Harness Analysis
| Candidate | Hermes | ChatGPT | Claude | Gemini | Perplexity | Board |
|---|---|---|---|---|---|---|
| Marcus Semien Under 1.5 Total Bases | — | T3 | T3 | T3 | T3 | Loss |
| Kody Clemens Under 1.5 Total Bases | — | T2 | T2 | T3 | T3 | Win |
| Michael Soroka Over 1.5 Walks Allowed | — | T3 | T3 | T3 | T3 | Win |
| Victor Wembanyama Under 12.5 Rebounds | — | T3 | T3 | T3 | T1 | Loss |
Final Decision
No Tier 1 consensus, two direct SKIP votes, MLB context/data-integrity caveats, fragile total-bases-under cluster, and no candidate cleared the survival-first threshold for a hard 3-loss experiment.
No official bet placed; Martingale sequence unchanged. Board went 2-2: Semien lost with 8 TB, Wemby lost with 15 REB. Clemens and Soroka both won. PASS avoided the Perplexity-driven Wemby loss.
Candidates
- Rithmm
- 61.2%
- BE
- 59.5%
- DTM
- 4.3%
- Recent
- L10 9/10
- Rithmm
- 62%
- BE
- 60.3%
- DTM
- 4.4%
- Recent
- L10 7/10
- Rithmm
- 64.7%
- BE
- 63.1%
- DTM
- 4.5%
- Recent
- L6 3/6
- Rithmm
- 69.2%
- BE
- 56.9%
- DTM
- 28.7%
- Recent
- L10 4/10
Harness Notes
Official PASS preserved the Martingale sequence. The paper lean (Wemby U12.5) lost with 15 rebounds in 26 minutes. PASS avoided the Perplexity-driven forced-action candidate.
PASS correct officially. Forced pick Clemens U1.5 TB would have won (0-for-2, 0 TB), but SKIP enforcement was good given the Coors/Semien and Wemby landmines on the same board.
Correctly identified Clemens as the best settled alternative (won, 0-for-2, 0 TB) while still framing it as not recovery-safe. Avoided the Wemby loss.
PASS correct officially. Correctly flagged board/data integrity issues and avoided Semien (8 TB, massive loss) and Wemby (15 REB, loss). Overly absolute confidence, but direction was right.
Worst forced-action signal. Overweighted model cushion and underrated superstar rebound ceiling and poor recent form. Wemby finished with 15 rebounds in 26 minutes — a clear loss.