The scoring window is dominated by macro rather than crypto-native catalysts. A below-forecast US second-quarter growth print, a central-bank decision to leave policy unchanged with the dollar posting its steepest two-week decline, and publicly held US debt crossing 100 percent of output all land in the same day. Broad crypto-market coverage nets bearish across 27 scoped reads while equity-market coverage nets bullish. Bitcoin remains our deepest coverage and nets bullish on the day; Ethereum and XRP both net bearish. Elsewhere in the window, two independent items cover a privacy-coin shielded-pool migration and a large mining-capacity expansion on the same network, and an exchange listing pair pushes a newly traded derivatives token into the top-covered set.
No new seven-day coin leans were issued. All seven seven-day slots are standing and unchanged, and the three experimental three-day slots are full; nothing matured for scoring this run. The quarter-length relative lane is also at its concurrency cap: a fresh relative read on Binance Coin versus Bitcoin cleared the model's gate this run but there was no free slot for it, so it was not issued and no number for it is being published. One new read was added, a macro event read: with growth softening but no contraction signal in the window's coverage, our forward read puts the chance the US avoids a dated recession through the end of 2026 near 88 percent.
On the record: nothing resolved since the last run. The last call to mature was a three-day upside lean on XRP, issued 2026-07-27, which resolved incorrect on 2026-07-30. The three resolutions before it, all stamped 2026-07-29, also went against us: seven-day upside leans on XRP and Solana, both issued 2026-07-21 and matured 2026-07-28, and a three-day upside lean on Convex. The headline skill figure is negative: the Brier skill score against a base-rate reference sits at -0.46 across the five resolved calls from the current model, meaning our probabilities have scored worse than simply quoting the historical frequency would have. The decomposition says why: reliability is 0.17, which is where over-confidence shows up, and resolution is 0.00, meaning the probabilities are not yet discriminating between outcomes at all. The claim pool is five resolved calls against a minimum of twenty-five, so the verdict remains insufficient data - but the direction of the evidence is not flattering and we are not going to dress it up. A directional-accuracy diagnostic, which is blind to the probability we publish and is never our headline, reads -34 percent on seven-day calls and -15 percent on the experimental three-day calls. The long-horizon structural cards, several of which are plain historical frequencies rather than signal-driven reads, contribute nothing to this figure either way - the nearest of them does not resolve until mid-August.
