Friday, May 15, 2026

Two models at 85% OVER with season average of 18.1 PTS, which is 25% above the line. Home average is even higher at 18.5. Mobley would need an historically bad shooting night to miss 14.5 — he's scored 15+ in the vast majority of games this season. Detroit doesn't have the interior defense to limit him significantly, and Cleveland's home crowd plus playoff intensity should keep him engaged. Haiku's UNDER at 72% is almost certainly wrong here, possibly confusing Mobley with a more inconsistent scorer.

Two models at 80-85% OVER. DiVincenzo's 30.1 MPG and 12.1 PPG need to go somewhere, and Dosunmu is the primary beneficiary. He should see 30+ minutes and take on a significant portion of DiVincenzo's shot attempts. The Spurs' elite offense will keep this game competitive, ensuring Dosunmu stays on the court in meaningful minutes. Haiku's UNDER likely assumes Dosunmu stays in a limited bench role, but the injury cascade clearly elevates him to starter-level usage.

Both deepseek (76%) and GPT-4.1-mini (75%) agree on UNDER. Cunningham faces Cleveland's elite defense at home — Allen and Mobley limit rim attacks while the perimeter D is solid. A PR of 32.5 requires roughly 27 points and 6 rebounds, which is Cunningham's ceiling, not his median outcome on the road. Cleveland's defensive structure specifically targets primary ball-handlers, and Cunningham's efficiency typically drops in these matchups. Haiku's OVER at 71% seems to project a best-case scenario rather than expected value.

Strong 2-1 UNDER consensus (GPT at 80%, haiku at 78%). Wembanyama's three-point shooting is high-volume but inconsistent — he can launch 6+ threes but only converts at roughly 33-35%. Making 3+ threes requires both volume and efficiency to align, which is unreliable. Against Minnesota's length and Gobert's influence on shot selection (pushing Wemby to the perimeter where he's less efficient), UNDER is well-supported. Deepseek's OVER at 60% is the weakest conviction among all three models.

Both deepseek (80%) and GPT (70%) agree OVER. DiVincenzo's absence creates a massive opportunity — Dosunmu absorbs both scoring and playmaking duties. If he scores 14+ and dishes 4+ assists (both reasonable in 30+ minutes), he clears 17.5 PA comfortably. The game script should be competitive given SAS's elite offense, meaning Dosunmu won't be benched in garbage time. Haiku's UNDER may not fully account for the magnitude of DiVincenzo's minutes redistribution.

Deepseek at 78% UNDER leads the consensus. Cunningham would need an elite game — roughly 27 points and 10 assists — to clear 36.5 PA on the road against Cleveland's defense. CLE limits opponent point guard production through their switching scheme and rim protection. Even Cunningham's best games rarely combine both high scoring AND high assists at these levels simultaneously. Haiku's 75% OVER is likely anchoring to Cunningham's season highs rather than median outcomes in tough road environments.

GPT at 85% UNDER and haiku at 78% UNDER create strong consensus. Champagnie is a role player whose three-point volume is limited to 3-4 attempts per game. Making 2+ threes requires a 50%+ conversion rate on any given night, which is unreliable even for good shooters. Against Minnesota's remaining perimeter defenders (McDaniels is elite), Champagnie's looks won't be clean. Deepseek's 65% OVER seems based on the 'he's a shooter' label rather than volume-adjusted probability.

GPT-4.1-mini at 85% OVER correctly identifies the key factor: DiVincenzo's 12.1 PPG absence creates a massive usage vacuum. Edwards is the primary scorer who absorbs the largest share of redistributed shots. Additionally, the Spurs bring an elite offense that will keep Minnesota in a competitive game script, requiring Edwards to score at high volume throughout. Against SAS's perimeter defense (which loses some pressure with DiVincenzo's absence being on MIN's side, not theirs), Edwards should attack aggressively. Deepseek and haiku's UNDER at 65-72% may be using season averages without properly adjusting for the teammate absence.

GPT at 85% OVER and haiku at 75% OVER form strong consensus. Gobert is one of the league's most consistent rebounders, and at home against a SAS team that plays through the paint with Wembanyama, there will be plenty of missed shots around the rim generating rebound opportunities. Wembanyama's shot-blocking creates long rebounds that favor the defense, and Gobert positions himself perfectly for these. Deepseek's 85% UNDER is a red flag — it's extremely unusual for Gobert to miss 8.5 rebounds at home. This looks like a deepseek model error.

Two models OVER (deepseek 70%, GPT 75%). With DiVincenzo out, Dosunmu becomes a primary ball-handler alongside Edwards. His playmaking ability is well-documented from his Chicago days, and in expanded minutes (30+), generating 4+ assists is very achievable. The Spurs' defensive rotations with Wembanyama as rim protector create kick-out opportunities that Dosunmu can facilitate. Haiku's UNDER may not fully model the role change that comes with DiVincenzo's absence.