The Clean Pick Won, and the Messier One Still Lost

The pick was Jalen Johnson under 17.5 rebounds plus assists, which is another way of saying the board chose the stat shape with fewer ways to go wrong. The model gap was real, the recent hit rate was the best on the sheet, and unlike the MLB singles options, it was attached to a confirmed playoff game instead of timing noise and box-score archaeology. Johnson finished with 7 rebounds and 3 assists, so the line was never under serious threat.

The signal was unusually cooperative. The final call, Claude, and Gemini all landed on Johnson. Perplexity and ChatGPT wandered off toward MLB singles, with Abrams and Jung Hoo Lee respectively. The irritating part, if you enjoy dramatic disagreement, is that they were right too. Abrams cleared at 1-for-4, Lee logged 2 singles, Gobert cashed both unders, Robinson stayed under, and the board’s only real failure belonged to Karl-Anthony Towns clearing 35.5 PRA with 37.

The result was a win. More specifically, it was a placed win, which matters because the 3-loss martingale experiment was on Johnson under 17.5 rebounds plus assists. So the sequence resets instead of doubling into tomorrow, which is the closest this format gets to a relaxing evening.

The learning is not that every model agreed, because they did not. It is that the cleaner stat construction held up exactly the way the triage said it might. Towns PRA was flagged as the shakier under because PRA gives a good player too many doors to beat you, and that was the lone miss. Johnson’s rebounds-plus-assists under had a narrower failure path and behaved accordingly. Structural simplicity still counts for something, even on a board full of winners.

The verdict is plain. The placed bet followed the final recommendation and Claude, and that alignment paid off immediately. Perplexity found a winner too, but on this slate the lower-noise NBA under was the one that actually got staked.