How the record stands
Our headline measure is the Brier Skill Score of the 7-day calls against the historical base rate they were issued against. Over the trailing 30 days to this run (30 September, engine epoch 4) it reads -0.046 on 12 scored calls: our stated probabilities have not yet beaten simply quoting the base rate. Twelve is well short of the 25 calls we require before making any skill claim, so the verdict remains "insufficient data", and a result this close to zero can move either way on one or two outcomes.
The Murphy split explains the shape: reliability 0.0185 (the gap between what we state and what happens) against resolution 0.0231 (how much our probabilities separate outcomes). Resolution that low means the 7-day probabilities are not yet discriminating much between calls that come true and calls that do not. As a directional-accuracy note only, not a skill measure: the 7-day calls picked the right direction 6 times in 12, about 1.1 points above the average base rate. A constant up-call would have picked 8 of those 12.
Receipts: resolved since the last run
All asset calls below were graded on direction alone. The quarter-long relative call was graded on whether XRP beat Bitcoin over its window.
| Call | Stated | Issued | Matured | Resolved | Outcome |
|---|---|---|---|---|---|
| Bitcoin - leans higher over the next day (experimental) | 56.9% | 29 Sep 02:57Z | 30 Sep 02:57Z | 30 Sep 03:03Z | Hit (+0.39%) |
| Solana - leans higher over the next day (experimental) | 55.7% | 29 Sep 02:57Z | 30 Sep 02:57Z | 30 Sep 03:03Z | Hit (+2.19%) |
| Pump.fun - leans higher over the next day (experimental) | 58.8% | 29 Sep 02:57Z | 30 Sep 02:57Z | 30 Sep 03:03Z | Hit (+24.05%) |
| Uniswap - leans higher over the next 3 days (experimental) | 60.9% | 27 Sep 02:25Z | 30 Sep 02:25Z | 30 Sep 03:03Z | Miss (-10.30%) |
| NEAR Protocol - leans higher over the next 3 days (experimental) | 62.9% | 27 Sep 02:25Z | 30 Sep 02:25Z | 30 Sep 03:03Z | Miss (-4.08%) |
| Zcash - leans higher over the next 3 days (experimental) | 61.8% | 27 Sep 02:25Z | 30 Sep 02:25Z | 30 Sep 03:03Z | Miss (-13.98%) |
| XRP lags Bitcoin over the quarter | 87.8% | 1 Jul 16:17Z | 29 Sep 16:17Z | 30 Sep 03:03Z | Miss (XRP +41.6% vs Bitcoin +39.4%) |
Three hits, four misses. The whole 3-day set resolved incorrect while the whole 1-day set resolved correct; with three calls each, that split is noise, not a finding about either horizon. Neither experimental horizon counts toward the public record. The XRP relative call was one of our most confident, and it resolved against us by about two points of relative performance.
New reads (experimental horizons, excluded from the public record)
The 7-day lane is full until two slots free up on 1 October, so today's new calls all sit in the short experimental lanes. Every number below comes from the ensemble engine, not from us, and is provisional and uncalibrated.
- NEAR Protocol, 3 days - leans higher, roughly 61% vs a ~48% base rate. Coverage is constructive but concentrated in one account.
- Pump.fun, 3 days - leans higher, roughly 60% vs a ~48% base rate. Analyst-only coverage; strong short-term momentum.
- Solana, 3 days - leans higher, about 58% vs a ~49% base rate. The broadest coverage of the set, including a note that one analytics firm's agent-payment transactions on Solana hit a weekly record.
- Avalanche, 1 day - leans higher, about 55% vs a ~46% base rate. Momentum carries this read; our coverage leg is near neutral.
- Bitcoin, 1 day - leans higher, about 56% vs a ~49% base rate. Wide coverage, constructive on balance, with cautious notes on nearby price structure and falling futures leverage.
- XRP, 1 day - leans higher at about 49%, below even odds but above its ~43% base rate.
Screened out on evidence rules rather than rank: SUI, CVX, CRV and LINK each ranked high enough to qualify, but each had all its directional coverage from a single account, so none was issued. Uniswap and Monero also qualified but ranked below the six issued; both short lanes are now full.
Regime and standing questions
No new regime row. Bitcoin closed the 29 Sep bar above its 20-day average, with a weekly move of about -2.95%, just inside the flat band. That borderline reading comes one day after the rotation call (which grades on breadth) went out, so the standing rotation row and the earlier risk-on row both keep running and will be graded as issued.
Coverage over the window was constructive on the majors and on DeFi and yield, from 48 analyzed items against 68 in the prior two days. The current window's second day is only a few hours old, so the counts are not comparable one-for-one. Two items came from a regulator account.
On the standing questions, the engine's fresh read on a new Bitcoin record differs materially from the older standing row, but that question already carries two standing calls, so a refresh has to wait for one to expire. The altseason and major-drawdown reads are essentially unchanged from their standing rows, so nothing was added there.
