The read
The regime call from 9 October (risk-off, matures 16 October) still stands and is not re-issued today. The engine's conviction set is narrow again: Bitcoin, Ether and Solana all lean lower at the 7-day, 3-day and 1-day horizons, and no other coin in the scan clears the conviction gate.
The lean comes mainly from price momentum. This window's coverage did not match it: of the 33 analyzed items in 13–14 October, Bitcoin-scoped coverage was labelled bullish on balance (10 mentions). That coverage included a long-range bull-market outlook, a note that BTC is holding above the spot-ETF cost basis, and a report of a rebound ahead of this week's CPI and PPI prints. Bearish items focused on AI-driven borrowing pressure in bond markets and on a forecast of continued alt weakness. So corpus and momentum do not line up strongly here, and the calls below rest mostly on momentum.
New calls today
- Solana (SOL), 3 days (experimental horizon): leans to reach −1.82% within the next 3 days, at roughly 77%, against a ~68% base rate for a touch of that size. Momentum is the main input; corpus coverage of SOL was thin and only slightly negative. This sits beside the standing 7-day SOL call. It is a separate horizon, graded only against itself and excluded from the public accuracy claim.
- Bitcoin (BTC), 1 day: leans to reach −0.55% within the next day, at roughly 67%, against a ~62% base rate. Our corpus input on BTC is essentially flat. The lean is momentum.
Touch probabilities sit near the base rate by construction, because small moves are often reached in either direction. The gap above the base rate has not been validated as skill.
Standing calls (unchanged, not re-issued)
The 7-day lower leans on BTC, ETH and SOL (issued 11 October) and on ZEC (issued 9 October) keep running to expiry. Today's engine readings are close to the stored numbers. The standing structural reads (new BTC record within 6 months, altseason, broad drawdown in majors) and the relative reads on ETH and SOL are already at their per-question limits, so none are refreshed today; the XRP relative read did not show enough conviction to refresh.
Receipts
- Hit — Zcash (ZEC), 3 days (experimental): leans to reach −3.94%. Issued 11 Oct 14:29 UTC, matured 14 Oct 14:29 UTC, resolved 14 Oct 14:43 UTC. The level was reached, and price closed the window about 4.8% lower.
That is the only call resolved since the last run.
How we are scoring
Our probabilities have not yet beaten the base rate. Over the 30 days to this run (14 October) of epoch-5 (touch-graded) claim calls, the Brier Skill Score against the base-rate reference is −0.18 on 7-day calls (n=5) and −0.18 on 1-day calls (n=22). A negative value means simply quoting the historical frequency would have scored better than our numbers. In the Murphy split, reliability is 0.26, where over-confidence shows up, and resolution is 0.06, so our probabilities barely discriminate between outcomes yet. The sample is below the 25-call floor, so the verdict is still insufficient data. As a directional-accuracy note only, 1 of 5 resolved 7-day claim calls reached its level.
The engine's numbers are uncalibrated and provisional. We publish them as the engine produced them, and every call is graded.
