Alyssa Thomas Went Triple-Double and the Model Had to Chase

One game on the slate — PHX against CHI — and it delivered. The kind of night where the box score reads like someone left a cheat code running. Alyssa Thomas put up a triple-double with 19 points, 15 rebounds, and 12 assists in 38 minutes. That's not a stat line, that's a documentary. Meanwhile, Chicago kept it interesting behind an injury-reshuffled rotation that pushed Courtney Vandersloot and Kamilla Cardoso into expanded roles — and both of them cashed. Kahleah Copper dropped 31 points on 34 minutes of work, also with a minutes bump. DeWanna Bonner quietly went 14-8-4 with three steals and two blocks. This was a legitimate basketball game, played by people who are very good at basketball, and the DraftKings scoring reflected it all the way down the board.

Eighteen players graded. The mean absolute error came in at 8.1 DraftKings points. That's the honest one-number receipt — not great, not catastrophic, and mostly explained by a handful of players who went stratospheric while the model was reasonable. Let's go name by name through the ones that matter.

Where It Happened

• Alyssa Thomas ($11,500) — Projected 40.1, scored 59.3. That's a 19-point miss, and the miss is entirely fine, because projecting a triple-double at that volume for any player on any night is an exercise in humility. The model liked her. The model did not like her that much. Nobody did. When you get 19/15/12 in 38 minutes, you're in a different conversation than projections.
• Kamilla Cardoso ($9,200) — Projected 44.4, scored 43.3. An injury bumped her minutes, and the model had it. Off by one point. That's the kind of call you frame.
• Courtney Vandersloot ($7,400) — Projected 27.8, scored 43.3. An injury bumped her minutes too, and Vandersloot responded with 16/5/10 and two steals in 27 minutes. The model was directionally right — she was a value at that salary — but the ceiling she showed was above what we projected. At $7,400, 43.3 DK points is 5.9 points per $1K. That is a number worth underlining.
• Kahleah Copper ($9,300) — Projected 33.7, scored 41.0. An injury bumped her minutes, and 31 points on 34 minutes of run is how you cover that gap. The model was low, but it flagged the opportunity. That matters for construction.
• DeWanna Bonner ($8,900) — Projected 25.5, scored 40.5. No minutes bump here — this was just Bonner doing Bonner things at a level the model underestimated. Fourteen points, eight boards, four assists, three steals, two blocks. The defensive stats were the kicker. At $8,900 and 40.5 DK points, that's 4.5 per $1K, and a clean miss on our end.
• Natasha Cloud ($7,500) — Projected 28.8, scored 31.8. An injury bumped her minutes. Closest to the projection of the bumped players, and solid floor value.

Where We Missed

• Elizabeth Williams ($5,400) — Projected 31.0, scored 15.0. This is the one we own plainly. An injury bumped her minutes in our model, and she played only 15 minutes, posting 5 points and 6 rebounds. The bumped minutes didn't materialize the way the model anticipated. That happens with rotation volatility — it's the nature of injury-based projections — but a 16-point miss is a 16-point miss. We'll say it out loud because that's how trust gets built.

A few things worth carrying into the next slate.

First: the injury-bumped minutes signal was real on three of four players. Cardoso, Vandersloot, Copper, and Cloud all delivered. Williams was the exception, and exceptions happen — but the model's core thesis on injured-teammate minutes redistribution held up more often than not tonight, and it held up on the players who had the biggest ceilings.

Second: Alyssa Thomas at the top of the salary range ($11,500) still underscores something important about walk-forward models. The model was walk-forward validated, beating naive baselines out of sample, and it still projected a reasonable Thomas game because no walk-forward validation captures triple-double outliers as a systematic expectation. What the model does is identify players in high-usage positions with favorable contexts. Thomas was both. The outcome is in the tail of the distribution, and the honest answer is: you put players in favorable positions, and sometimes they go supernova.

Third: Bonner and Vandersloot are a reminder to watch defensive stat ceilings on players with the usage and athleticism to collect them. Bonner's three steals and two blocks were not out of character historically — they were just above what the single-game projection assigned. That's a modeling angle worth revisiting on the next Phoenix game.

The model earned its night on the cheap value plays. It left points on the table at the top. That's a scorecard we can live with.