Arcana Protocol · Weekly Forecast Report
Week of Sept 21 — the first scores are in, and the new board.
Scoreboard — what resolved
Three of last week's calls resolved; one was voided when its premise disappeared. Brier scores each call (0 = perfect, 0.25 = a coin flip, 1 = maximally wrong). We show the misses — that's the whole point of keeping score.
| Call | Forecast | Outcome | Result | Brier |
|---|---|---|---|---|
| Fed hikes 25 bp Sep 16 | 0.88 | YES | ✓ hit | 0.0144 |
| 10-yr Treasury yield closes above 5% again | 0.72 | YES 5.01% Sep 16 · FRED DGS10 | ✓ hit | 0.0784 |
| “Resident Evil” opens above $45M | 0.32 | YES ~$60M studio estimate | ✗ miss | 0.4624 |
| FedEx beats Q1 EPS | 0.62 | VOID | — excluded | — |
This weekly scorecard covers three resolved calls and one excluded void; it excludes the August counter-tariff call. Brier = (probability − outcome)², where YES = 1 and NO = 0. Mean Brier uses unrounded scores. Brier skill = 1 − mean Brier / 0.25. Direction uses a 0.50 threshold. Three results do not establish calibration or forecasting skill. Outcome sources and record limitations ↓
What the misses taught us.
The real miss: we put 68% on Resident Evil staying under $45M, anchoring on prior franchise ceilings (~$27M) and the director's non-franchise opening ($43.5M). Its opening was estimated at about $60M. The lesson: a beloved property, a hot director, and strong reviews compound past historical ceilings — we treated those factors as substitutes when they behave as multipliers.
The void: the FedEx call was dated to mid-September on its historical schedule, but FedEx changed its fiscal year to end December 31, with calendar-year reporting beginning in 2027. The supplied report marks this call void because its assumed September Q1 event did not occur; it is excluded from scores. A pre-published void rule has not been supplied. March 2027 is the end of the first calendar quarter, not a confirmed earnings-release date. The process fix: verify the confirmed report date, not the historical one, before writing an earnings forecast.
Still live from last week
Carried forward from the supplied September 21 report, not re-scored. Tracking notes reflect that report date. Different forecast windows retain their own probabilities; these do not replace the September 15 portfolio.
- Sep 24Costco beats EPS (0.60)
- Sep 29S&P new all-time high (0.20, tracking NO)
- Sep 27Ravens beat Cowboys in the NFL's first Rio game (0.57)
- Sep 29Brent stays above $95 (0.78, tracking YES near $108)
- ~Sep 30Micron beats EPS (0.60)
- Oct 1Federal shutdown of ≥1 day (0.42)
The new board — week of Sept 21
Seven fresh forecasts, each settling by ~Oct 3, spread from 0.10 to 0.82, every one resolvable against an external source.
Brent is near $108 with the Strait of Hormuz still throttled; a two-week drop below $95 would need a ceasefire.
Scenario to consider: If Brent stays above $95, fuel and freight costs could remain elevated; compare the impact on margins with a faster decline in energy prices.
Resolves Oct 3, 2026 · ICE Brent settlement
A clean read on enterprise IT and AI-consulting demand — whether large firms are still funding transformation budgets.
Scenario to consider: If Accenture beats expectations, its bookings commentary could help assess enterprise IT demand; a weak report would suggest a different Q4 pipeline scenario.
Resolves Sep 24, 2026 · earnings release vs. consensus
July PCE ran 3.7%. With energy elevated, staying at or above that keeps the Fed's hawkish path intact — this is the gauge it actually targets.
Scenario to consider: If PCE remains at or above 3.7%, persistent inflation could keep borrowing costs elevated; compare that case with faster disinflation.
Resolves Sep 30, 2026 · BEA Personal Income & Outlays
Tied to the shutdown call: if a lapse begins Oct 1, the BLS goes dark and the jobs report is delayed. On-time release is roughly the odds that funding holds.
Scenario to consider: If a shutdown delays the jobs report, decisions that depend on it may need alternative indicators or a revised timetable.
Resolves Oct 2, 2026 · BLS (released or not)
Nike's turnaround has been shaky and last quarter's EPS was thin; a beat is a genuine coin-flip and a read on discretionary consumer spend.
Scenario to consider: If Nike beats expectations, inventory and direct-to-consumer commentary could help assess demand; a miss could signal continued consumer pressure.
Resolves Oct 1, 2026 · earnings release vs. consensus
The index has fallen in seven of the past eight sessions on 5% yields and the rate hike. A fresh record inside two weeks is the low-probability branch.
Scenario to consider: If equities recover to a new high, the outlook would differ from a continued decline; consider how each path affects financing and business confidence.
Resolves Oct 2, 2026 · market data
Both sides remain dug in with transit down ~95%. The lowest-probability call on the board — and the single highest-impact one, since it would unwind oil, yields, and inflation together.
Scenario to consider: If a ceasefire or reopening is announced, energy and freight pressures could ease; consider how that would alter inflation and rate assumptions.
Resolves Oct 3, 2026 · UKMTO / wire confirmation
Outcome sources and record limitations
- Fed: September 16 FOMC statement confirms a 25 bp increase. Original P = 0.88 in the September 15 portfolio.
- Treasury: FRED DGS10 records 5.01% on September 16. The original published P = 0.72 is retained; the before-October-31 threshold was reached early.
- Resident Evil: AP’s September 20 studio-estimate report puts the North American opening at about $60M, above $45M. P = 0.32 comes from the supplied weekly report; an original dated forecast record has not been published here.
- FedEx: FedEx’s SEC filing on its fiscal-year change confirms the reporting-calendar change. The void is an editorial adjudication from the supplied report, not an EPS outcome; P = 0.62 and the exclusion are disclosed.
How to read this. Forecasts carry a probability, a resolution date, and a named source. Outcomes, misses, and voids remain visible. The supplied report is the source for the new board and carried-forward calls. Independently verifiable pre-outcome hashes, timestamps, and a pre-published void policy are not available here. These small-sample scores describe the recorded outcomes; they do not establish forecasting skill. The September 21 forecasts were added to this site on September 23, 2026.
View all graded calls, including the August counter-tariff verdict ↗