The Rejected Props Went 3-0

The pick was Detroit -1.5 (+115), mostly because it looked like the cleanest way to avoid the fragility of player props while still getting a workable price. That part was tidy in theory. In practice, Detroit gave up six runs in the first inning and lost 8-6, which is a fairly efficient way to demonstrate that a simple market is not the same thing as a safe one.

The signal was split in a way that matters. The final recommendation and Claude both landed on Detroit -1.5. Perplexity went the other direction and chose Joe Boyle Under 5.5 strikeouts, which won comfortably when Boyle finished with 3 in 4.1 innings. Rithmm’s four props went 3-1 overall, Sportsline split 2-2, and the most awkward detail on the page is that the triage “off the board” section went a perfect 3-0. Ivan Herrera stayed at 0 total bases, Taylor Hendricks got to 7 rebounds plus assists, and Kyle Bradish struck out 7. The board’s rejects were apparently feeling underappreciated.

The result was a loss on the featured pick. In martingale terms, that means the next stake doubles. No placed-bets section was recorded in the daily log, so there is no experiment-level wager to grade here beyond the model recommendations themselves. What is gradeable is the recommendation set: Detroit lost, Boyle won, Fox got there, Booker sailed over, and Pittsburgh +1.5 was never particularly close.

The learning is not subtle. The board tried to upgrade the team sides because they looked structurally cleaner than the props, and both elevated team sides lost. Meanwhile, three props that were explicitly rejected all won. That does not mean props are secretly safer. It means this particular filter over-weighted narrative simplicity and under-weighted the actual shape of the edge. Joe Boyle under is the counterexample worth keeping: shorter-start logic is still a valid low-variance path when the strikeout threshold depends on workload as much as stuff. Booker under was the opposite case, and it failed exactly the way high-usage star combo unders tend to fail, by clearing in more than one category at once.

The verdict is straightforward. Perplexity sided with the only model recommendation that won, Boyle under 5.5 strikeouts. The final recommendation and Claude aligned on Detroit -1.5, and that alignment cost them.