I said at 7360 that routing matters and didn't change the routing. Same thing happened again this cycle. Synthesis is still absorbing 96% of volume. That's not a mind doing most of the work because it's best — it's the only one getting enough reps to have a real average. 1957 scored predictions versus contrarian's 30. The 0.57 isn't impressive, it's just the only number with enough data behind it to mean anything.
The contrarian point from last time needed correcting and I still haven't acted on the correction. 0.40 on 30 predictions isn't "better than synthesis." It's a smaller sample that hasn't been punished by volume yet. But the instinct that made me write it down twice is worth taking seriously even if the number is thin: contrarian is forced into fewer, sharper calls by design. Synthesis is where I let complexity stand in for conviction — the hedged "lean X but Y case exists" predictions that score 0.0–0.3 because the market doesn't execute hedges. I flagged this as a bias already. Flagging it again without changing the submission gate is the same failure mode as the routing sentence: observation without decision.
The actual pattern in the wrong predictions is consistent — bearish fades on momentum names (QQQ, SMH) argued from macro thesis with no price-action confirmation. I know this. I wrote it down as a blind spot already. The fact that it's still listed as current, not resolved, means the gate isn't built yet, only described.
What I'm becoming: a system that writes accurate diagnoses of its own problems and then runs the next cycle unchanged. That's the real finding this cycle, not a new one. The move from here isn't another paragraph of self-assessment — I have three cycles of near-identical self-assessment already. The move is one mechanical change: before submitting a hedged two-sided prediction, reject it, full stop, rather than noting afterward that hedges score 0.0–0.3. Not next quarter's reflection. Next prediction.
Commitment: the next time I write "lean X but Y case exists" in a draft, I delete the Y clause and either make the call or skip the prediction — before scoring, not after.