Today's read
The analyzed corpus for 20–21 October is thin: 13 items, down from 39 in the prior window. Bitcoin and Ethereum coverage leaned constructive. Equity-market coverage leaned bearish, and one analyst flagged that an equity correction could push crypto lower. A reported exchange exploit was the highest-importance story in the window. The standing market regime call is choppy and runs until 23 October, so no new regime call is issued today.
New calls (touch-graded: a call scores if price reaches the stated level at any point within the horizon)
| Call | Probability | Base rate |
|---|---|---|
| NEAR — reach +8.96% within 7 days | ~69% | ~65% |
| NEAR — reach +3.39% within 1 day | ~68% | ~64% |
| Bitcoin — reach +0.88% within 1 day | ~62% | ~57% |
| Zcash — reach +2.30% within 1 day | ~59% | ~55% |
| Solana — reach −1.65% within 3 days (experimental) | ~73% | ~68% |
NEAR is the only coin outside the large-cap majors that passed the model's conviction filter today. Its read comes from NEAR-specific coverage, including a report that Zcash-linked holders are rotating into the token, plus upward momentum. It is not simply tracking Bitcoin.
Bitcoin and Zcash now lean higher over the next day, while their 7-day lower leans from 18 October stay open. Each horizon is scored separately, so a short-term upward lean does not cancel or replace the 7-day call. The Zcash read is driven almost entirely by momentum after a strong rally, since coverage of Zcash was mixed.
Every gap above is a few points over the base rate. A touch probability near 60–70% mostly reflects how often moves of that size happen, which is a matter of units and not of skill.
Not issued:
- XRP: already carries two open calls.
- Solana 1-day: one was issued less than 48 hours ago and the read has not changed materially.
- Ethereum: no lean in either direction.
- AVAX, DOGE, ADA, SUI, CRV: momentum alone, without enough coverage behind it, so they did not pass the conviction filter.
- Standing questions (Bitcoin all-time high, altseason, majors drawdown): each already has two open reads.
- A NEAR-versus-Bitcoin relative read: the six standing-question slots are already in use.
Receipts: resolved since the last read
| Call | Issued | Expired | Resolved | Result |
|---|---|---|---|---|
| Solana — reach −0.93% within 1 day (72%) | 20 Oct 21:20Z | 21 Oct 21:20Z | 21 Oct 21:24Z | Hit: level reached; closed +1.21% at expiry |
| Zcash — reach −3.75% within 3 days (67%, experimental) | 18 Oct 17:45Z | 21 Oct 17:45Z | 21 Oct 21:24Z | Hit: level reached; closed +24.84% at expiry |
| Ethereum — reach −2.20% within 3 days (67%, experimental) | 18 Oct 17:45Z | 21 Oct 17:45Z | 21 Oct 21:24Z | Miss: level not reached; closed +6.17% |
That is 2 hits and 1 miss. Both hits touched their downside level early and then closed well above where they started. Under touch grading they count, but they are not evidence that prices fell.
Accuracy (trailing 30 days to 2026-10-21)
- 7-day calls (counted in the accuracy record), n=10: the Brier Skill Score against the base rate is −0.0034. That means our probabilities have not yet beaten simply quoting the historical base rate, though they are close to level. Reliability error is 0.0375 (this is where overconfidence shows up). Resolution is 0.045 (this measures how well our probabilities separate outcomes).
- 1-day calls (counted in the accuracy record), n=26: the Brier Skill Score is −0.1464, clearly worse than the base rate.
- The sample is below the 25-call floor for 7-day calls, so the verdict is insufficient data.
- A separate direction-picking measure, which ignores the stated probability, reads −0.0113 on the 7-day calls.
- The 3-day experimental lane is excluded from every figure above.
