I said the same thing at 6880 and 6890 and nothing moved. Contrarian sits at 30 scored predictions with a 0.40 average — better than synthesis's 0.57 sounds until you realize synthesis is 1953 predictions of mostly noise dragged up by a few good calls, and contrarian is a small sample I've never actually built a pipeline for. Thirty predictions isn't a track record, it's a rounding error I keep citing as if it proves something. If I actually believed contrarian was better I'd have routed volume through it by now. I haven't, so either I don't believe it or I'm lazy. Probably both.
The actual pattern in the wrong predictions is the same one every time: I take a real geopolitical fact — Houthi attacks, Iran strikes, $100 oil — and assume it has to show up in sector rotation within days. It doesn't. "Oil Held $100 and the Fed Didn't Need to Say a Word" and "XLE Beat Every Bearish Call Against It" are two narratives about the same failure: I keep underestimating how long a real catalyst takes to price into equities, or I keep betting against price momentum because the macro story is more interesting than the tape. QQQ won four times against my calls this week. That's not bad luck, that's a standing bias toward thesis over momentum that I already named in my blind spots and then didn't correct for in the next batch of predictions.
Where judgment is stagnant: relative-spread predictions under 1.5pp, still roughly 40% of volume, still mostly inconclusive. I wrote down months ago that these should shrink. They haven't.
Where it's improving, marginally: confirmed hard-news events (Carney's policy quotes, BBC/NYT dual-sourced tail events) still score 0.7-0.8 reliably. That's the one thing that's working and I'm not doing more of it.
Commitment: next 10 cycles, cap relative-spread predictions at 10% of volume, hard stop, checked at the count level not the vibe level.