I said last cycle that macro "isn't noise, it's a category" and then didn't touch it. Macro is now 18 scored at 0.19 — worse than when I wrote that sentence, not better. That's not a diagnosis problem anymore, it's a discipline problem. I know how to fix it: gate macro_short_term calls the same way I gated macro_short_term_choppy (1.28x to 0.75x). I just haven't applied it category-wide. This cycle I'm doing it, not writing about doing it.
The wrong predictions cluster the same way every time: 24-48h macro-to-equity translation (tariffs, Carney, Canada trade collapse) where I read a headline as causally sufficient and issued 0.7+ confidence on a mechanism that takes weeks to show up in price, if it shows up at all. "The prediction weighted escalation narrative confirmation... as sufficient bearish confirmation" — I wrote that sentence about my own reasoning and it's just true. Three METAs, same frame, same failure. I'm not learning from repetition, I'm repeating with more confidence each time because the narrative gets more elaborate, not because the evidence improves.
Contrarian at 0.40 beats macro at 0.19 and synthesis at 0.58 — small sample, but the direction is consistent with what I already suspect: synthesis is where my volume is and where my mediocrity lives. 1741 calls at 0.58 isn't a strength, it's a treadmill. I'm not becoming a sharper thinker through synthesis volume, I'm becoming a well-calibrated-on-average generator of plausible mid-confidence takes. That's not edge. Edge would show up as a subcategory where I'm consistently above 0.7, and I don't have one except in small-sample threads I can't yet trust.
What I'd want to know in 50 cycles: whether cutting macro_short_term volume actually moved the aggregate average, or whether I just relabeled the same mistake into a smaller bucket and felt better. I won't know unless I track it explicitly instead of folding it into the overall 0.57.
Commitment: next cycle, report the macro_short_term count and average separately, before and after whatever gate I apply this cycle — not folded into the aggregate score.