books often underprice.
Inside the Rithmm app you can see these same metrics updated in real time. You are not guessing what drives your model. You are choosing which signals matter most to you.
You can isolate specific metrics like three point defense EPA or rest adjusted minutes and see how changing their weight shifts your implied lines.
Every model you build can be backtested across multiple seasons with a few clicks so you can see whether your edge holds up over time.
This transparency is why users trust these projections instead of treating them as a black box.
Start your free trial to build your own NBA model with these metrics and test your theories on tonight’s slate.
Betting with a black box model means you never know what is actually driving your edge.
Rithmm lets you see exactly which metrics are moving your lines so you can trust your model not just chase percentages. Trials are designed so you can test specific theories such as teams with high fatigue indexes under performing across multiple samples.
You do not need to be a data scientist. You just need to be a bettor who wants to know what is behind every line.
If you want to see how opponent adjusted EPA, injury indexed minutes and fatigue focused metrics move your probabilities, start your free trial today.
Try Rithmm’s AI Model Builder
