Demo · Live scores
DEN79·OKC84Q3 4:12
NYK28·BOS32Q2 1:48
LAL119·DAL112FINAL
MIN61·GSW56Q3 9:22
DEN79·OKC84Q3 4:12
NYK28·BOS32Q2 1:48
LAL119·DAL112FINAL
MIN61·GSW56Q3 9:22

Award case lab

MVP Race

A refined four-layer race model: Basketball Value, Award Case modifiers, context/confidence signals, and structured analyst lenses in one transparent case workspace.

Refined Methodology

Each block below tells you what job a metric has. Core score inputs build the Basketball Value base; award modifiers shape the MVP ballot case; context signals explain the evidence; analyst lenses translate the data into basketball terms.

Basketball Value

Season-long on-court value from impact, efficiency, scoring load, playmaking load, team value, and availability.

Award Case

The main leaderboard rank. Starts with Basketball Value, then applies capped voter-facing modifiers.

Confidence

Coverage, sample stability, and signal agreement make uncertainty visible instead of hiding it.

Analyst Lenses

Role difficulty, scalability, game control, two-way pressure, and playoff translation standardize the qualitative layer.

MVP methodology reliability

methodology

Reliability n/aConfidence registry

Methodology

How the MVP case engine works

The v3 tracker separates season-long Basketball Value from voter-facing Award Case logic. The main rank is Award Case: Basketball Value plus capped modifiers for team framing, eligibility pressure, clutch, momentum, and signature games. Gravity, support burden, opponent splits, play-style translation, and coverage warnings remain labeled context signals unless their samples are stable enough to join a core pillar.

Box-First

Classic MVP index — production, efficiency, BPM/VORP/WS, team context.

Production 25% · Efficiency 20% · Impact 25% · Team 15% · Momentum 10% · Style 5%

Balanced

Default. Blends box totals with multi-metric impact consensus and clutch signal.

Prod 18 · Eff 15 · Box-Impact 15 · Impact Consensus 20 · Clutch 10 · Team 12 · Momentum 7 · Style 3

Impact-Consensus

Leans on EPM/LEBRON/RAPTOR/PIPM/DARKO consensus + clutch; tempers box-total reliance.

Impact Consensus 35 · Clutch 15 · Eff 15 · Team 15 · Prod 10 · Style 5 · Momentum 5

These legacy profiles are now sensitivity checks, not the main methodology. They help explain whether a candidate's case depends more on box totals, balanced weighting, or impact consensus.

Consensus over any one metric: when impact metrics disagree we surface the disagreement (σ) rather than picking a winner.
Confidence gating:clutch stats under ~100 possessions and opponent splits with < 4 games per bucket render dimmed with a warning icon, not hidden.
Attribution: every external metric is labeled with source and as-of date. Gravity continues to distinguish official NBA rows from CourtVue proxy values.