Where the read sits
The corpus window for this run covers 28-29 August: 56 analyzed items, ten distinct sources spanning analysts, movers, insiders and regulators. Coverage bias reads bullish on the market aggregate and on each of the three most-covered coins - Bitcoin (8 mentions), Solana (3) and Ethereum (2) - though every one of those mention counts is down on the prior window, so the tone is constructive on thinning volume rather than broadening.
Sector coverage tilts to AI (4 mentions, bullish) and derivatives (4, neutral). The derivatives colour is the more useful of the two: within the window we have an options-expiry note pointing at pressure below spot on Bitcoin, an observation that Ethereum's weekly price rise diverged from net taker volume (possible absorption), and a report of a large short-liquidation surge. That is a market being carried by positioning as much as by flow.
What was issued
One new call: Solana leans higher over the next seven days. The engine puts it at roughly 57%, against a historical frequency of about 44% for a seven-day gain in this asset. Both available signals - the analyzed corpus and price momentum - point the same way, which is what cleared the conviction gate; the third leg (an external consensus input) is not present on this read, as it has not been on any price call for some time, so the blend is our own two signals.
Read the provenance honestly: 8 of the 10 Solana items in the window come from one analyst account. The independent corroboration is a single on-chain data report of record Solana ETF inflows, plus one bearish options note tied to an expiry that has since passed. The lift over the base rate is real but small, and it is not evidence of demonstrated skill - the weight the engine puts on signal strength is still an unfitted placeholder.
Nothing else was issued, and that is deliberate. Bitcoin and Ethereum both cleared the conviction gate too, but each already carries a live call in the same direction (a seven-day Bitcoin lean running to 3 September, a three-day Ethereum lean running to 31 August). Re-stating an existing view on an overlapping window is not a second forecast, so both were left alone. The remaining scan entries - BNB, XRP, DOGE, AVAX, ADA - convict on momentum alone with no corpus behind them; a single signal wearing the clothes of two does not get a slot.
The week-ahead regime call issued yesterday (choppy, running to 4 September) stands unchanged. On the structural board, the Bitcoin-record question re-scored higher this run, but that question already carries its two permitted standing reads and none has matured, so there was no room to add a third.
Receipts
Resolved since the last run:
| Call | Issued | Matured | Resolved | Stated | Outcome |
|---|---|---|---|---|---|
| Ethereum - leans higher over 7 days | 21 Aug 20:18Z | 28 Aug 20:18Z | 29 Aug 18:54Z | 55.7% | Correct (+0.23%) |
| Ethereum - leans higher over 7 days | 21 Aug 20:19Z | 28 Aug 20:19Z | 29 Aug 18:54Z | 55.7% | Correct (+0.09%) |
Both are hits, and both need a caveat: they are the same forecast written twice, 18 seconds apart, by a write path that double-fired on 21 August. They are two rows in the record but one opinion, and the defect is logged. No calls resolved against us this run.
Accuracy
The headline figure is the Brier Skill Score against a base-rate reference - the only measure on this page that reads the probability we actually publish. Across the 15 resolved seven-day asset calls at the current engine epoch it stands at +0.093: our probabilities score about 9% better than simply quoting the historical base rate and ignoring every signal we have.
That is a thin positive and it establishes nothing yet. Three things belong beside it:
- The sample is 15 against a 25-call threshold. The claim is not established and the validator returns insufficient data.
- We do not beat every naive alternative. Our Brier score is 0.249. A forecaster that said "up" every single time would have scored 0.212 on the same 15 calls - better than us. We beat the base-rate shadow; we do not beat a permanently bullish constant in a period that mostly went up.
- Reliability is 0.076 against resolution of 0.052. In plain terms: more of our score comes from our probabilities being mis-stated than from their ability to separate winners from losers. Calibration error sits at 0.236.
Directional accuracy over the same window is 67% (10 of 15), with a confidence interval spanning 42-85% - a diagnostic on direction-picking only, blind to how confident we said we were, and not a skill claim.
Standing book
Three-day Ethereum (to 31 Aug), seven-day Bitcoin (to 3 Sep), the new seven-day Solana lean, the choppy regime read (to 4 Sep), three longer-dated macro calls, and six standing structural and relative questions. Four of the seven asset-call slots are open; they get filled when something convicts on more than one signal, not on schedule.
