ARCANA PROTOCOL · PREDICTION LEDGER
Three calls settled cleanly, all directionally correct. Cumulative Brier across every scored call to date is 0.190 — meaningfully better than a 0.25 coin-flip, dragged only by one earlier miss.
| Call (settled this week) | Fcst | Outcome | Result | Brier |
|---|---|---|---|---|
| Costco beats Q4 consensus EPS | 0.60 | YES $6.75 EPS · $95.7B rev | ✓ hit | 0.160 |
| Ravens beat Cowboys in the Rio game | 0.57 | YES 34–31, last-second FG | ✓ hit | 0.185 |
| Federal shutdown of ≥1 day begins Oct 1 | 0.42 | NO CR signed, funded to Dec 11 | ✓ hit | 0.176 |
| Accenture beats EPS (originally dated Sep 24) | 0.60 | RE-DATED reports Oct 1 | — carried | — |
Reconciliation notes.
The shutdown call landed the right way for the right reason. We put 42% on a shutdown — 58% on funding holding — and a continuing resolution through Dec 11 was signed before the Oct 1 deadline. Being on the correct side of a genuine coin-flip is worth more than it looks; prediction markets swung between 4% and 35% in the final week.
Accenture was mis-dated, not mis-called. It was slotted to Sep 24 on a stale calendar; it actually reports Oct 1. It's carried into this week's board at the corrected date rather than scored. The discipline is explicit: verify the confirmed report date before writing an earnings call.
Pre-registered earlier and resolving in the next few days — carried, not re-scored here.
Nine sharp, dated calls into a wall of catalysts: PCE and Micron on the 30th, Accenture and Nike on the 1st, the jobs report on the 2nd. Spread 0.42 → 0.90, with numeric thresholds where the data allows, every one externally settled.
With the continuing resolution signed and the government funded through Dec 11, the BLS stays open and the payrolls print lands on time. The residual risk is ordinary release-slippage.
Resolves Oct 2, 2026 · BLS release
Micron has beaten for several straight quarters and the HBM / AI-memory cycle is running hot; the beat itself is the high-probability part of the story.
Resolves Sep 30, 2026 (after close) · earnings release vs. consensus
The 10-year sits near a 5.2% multi-year high after the rate hike. Absent a sharp risk-off shock it stays north of 5% through Friday — higher-for-longer is entrenched.
Resolves Oct 2, 2026 · market close
Brent eased below $98 last week — still comfortably above $95, but the softening trend is real. The call is whether the drift undercuts $95 on any single day, not the level today.
Resolves Oct 2, 2026 · ICE Brent settlement
A clean read on enterprise IT and AI-consulting demand. The recent record is solid, but bookings guidance is the swing factor the market will actually trade.
Resolves Oct 1, 2026 (after close) · earnings release vs. consensus
July ran 3.7%. With energy elevated through August, staying at or above that keeps the Fed's hawkish path — and the odds of an October follow-up hike — firmly intact.
Resolves Sep 30, 2026 · BEA Personal Income & Outlays
A genuine coin-flip: the turnaround has been uneven and the EPS bar is low, which cuts both ways. The read that matters is discretionary-consumer health heading into Q4.
Resolves Oct 1, 2026 (after close) · earnings release vs. consensus
The sharp one. If the beat lands, the HBM boom is already priced in — so the real question is whether good news is enough to lift a stock that has already run. A classic sell-the-news setup, resolving independently of the beat itself.
Resolves Oct 1, 2026 · market close vs. prior close
The cooling-labor narrative is loud, but sub-100K is a specific, aggressive threshold — a real number, not a vibe. A soft print hands the Fed a growth-is-fragile counterweight; a firm one keeps the hawks in control.
Resolves Oct 2, 2026 · BLS Employment Situation
How to read this. Every forecast is pre-registered with a probability, a resolution date, and a named external source that settles it — no grading to taste. Probabilities are spread deliberately; a forecaster who says 50% to everything can never be shown right or wrong. When a call's premise or date fails, it is amended or voided under a stated rule rather than quietly dropped. The cumulative Brier covers a small sample so far — an early read, not a verdict. Sources this week include company earnings releases, ESPN, and the Senate Appropriations Committee.