NBA Model Audit: +3.30 Units Across 21 Tracked Window Picks

Across 14 games from October 25 to October 28, 2026, the model portfolio logged a 12-7-2 overall record, delivering +3.30 net units with a 63.2% win rate across 21 total tracked picks. Player props drove the net gain with a 12-6-2 mark (+4.30 units) via the xgb_prop_v4 model, while the meta_v2 model's single spread pick dropped -1.00 unit. Four days of NBA results represent a small sample size that should be evaluated plainly rather than taken as proof of model performance.

Breakdown of Tracked Market Performance

Market	Model	Picks	W-L-P	Units
Prop	xgb_prop_v4	20	12-6-2	+4.3
Spread	meta_v2	1	0-1-0	-1.0
Total	Portfolio	21	12-7-2	+3.3

Across all 21 plays, the average closing line value (CLV) finished at -3.0. High-confidence prop projections anchored the window's profitability, led by successful under evaluations across multiple statistical categories.

Prop Results and Key Model Divergences

The xgb_prop_v4 model saw several high-confidence under picks convert comfortably against market prices:

• Anthony Black: Under 4.5 assists (-106 american price, 0.82 confidence score) covered with an actual value of 1.0 assist (+0.94 units).
• Jordan Goodwin: Under 2.5 assists (-192 price, 0.83 confidence score) covered with an actual value of 2.0 assists (+0.52 units).
• De'Andre Hunter: Under 1.5 assists (+134 price, 0.81 confidence score) covered with 1.0 assist in a 123-112 Memphis win over Sacramento (+1.34 units).
• Desmond Bane: Under 3.5 assists (-104 price, 0.81 confidence score) covered on Oct 25 with 2.0 assists (+0.96 units), though his Oct 28 under 3.5 assists pick (-125 price, 0.75 confidence score) lost with an actual value of 4.0 assists.
• Isaiah Jackson: Under 6.5 rebounds (-110 price, 0.68 confidence score) covered with an actual value of 5.0 rebounds (+0.91 units).
• DeAndre Jordan: Under 4.5 points (-148 price, 0.54 confidence score) covered with an actual value of 3.0 points (+0.68 units).

Over selections also generated positive returns. Ziaire Williams went over his 8.5 points line (+100 price, 0.69 confidence score) by logging 11.0 points (+1.00 unit) in Portland's 136-129 victory over Los Angeles. In that same game, Austin Reaves cleared his 23.5 points line (-107 price, 0.53 confidence score) with 26.0 points (+0.93 units).

On the losing side, Ty Jerome missed his under 6.5 assists projection (+102 price, 0.78 confidence score) by recording 8.0 assists (-1.00 unit). Paul Reed went over his 2.5 assists under line (-150 price, 0.66 confidence score) with 3.0 assists (-1.00 unit), while Ajay Mitchell exceeded his 1.5 three-pointers under line (-155 price, 0.62 confidence score) with an actual value of 2.0 threes (-1.00 unit). Isaiah Collier fell just short of his over 6.5 points line (-130 price, 0.54 confidence score) with 6.0 points (-1.00 unit), and Ryan Kalkbrenner went over his under 5.5 points line (+104 price, 0.57 confidence score) with 6.0 points (-1.00 unit).

Two prop picks resulted in pushes (0.0 units): Kawhi Leonard under 3.5 assists (-130 price, 0.79 confidence score) and Keegan Murray under 1.5 assists (-108 price, 0.66 confidence score), both finishing with actual values of -1.0.

Spread Outcome and Standout Performances

The meta_v2 model's lone spread selection came on October 26, backing the home Utah Jazz at -1.50 against the Memphis Grizzlies (-104 price, 1.00 confidence score). Memphis won 110-105 (actual margin -5), producing a -1.00 unit result and a CLV of -3.00.

Outside the betting board, several notable individual box scores occurred during the window. Luka Dončić recorded the highest game score of the period (43.6), producing 52 points, 8 rebounds, and 10 assists on October 25. Jayson Tatum delivered 40 points and 8 assists on October 27 (35.0 game score), Brandon Miller scored 39 points on October 26 (29.5 game score), and Karl-Anthony Towns grabbed 20 rebounds alongside 27 points on October 27 (19.4 game score).