Forecasts
A forecast is worth something only if it was made before the answer was known. Each forecast here was sealed with Pith Seal before the measurement that settles it. When that measurement arrives, the forecast is revealed, checked against its seal, and scored beside the consensus forecast for the same question.
Forecasters
| Forecaster | Sealed | Revealed | Resolved | Hits |
|---|---|---|---|---|
| Recognition Science | 3 | 0 | 0 | none resolved yet |
Recognition Science: Each forecast comes from named theorems in the Recognition Science Lean library, which state the quantity and any measured input they use.
Benchmark. The consensus forecast: the latest published measurement or global fit for the same question, read by the rule frozen in each entry.
Seals
- Seal
- cac5b659f3163d19
- Sealed
- 2026-09-26 04:49:24 UTC
- Bitcoin
- confirmed in block 968,643
- Rules
- rs-forecasts-contract-v1.json, revealed from the seal
- SHA-256
- 5e14484ef9f3745bdc10396220cca7e8dc04355221c0c393308143e10cc28fc3
| Forecast | Question | Settled by | Status |
|---|---|---|---|
| RSF-0001 | Neutrino mass ordering | The first result that reports the ordering at 3 standard deviations or more. | sealed |
| RSF-0002 | Ratio of the atmospheric to the solar neutrino mass-squared splitting | The first result that reports both splittings with the ratio known to 0.5% or better. | sealed |
| RSF-0003 | Sum of the three neutrino masses, given the solar splitting | The first determination of the sum to 0.010 eV or better. | sealed |
A sealed forecast's value and reason stay private until its measurement arrives. The seal proves they were fixed no later than the times on its receipt.
The rules
Which questions. A quantity enters when a theorem pins its value, relation or sign; a measurement underway will report it more precisely than today; and the forecast differs from the consensus forecast in value or in width. Every quantity that passes enters, whatever outcome it is likely to have.
What counts as a derivation. The theorem has to state the quantity. A statement that some number lies in a range, proved by exhibiting a chosen number, does not count. Neither does a yes-or-no value defined to be true. Each forecast records the weakest assumption in its chain and how many candidate forms were examined before the one it states.
How a forecast is settled. By measurement. Retractions, reviews and consensus statements never settle a forecast. When a measurement depends on a model the forecaster disputes, the forecast says so in advance, and both readings are scored and published.
How it is scored. A point forecast hits when the reported value lies within two reported standard deviations of it. Both forecasters also get a log score at the reported value, so a sharp forecast that lands gains more than a broad one. The stated reason is scored too: right for the wrong reason counts as a miss on reason. An unresolved forecast is reported as unresolved and never counts as a hit.
The complete rules, byte for byte as sealed: rs-forecasts-contract-v1.json.
Not forecast, and why
These results from the same library did not meet the rules when the first seal was made. Some are already contradicted by measurement. They are listed so the record starts complete.
- fine-structure constant (Predictions.Registry.alpha_inv_row)
the registry formula is marked stale; the 0.6% remainder is assigned to an un-derived layer (D-alpha-remainder-to-t10-20260906). Enters when that layer is derived. - weak mixing angle and W/Z mass ratio (Registry.sin2thetaW_row, Registry.MW_MZ_ratio_row)
already compared with data: about 14 and 23 standard deviations from the measured values. Retrospective record, not a forward forecast. - primordial spectral index n_s = 1 - 2/phi^3 (Unification.CosmologicalPredictionsProved.spectral_index_formula)
evaluates to about 0.528 against measured 0.965 to 0.974. Already contradicted; flagged for repair. - muon g-2, R_K, gallium cross-section (Experimental.Anomalies)
existence statements witnessed by chosen numbers; inadmissible. - no WIMP signal, no sterile neutrinos (Experimental.DAMAModulation.no_wimp_expected, Experimental.MiniBooNELSND.rs_no_sterile)
Booleans defined true; inadmissible until derived. - charged-lepton and quark absolute masses, alpha_s, CKM (Registry Tier D rows)
no interval proof and, for alpha_s, no named scale.
Checking it yourself
Each seal's receipt carries its statement and proofs, which standard tools verify without Pith. When a forecast is revealed, its file, salt and path appear on the receipt, and anyone can recompute the seal's root from them. The rules file above is the seal's first file; its SHA-256 is the fingerprint shown beside it. The same data is at /api/v1/forecasts.