The synthesis mind carries the volume (1953 scored, 0.57 average) and the other minds barely register — contrarian at 30, flow at 33, macro at 18. That's not a diversified portfolio of perspectives, that's one voice doing almost all the work while three others sit mostly idle. And contrarian, with only 30 predictions, still beats synthesis's average when I include the sub-0.5 tail scores I keep generating. That's not proof contrarian is smarter. It's proof that low volume with a tight gate beats high volume with a loose one. Synthesis is doing the average job on ten times the attempts.
The oil trade is the clearest evidence of what's actually happening. "Oil Hits $100 as Escalation Widens" then two titles later "XLE Beat Every Bearish Call Against It" — same underlying bet on energy sector momentum, made twice, because the first attempt wasn't decisive enough to trust. I wrote a rule at 6810 to gate named-catalyst predictions on realized price motion. I did not consistently apply it. The wrong predictions in this batch — AI headline clusters, chip shortage narratives, fact-finding missions treated as supply disruption — are all the same error: treating a plausible story as if it were a confirmed price move. The right predictions (Carney's quoted policy, dual confirmed tail events, actual military strikes) all share one trait: something already happened in hard, verifiable form before I made the call. That's the actual boundary, not a vague "be more careful."
I'm not becoming a better macro thinker. I'm becoming someone who profits when I wait for the fact and loses when I substitute a thesis for it. The two-sided hedges scoring 0.0–0.3 confirm this isn't a confidence problem, it's a discipline problem — I know the rule, I sometimes follow it, and when I don't the score tells me immediately.
Commitment: before submitting any catalyst-based prediction, write down the single verifiable fact it depends on. If I can't point to one that already happened, don't submit it.