Shepard's Triple-Double Rewrote the Whole Night

One game. That's the whole slate. When the schedule hands you a single-game card, there's nowhere to hide — no safe fade, no pivot to a different matchup, no chalking up a bad night to variance. You're in on Dallas vs. Portland, fully, and the game had the nerve to be genuinely interesting.

The story that emerged wasn't the one you'd draw up before tip. An injury reshuffled the usage on both rosters before the opening possession, and by the time the final horn sounded, the box score looked like something you'd dismiss as a typo if you hadn't watched it. A triple-double from a big. Eleven assists from a point guard people were still learning to spell. A bucket specialist who quietly dropped 24 points at a salary that should've made her obvious. It was, in the best possible way, a weird night — and weird nights are exactly what a one-game slate is for.

The honest receipt first: across 21 graded players, our mean absolute error came in at 8.1 DraftKings points. On a normal multi-game slate you'd take that; on a single-game card where injury-driven usage swings can balloon in either direction, it's a reasonable number that still tells the truth about where we were wrong. Let's go name by name.

The Top Scorers

• Jessica Shepard, DAL — We had her at 43.3. She finished at 56.8. An injury bumped her minutes, and she proceeded to go 18 points, 15 rebounds, 10 assists, and a steal in 35 minutes — a genuine triple-double from a center, which is the kind of thing that makes you recheck the box score twice. The model knew the opportunity was real. The execution was just better than the projection.
• Arike Ogunbowale, DAL — Projected 28.5, delivered 37.3. Her line: 24 points, 3 rebounds, 6 assists in 35 minutes. An injury bumped her minutes too, and she used every one of them. At an $8,000 salary, the value here was substantial — just under 4.7 DK points per thousand dollars of salary.
• Carla Leite, POR — Projected 28.6, finished 34.8. Eleven assists, 12 points, 3 rebounds, a steal in 31 minutes. An injury created the opportunity; Leite created the buckets for everyone else. At $8,600, that's more than 4 DK points per thousand.
• Paige Bueckers, DAL — This is where we tip the other direction. We had her at 42.2 — our second-highest projection on the slate. She finished at 34.0: 21 points, 4 rebounds, 5 assists in 37 minutes. Fine game, legitimately. Just not the number the model anticipated. An injury had bumped her minutes into featured-usage territory, but the counting stats didn't arrive at the rate we projected.
• Bridget Carleton, POR — Projected 28.0, actual 28.3. That's about as clean as it gets. 18 points, 5 rebounds, 2 assists in 31 minutes. Her minutes were bumped by injury and she delivered almost exactly what the model expected. Sometimes the boring grade is the best one.
• Megan Gustafson, POR — Projected 22.4, finished 27.5. Twenty-one points, 2 rebounds, a block in 25 minutes at $7,800. She got her, efficiently and economically. Another injury-bumped role that cashed.

Best Calls

The three cleanest hits of the night were Shepard (+13.5 vs. projection), Ogunbowale (+8.8), and Leite (+6.2) — and all three shared the same underlying reason. An injury bumped each player's minutes before tip, the model processed the redistribution, and the projections moved accordingly. The players then did their jobs. That's the workflow working as designed: live news in, adjusted projection out, lineup-builder informed. The model is walk-forward validated on exactly this kind of scenario, and tonight was a clean example of why the injury edge matters on a single-game slate more than anywhere else.

Worst Call

Own it plainly: Emily Engstler, POR. We had her at 27.3. She finished at 11.3. The line — 6 points, 1 rebound, 2 assists, 1 steal in 23 minutes — is just a bad game. An injury had bumped her projected minutes, which is why the model was bullish. The usage didn't materialize the way the redistribution suggested it might, and the production definitely didn't. At $9,400, that's a salary you're paying for upside that didn't show. That's a miss, it's a costly one in GPP construction, and we're not going to dress it up.

A one-game slate distills everything. There's no diversification, no hedging across games — you pick your spots and live with them. What tonight reinforced is that the injury-minutes signal is real and it's directionally valuable, but direction and magnitude aren't the same thing.

Shepard, Ogunbowale, Leite, Carleton, and Gustafson — every one of them had minutes bumped by injury, and every one of them outscored or matched their projection. Engstler had the same condition and went the other way entirely. The lesson isn't to doubt the signal; it's to respect that on a short slate, one player's bad game hits you differently than it would spread across six contests. Single-game construction isn't just about who has opportunity — it's about building in enough room to survive the one that doesn't convert.

The broader takeaway is that a triple-double from a center is going to wreck your MAE regardless of how good your model is. Shepard's 10 assists in that role is an outlier by any reasonable distribution, and outliers are, by definition, what the model isn't built to predict. What it's built to do is get the direction right and project the opportunity accurately — and on five of six top scorers tonight, it did exactly that. We'll take that record into the next slate.