Rankings powered by the Gaussian Hoops Star Model — a hybrid of Machine Learning and historical NBA comp analysis. The model combines a Logistic Regression (AUC 0.812, 5-fold CV) trained on 494 historical draft picks (2015–2024) with a KNN comp engine that matches each prospect to the five most similar historical draftees by playing style. A Mock Draft Prior adjusts probabilities for consensus top-20 prospects. Key inputs: scoring, playmaking, shooting efficiency, athleticism, age at draft, wingspan, and position-specific weights. Competition level is adjusted — NCAA Power 5 stats count more than Low-Major.

Star% is a probability estimate, not a guarantee. The model intentionally favors young, high-usage, athletic prospects — known predictors of NBA star outcomes. DCS v2 (Draft Ceiling Score) provides a secondary physical-tools signal.

Three-layer prediction pipeline: LR Model (40% weight) + KNN Historical Comps (60% weight) → Wingspan Boost → Mock Draft Prior → final Star%.

LayerWhat it does
LR ModelLogistic Regression on 22 features (stats + age + wingspan + position). AUC 0.812 (5-fold CV). Trained on 494 picks, 2015–2024.
KNN CompsCosine similarity on percentile stats → 5 nearest historical draftees → weighted outcome probabilities (Star/Rotation/Fringe/Bust).
Wingspan BoostPosition-adjusted wingspan delta adjusts Star% ±6–12%. Wingspan >3″ above avg: +12%. Below avg: −10%.
Mock Draft PriorConsensus mock rank boost: picks 1–5 +15%, 6–10 +8%, 11–20 +3%. Updates without retraining.
DCS v2Physical-tools ceiling score (PUI, STI, AC, PMI, SEU, RBI) with position-specific weights and age multipliers. Reference metric.
Filter:
2026 NBA Draft Rankings — loading…
Sorted by Star% (Model) ↓ — click any header to re-sort
# Player Pos Cls Age Team Tier Ht PUI STI AC PMI SEU RBI DCS v2 ⭐ Model