**Grade: A** — Called the winner, score was close. **Result:** GSW 110 - ATL 126 (Winner: ATL ✓) **Prediction:** GSW 109 - ATL 119 Score error: 1/7 | Total: pred 228, actual 236 (+8) **Standout Players:** Gui Santos (GSW): 2p actual vs 12.7p pred (+10.7) Dyson Daniels (ATL): 28p actual vs 11.8p pred (-16.2) Mouhamed Gueye (ATL): 16p actual vs 3.8p pred (-12.2) Brandin Podziemski (GSW): 5p actual vs 15.1p pred (+10.1) **Value Bet Results:** ✗ TOTAL UNDER 229.5 — LOST ✗ TOTAL UNDER 229.5 — LOST ✓ MONEYLINE ATL -455 — WON ✓ MONEYLINE ATL -480 — WON

**Prediction:** GSW 106 - ATL 115 (ATL by 9, total 221) **Confidence:** 94% **GSW** 29-25 (#8) | Net: 2.0 | Off: 112.2 | Def: 110.2 | Rest: B2B **ATL** 26-29 (#9) | Net: -1.2 | Off: 111.7 | Def: 112.9 | Rest: B2B Projected pace: 103.7 **Injuries:** GSW: Jimmy Butler III (Out), Stephen Curry (Out), Seth Curry (Day-To-Day), Al Horford (Day-To-Day), Malevy Leons (Day-To-Day), Moses Moody (Out), Kristaps Porziņģis (Day-To-Day), Quinten Post (Day-To-Day) (impact: -10.0 pts) ATL: Jonathan Kuminga (Day-To-Day) (impact: -0.8 pts) **Key Players:** GSW: De'Anthony Melton 15.5p/4.1r/2.9a | Kristaps Porziņģis 13.9p/4.2r/1.7a | Brandin Podziemski 13.6p/4.7r/3.7a ATL: Nickeil Alexander-Walker 22.2p/2.6r/2.8a | Jalen Johnson 19.9p/10.0r/7.0a | CJ McCollum 16.0p/2.0r/3.5a **Value Bets:** TOTAL UNDER 229.5 (strong, EV $0.360/dollar) TOTAL UNDER 229.5 (strong, EV $0.338/dollar) MONEYLINE ATL -455 (moderate, EV $0.141/dollar) **Applicable Insights:** - [model] PROP TRAP: Season-avg-based unders on role players from depleted/tanking teams are unreliable. When star players are out - [model] PROP CALIBRATION: 'Strong edge' plays (biggest gap between season avg and book line) are the WORST tier, not the best. M - [model] PROP SIGNAL NOISE: EDGE, REGRESSION, and FADE signal types show extreme day-to-day variance and should not be relied on - [team] GSW without Curry+Butler is unpredictable for props. Role players (Melton, Podziemski, Horford) can explode beyond seaso - [model] USAGE BOOST OVERCORRECTION (Mar 6): First night with usage adjustment. Props 25/64 (39%). Big-boost (>25%) plays 11/28 ( - [model] BLOWOUT MINUTES KILL PROPS: NYK@DEN 142-103 — Murray played 17:36 (adj avg 33.9, actual 12p). Brunson 33:48 but only 9p - [model] 3PM PROPS CATASTROPHIC (Mar 6): 3/15 (20%) hit rate. Usage boost inflates 3PM predictions but shooting is high-variance - [model] GAME MARGIN vs PROP CORRELATION (Mar 6): POR@HOU 7pt game: 7/12 props (58%). LAC@SAS 4pt: 3/9 (33%). IND@LAL 11pt: 3/6 ( - [model] OVER BIAS EXTREME (Mar 6): 62/65 props were OVERs (95%). Usage adjustment inflates all averages upward, turning everythi - [model] TOTAL UNDER SYSTEMATIC MISS (Mar 6): Under bets went 2/7. Avg predicted total 218, avg actual 235 — off by +17. Depleted - [model] GAME PREDICTIONS STRONG (Mar 6): 6/7 winners (86%), value bets 22/32 (69%). NYK ML best bet hit massively (142-103). Mod - [model] Mar 6 calibration changes validated: props improved from 39% (Mar 6) to 54.8% (Mar 7) after reducing PROP_USAGE_CONFIDEN - [model] Close game prop thesis confirmed 2 nights running: LAC@MEM (pred margin 3, actual margin 3) went 6/8 (75%) on props. Com - [model] Big usage boost (>25%) still underperforms at 36% (4/11) on Mar 7. Keyonte George had +60% boost (adj avg 38.5, scored 2 - [model] REGRESSION_OVER signal hit only 38% (3/8) on Mar 7 vs SEASON_EDGE at 61% (14/23). Two nights of data now show regression - [model] Mar 7 game predictions 4/6 (67%). Two F grades: ORL@MIN predicted MIN by 9, ORL won by 27 (36pt error). BKN@DET predicte - [model] Blowout discount working correctly: BKN@DET (pred margin 18) had 0 prop plays filtered out entirely. Mar 6 NYK@DEN (pred