Saturday, May 16, 2026

All 3 models agree on under, and the data is extraordinary: Reaves has 0.0 steals in his last 5 straight games. That's not a trend, that's a pattern of complete disengagement from the steal category. Whether it's a change in defensive assignment, effort allocation, or scheme, he simply isn't generating steals right now. Against OKC's ball-secure offense led by SGA (one of the best ball-handlers in the league), the chances of a steal spike are minimal. The 1.2 line almost requires him to get 2 steals to go over, which would be a massive outlier from his current form.

All 3 models agree on over, and the data is compelling: Hachimura has averaged 4.2 3PM over his last 5 games with 4 of 5 going over, and last 3 straight at 4.3 3PM. He's nearly doubling this line. This isn't just a hot streak — it represents a genuine role change where Hachimura is launching more threes in LAL's current offensive scheme. Against OKC, whose perimeter defense focuses heavily on shutting down primary ball-handlers (SGA matchups), role players like Hachimura often get open looks. The 4-day rest for LAL means fresh legs and likely sharp shooting.

SGA has been under 1.5 steals in all 5 of his last 5 games and 9 of his last 10, averaging just 0.8 steals. GPT's OVER 70% call is clearly an error — perhaps their model is anchoring on SGA's career reputation as a steal artist rather than the current data. The 0.8 average means he'd need to nearly double his recent output. Against LAL's relatively careful ball-handling guards (Reaves, Smart), this is a strong under play.

All 3 models agree, and the data backs it: L5 average is 19.6 PTS, L10 is 19.9, and 7 of 10 recent games went under. Reaves has been consistently below this line. OKC has the league's best defense and will likely assign Dort or Caruso to shadow Reaves, both elite perimeter defenders. Even with LAL's 4-day rest, the matchup difficulty here is severe. The 22.5 line appears to be set based on Reaves' peak performance rather than his current production level.

Every single data point screams under: L5 avg 6.2, L10 avg 6.7, season avg 6.1, home avg 5.9, and 8 of 10 recent games under. LeBron has been less engaged on the glass, possibly conserving energy for offense. OKC's dominant frontcourt of Holmgren and Hartenstein will dominate the boards and limit LeBron's easy rebound opportunities. The 7.5 line is set based on a younger LeBron's rebounding norms, not his current output. Two of three models agree on under.

Hartenstein has been under this line in 9 of his last 10 games, averaging just 1.9 STL+BLK. His season average of 1.8 is even lower. The 2.5 line requires a significant outlier performance — essentially needing 3+ combined steals and blocks. While Haiku went over, the overwhelming data trend makes this a strong under play. LAL's offense doesn't generate many block-prone shots, and their guards are reasonably secure with the ball.

All 3 models agree on over, and the numbers support it: L5 average is 2.0 STL with 3 of 5 over, and L10 average is 2.4 STL with 6 of 10 over. Smart is one of the NBA's most active defenders and has been generating steals at an elite rate recently. Against OKC, Smart will be guarding secondary ball-handlers and will be active in passing lanes. His defensive intensity typically elevates in marquee matchups. The 1.5 line should be cleared by his current form.

Holmgren has averaged 20.8 PTS over his last 5 games, clearing this line by 2+ points consistently. His L10 average of 19.4 also clears comfortably. GPT's under call at 80% seems to be based on an expectation of regression that isn't supported by the data. Against LAL's interior defense (Ayton, Hayes), Holmgren has the shooting versatility to score from multiple areas. In a competitive game where OKC will need his scoring, expect him to maintain his recent production levels.

Mitchell's season average is only 1.1 3PM, home average is 1.1, and road average is 1.0. Every single split is well below the 1.5 line. GPT's under at 80% is the correct call. Deepseek and Haiku going over at 60-62% seem to be hoping for variance rather than reading the data. Mitchell simply doesn't shoot enough threes or make them at a high enough rate to consistently clear 1.5. The line is set too high relative to his actual production.

Smart's L5 average is 9.6 PTS with 4 of 5 under. His season and home averages are both 9.3 — well below the 10.5 line. While his L10 average of 12.9 looks better, it's inflated by a couple of outlier scoring games. Against OKC's elite defense, Smart's scoring opportunities will be limited, and he's more likely to contribute through playmaking and defense. Deepseek's OVER 80% is overweighting the L10 average while ignoring the more recent and more relevant downward trend.