The read
The conviction set widened by one from yesterday: Bitcoin and Ethereum again, now joined by Solana, clear the engine's gate, and all three lean lower at every horizon we scan. A wider group of large caps (XRP, DOGE, BNB, AVAX and others) also leans lower on momentum, but their corpus support is too thin or too concentrated to pass, so we issue nothing on them.
Bitcoin and Ethereum already carry a full set of standing calls (a 7-day and a 3-day each), so the only open slot is on Solana, where the 3-day slot freed up when its earlier 3-day call matured today.
New call — Solana, 3 days (experimental horizon). The engine puts roughly 76% on Solana touching −1.62% from issuance at some point within three days, against a base rate of about 68% for a move that size. Momentum carries most of the lean; this week's coverage reads less constructive than Solana's usual tone, with eleven items from a mix of analyst, insider and mover sources. The gap over the base rate is modest and the engine's lift is unvalidated, so treat it as a lean, not a forecast with earned skill. 3-day calls are a pre-registered experiment and sit outside the public accuracy claim.
Standing reads. The week-ahead regime stays choppy (issued 16 Oct). The structural scan re-scored a new Bitcoin record within six months, a broad altseason and a multi-major drawdown, and Solana versus Bitcoin over 90 days; each either matches its standing call closely or its question already holds the maximum two standing reads, so nothing new was issued there.
Receipts
Two calls resolved since the last run, both hits:
- Solana touches −1.82% within 3 days (experimental) — issued 14 Oct 14:49 UTC, matured 17 Oct 14:49 UTC, resolved 17 Oct 17:34 UTC. Hit: the level was reached on the path, even though price closed the window up 0.56%. Touch calls grade on whether the level is reached, not where price ends.
- XRP versus Bitcoin over the quarter — issued 19 Jul, matured 17 Oct 08:19 UTC, resolved 17 Oct 17:34 UTC. Hit: XRP gained 17.7% against Bitcoin's 25.6%, so it finished behind on return. This card was a historical-frequency read, not a signal-driven one, so the hit says the base rate held, not that our signals had skill.
How the record stands
Our probabilities have not yet beaten the base rate. On claim-eligible 7-day calls under the current grading rule, over the trailing 30 days to this run (17 October), the Brier Skill Score against the base rate is −0.0735 (n=7): our stated probabilities score slightly worse than quoting the historical frequency and ignoring every signal. The breakdown shows why: reliability error of 0.0989 (where over-confidence shows up) outweighs resolution of 0.0544 (how well the probabilities separate outcomes). The 1-day lane, also part of the claim, reads −0.171 (n=23). Both samples are far below the 25-call minimum, so no skill claim is made in either direction.
As a direction-only note, which does not read the probabilities we publish, the 7-day hit rate sits 17.8 points below its average base rate.
