CourtVision uses a layered metric architecture where no single number captures basketball truth. Each metric measures a specific dimension of value, and composite projections blend them with context-aware weighting. Everything is designed to be explainable and transparent about uncertainty.
Season averages mask what's happening now. CourtVision uses three overlapping rolling windows to balance recency with stability:
Captures hot/cold streaks and immediate form changes. High signal, high noise.
Smooths single-game variance. Detects sustained performance shifts.
Baseline stabilizer. Separates real trends from small-sample flukes.
Stable estimate of overall player value using the strongest existing impact metrics.
PER, BPM, VORP, WS/48, plus/minus data, usage rate
Weighted composite: 30% BPM + 25% VORP-adjusted + 20% WS/48 + 15% PER + 10% on/off differential. Normalized to 0-100 scale.
Jokic (BIS 85): Elite across all impact metrics — highest BPM, top-3 VORP, exceptional on/off swing.
Captures current form with noise filtering — what is actually happening right now.
Last 5/10/15 game windows of PPG, FG%, plus/minus, usage trends
Rolling z-scores across 3 windows (5g, 10g, 15g) weighted 50/30/20. Streak classification via bayesian regression to filter noise.
Estimates whether a player actually improves team defense, beyond steals and blocks.
DBPM, steal/block rates, opponent FG% at rim, team DRTG on/off splits
40% DBPM + 25% on/off DRTG differential + 20% opponent shot quality impact + 15% steal+block volume adjusted for position.
Measures how hard a player's offensive role is — separating easy-efficiency from high-burden creation.
Usage rate, assist ratio, pull-up vs catch-and-shoot split, creation frequency, TS% under high usage
Usage-adjusted efficiency: TS% × usage-difficulty multiplier. Higher scores = maintaining efficiency at high creation burden.
Measures whether player value transfers across different contexts, roles, and lineups.
On/off with various lineups, teammate quality adjusted metrics, role versatility index
How much value persists when teammates change. Low variance across lineup contexts = high portability.
Estimates value created without the ball through spacing, movement, and ecosystem effects.
Off-ball scoring frequency, gravity metrics (opponent help rate), screen assists, hockey assists
Off-ball points + teammate efficiency lift when player is on-court without the ball. Adjusted for minutes and role.
Baseline team quality estimate adjusted for schedule, injuries, and recent form.
Win%, top-5 player BIS, roster depth (BIS of players 6-10), SOS adjustment
35% win-pct-adjusted + 30% top-player-composite + 20% roster-depth + 15% SOS-adjusted net rating.
Estimates how strong a team is right now based on rolling performance and context.
Last 5/10/15 game team record, point differential trend, home/away splits
Rolling team form: weighted game results with recency bias. Recent blowout wins count more than close losses.
Measures how well a team's lineup units function together across spacing, creation, and defense.
5-man lineup net ratings, minutes-weighted lineup combos, spacing metrics
Minutes-weighted average of top-8 lineup net ratings, penalized for over-reliance on one lineup.
Estimates whether a team's style and depth survive high-leverage playoff basketball.
Playoff experience, crunch-time metrics, half-court offense efficiency, defensive switchability
50% half-court offense rating + 25% defensive scheme versatility + 15% experience factor + 10% crunch-time performance.
Measures fragility if the top player is absent or limited.
Top player's share of team value, backup quality, injury history, clutch dependency
Gap between team performance with vs without best player, adjusted for backup quality. Higher = MORE risk.
Penalizes overlapping skill sets that reduce team fit and flexibility.
Player skill overlaps, positional redundancy, shot-type distribution
Penalizes teams where multiple high-usage players operate in same zones. Lower = less redundancy (better).
CourtVision generates win probabilities and projected scores for every game using a multi-factor model:
FiveThirtyEight-style power rating, updated after every game with margin-of-victory adjustment. Best single predictor of future wins.
Point-in-time ORTG/DRTG from last 15 games (weighted-recency). 70% rolling / 30% season blend prevents overfitting.
Team-specific home advantage derived from home vs away win rate differential. Ranges 1-6 pts (not flat 3.2).
Back-to-back teams penalized -4 pts. Rest advantage computed from days between games for each team.
Star players (BIS 75+) missing = -3.5 pts. Starters (BIS 55+) missing = -1.5 pts. Capped at -12 total.
TSC differential (0.15 pts/point) and LTFI momentum (0.08 pts/point) for roster quality and form.
Every metric includes a confidence score (0.0-1.0) that reflects data quality, sample size, and input availability.