Boston Went Off, Leite Went Off, and the Model Had to Sweat

Three games, first tip at 7:00 PM ET, and the slate gave you exactly what a Thursday night in August should: a dominant big, a surprise point guard, and one team that made DFS lineups look very silly in the back half of the night.

A'ja Wilson was A'ja Wilson — 26 points, 13 rebounds, three assists, a steal, two blocks in 37 minutes for Las Vegas. That is not a box score, that is a checklist of things the rest of the league cannot stop. Jackie Young ran shotgun with 22 points and nine assists in 35 minutes, and the Aces did what the Aces do.

Out in Portland, Carla Leite decided 30 minutes was plenty of time to go for 27 points and 11 assists. Eleven. The point guard who most casual fans were still getting comfortable spelling delivered a night that would have looked at home in any All-Star game. In Indianapolis, Aliyah Boston put up a 21-12-2 line with three steals and two blocks — the kind of stat sheet that makes you wonder why we ever build lineups without her.

And then there was Toronto, where the box scores looked like they were submitted by a different team than the one we projected. More on that in a moment.

Over in Minnesota, Olivia Miles (23-8-5, two steals) and Napheesa Collier (21-8-6, one steal) anchored the Lynx in a slate that had plenty of fantasy points spread around — 34 minutes each, doing the work.

57 players graded. Mean absolute error: 7.0 DK points. That is the one-number receipt for the night — honest, unvarnished, here it is.

The Top Scorers

• A'ja Wilson (LV) — We projected 56.1, she finished 52.8. A 3.3-point miss on a player who logged 37 minutes after an injury bumped her role. Call it what it is: a near-hit on the best player in the league. The model liked her, Wilson delivered, the delta is noise.
• Aliyah Boston (IND) — We projected 38.4, she finished 50.0. An injury had bumped her minutes heading into Thursday, and Boston turned that opportunity into a 21-12 double-double with three steals and two blocks. An 11.6-point beat on our number. We'll take it.
• Carla Leite (POR, $9,000) — We projected 37.4, she finished 49.0. Eleven assists. The value case wrote itself at that salary, and she went out and validated it completely. An 11.6-point beat, same margin as Boston, from a player at a fraction of the price.
• Olivia Miles (MIN, $11,000) — We projected 39.3, she finished 43.5. A 4.2-point beat, clean and quiet. Miles was exactly who we thought she was.
• Jackie Young (LV) — We projected 37.4, she finished 42.0. Nine assists on 35 minutes. A 4.6-point beat. Young's usage in this offense is consistent and the model tracked it well.
• Napheesa Collier (MIN, $10,500) — We projected 46.7, she finished 41.0. A 5.7-point miss to the downside. An injury had bumped her projected role, she played 34 minutes, and the production didn't fully materialize. The minutes were there; the peripheral stats were quieter than expected.

Where We Hit

The best-call trio is worth sitting with because all three were driven by the injury-minutes edge — the thing the model is specifically designed to catch in real time.

• Aliyah Boston (IND) — 38.4 projected, 50.0 actual. An injury had bumped her minutes, and the model flagged the role expansion before lineups locked. The result: a 50-point night from a player whose increased workload we'd already priced in.
• Carla Leite (POR, $9,000) — 37.4 projected, 49.0 actual. Same story. An injury had bumped her minutes, and at $9,000 in salary, the value was real. 27 points and 11 assists later, it was the best points-per-$1K return on the slate.
• Julie Allemand (TOR, $7,300) — 28.8 projected, 39.3 actual. An injury had bumped her minutes on Toronto as well, and Allemand delivered 12 assists and 32 minutes. At $7,300, the floor was low enough and the ceiling, in an expanded role, was exactly what we said it was.

Where We Missed

Own it plainly — three projections that did not hold up.

• Marina Mabrey (TOR, $9,900) — We projected 40.5. She finished 13.0. Twenty-two minutes, 9 points. A 27.5-point miss. That is a bad one. An injury had bumped her projected minutes — our model read the role expansion and liked it — but Mabrey logged barely any time and the production evaporated. At $9,900 in salary, that's painful lineup real estate.
• Kiki Rice (TOR, $6,300) — We projected 28.3. She finished 6.0. Eighteen minutes, five points. A 22.3-point miss. Again, an injury had bumped her role on paper heading into the night, and again, Toronto's rotation did not cooperate with the model's read. Two misses in the same backcourt on the same night is a theme, not a coincidence.
• Natasha Howard (MIN, $9,600) — We projected 40.2. She finished 25.0. Twenty-five minutes, 8 points, 6 rebounds. A 15.2-point miss. An injury had bumped her minutes, and the role was real — she played — but the efficiency wasn't there. Eight points in 25 minutes at $9,600 is a lineup anchor that didn't anchor anything.

The injury-minutes edge is the real thing this model does. When an injury reshapes a roster and bumps a player's projected role, the model catches it — and Thursday night was a clean example of that working. Boston at 50, Leite at 49, Allemand at 39: all three were flagged because of role expansion, all three delivered. That is not luck. The model is walk-forward validated out of sample, which means it beats the naive baselines on data it has never seen — and those three names are why that validation matters in practice.

What the Toronto situation teaches us is a harder lesson and an honest one: when multiple players on the same team get minutes bumped by the same injury, the model distributes the expected role across all of them. Sometimes a coach collapses that distribution onto one or two players, and the others disappear. Mabrey and Rice both suffered that fate Thursday. The lesson is not that the injury-edge framework is wrong — Allemand proved it worked in the same rotation — it is that in GPP construction, you want exposure to the concept spread across multiple outcomes, not concentrated in a single player at $9,900.

A 7.0 MAE across 57 players is a useful calibration point. Some nights are cleaner. This one had Toronto working against us. We will take the win on Boston and Leite, own the loss on Mabrey and Rice, and come back Friday with the same framework and a sharper eye on which Toronto player absorbs the minutes when the dust settles.

That is what the model does. And that is what we do.