Predictions we freeze in public — including the misses.
LEXUN publishes forecasts here before their outcomes are known — each dated, frozen with a named resolution source and criterion, then scored with the Brier rule when reality answers. Misses are published permanently. There is no fixed cadence: a cohort is issued when its questions are written and sourced, and the cohorts so far are listed below from the ledger.
EDITORIAL FORECAST REGISTER operator judgement about public statistics — scored here, and not evidence for the decision models, whose calibration is kept separately on the Accuracy Centre.
The register opened on 14 August 2026 with its first ten forecasts; four of those were withdrawn the same day, with the reason kept. Everything below renders live from the ledger itself — counts, states and outcomes are read, not typed — and no accuracy figure will appear until a domain reaches ten resolved outcomes. Nothing here was backfilled: each entry was written, validated and sealed before its resolution date, and the receipt beside it seals the question, the band, the criterion and the date together. They are judgement forecasts about public statistics, scored on their own scoreboard — being right about the Bank of England would say nothing about whether the cash-runway model works, and the Accuracy Centre keeps the two apart.
Whose forecasts these are. Every entry names its forecaster and the board it is scored on, and scores are never pooled across boards. The current cohort is operator judgement about public statistics, on the editorial board. None of LEXUN’s three decision models forecasts CPI, Bank Rate, unemployment or insolvency counts, so nothing here is evidence that any of them is calibrated — and a future table showing a good editorial Brier score must never be read as if it were.
Live view (same ledger, re-read in your browser)…
Download the register: forecast-ledger.json (the record itself, append-only) or forecast-ledger.csv (one row per forecast, written from the JSON at every release). Cite it via the citation page.
The rules every entry follows
- Frozen before the event, with a visible publish date — never edited afterwards.
- A named public resolution source (e.g. ONS release, Bank of England decision) and an unambiguous criterion, fixed at publish time.
- A probability, not a hedge — scored with the Brier rule on resolution.
- Misses stay on this page permanently. Deleting a miss would end this product's reason to exist.
- Withdrawal is bounded: within 24 hours of issue, only for a defect in the forecast's own construction, never after. A withdrawn entry stays in the file as void, is never scored as a hit or a miss, and is never replaced silently — a replacement names what it supersedes. The release gate enforces the window.
Withdrawal policy
- When. Only within 24 hours of issue. After that a forecast stands and is scored, hit or miss; the release gate refuses a later withdrawal.
- Why. Only for a defect in the forecast’s own construction: reasoning that was wrong on the evidence available at issue, or a criterion or due date that makes it unresolvable as written. Never because the outcome looks likely to go against it.
- The reason. Written into the entry at the moment of withdrawal and kept permanently, with the time.
- Supersession. A replacement is a new entry with its own number and receipt that names the entry it supersedes; the withdrawn entry’s page links to its successor.
- Permanence. The withdrawn entry keeps its sealed receipt, its original band, its timestamps and its permanent page.
- Scoring. A withdrawn entry is void: it is never scored, as a hit or as a miss, and it is never counted towards a resolved total.
Why forecasts are withdrawn, and why that is not editing
Four of the twelve forecasts on this register have been withdrawn, each on the day it was issued. The four withdrawn on 14 August 2026 each say so in their own row with the reason on record: two because the reasoning behind the band was wrong on the evidence then available (the July 2026 MPC minutes), two because the due date fell before or exactly at the instant their source publishes, which made them unresolvable as written. Each was superseded by a new entry that names the one it replaced, and the corrections log records that the first wording of two reasons — “withdrawn before publication” — was itself wrong and was corrected. A sceptic can read that as a do-over; the register is built so that the reading is checkable rather than taken on trust: the withdrawn rows keep their sealed receipts, their original bands and their timestamps, they count as void rather than vanish, and after the 24-hour window a forecast cannot be withdrawn at all — it stands and is scored, hit or miss.
Can “frozen” be checked from outside? Partly, today. Ledger fingerprint sha256:00103d84783d255a0fb875e736b88f4fb053a848a606efe081903a2a8e419172 (anchor dated 2026-11-05, 12 entries); not yet anchored outside this site — the operator's next step is a public git commit or an OpenTimestamps proof of this hash, and this sentence changes when it is done. The chain of fingerprints is at /ledger-anchors.json: each names the one before it, so a rewritten file would break the chain; to check the current one, take the SHA-256 of the bytes of /forecast-ledger.json and compare.
These editorial forecasts are separate from the three release-gated decision models and are labelled as such — they demonstrate scoring discipline, not model output.
Why publish predictions at all?
Anyone can claim to be well calibrated after the fact, because memory quietly edits the near-misses into hits. The only defence is a record written before the answer is known and left alone afterwards. Publishing dated predictions in frozen cohorts — 12 so far, ten issued on 14 August 2026 and two issued on 30 August 2026 — with the resolution source fixed in advance, converts a claim about judgement into something a stranger can check. It also imposes a cost: a bad call is visible permanently, and cannot be deleted or reframed.
How Brier scoring works
The Brier score is the squared difference between the probability given and what actually happened, where the outcome counts as 1 if the event occurred and 0 if it did not. Saying 80% about something that happens scores 0.04; saying 80% about something that does not scores 0.64. Lower is better, 0 is perfect, and 0.25 is what you get by saying 50% to everything. Because the penalty grows with the square of the error, the rule punishes confident wrongness far harder than honest uncertainty — which is exactly the incentive a forecasting record should create.
How to check a forecast’s receipt
Each entry’s receipt is a SHA-256 over the eleven fields frozen when it was issued: id, question, issued_at, data_cutoff, probability, criterion, resolution_source, due_date, model, seed and inputs_hash, written as JSON with the keys sorted and no spaces. The outcome is added later and sits outside the seal. To check an entry, download forecast-ledger.json and run this with Python 3, the entry’s id at the end:
python3 -c "import json,hashlib,sys;e=[x for x in json.load(open('forecast-ledger.json',encoding='utf-8'))['forecasts'] if x['id']==sys.argv[1]][0];k=('id','question','issued_at','data_cutoff','probability','criterion','resolution_source','due_date','model','seed','inputs_hash');print(hashlib.sha256(json.dumps({x:e.get(x) for x in k},sort_keys=True,separators=(',',':')).encode()).hexdigest())" lxf_external_010
The value printed should equal the entry’s receipt. It proves the frozen fields have not changed since they were sealed; it does not prove who wrote them or when. The verifier checks the receipts the workspace produces, which are a different format.
What would make this page dishonest
Three things the register is built to prevent: editing a probability once published, quietly dropping an entry that resolved badly, and replacing a forecast without saying so. A fourth — choosing questions that are easy to get right — cannot be prevented by construction; it can only be seen, which is why every question, band and criterion is published in full. Each entry is frozen with its publish date and its resolution criterion at the moment it goes up. If a criterion turns out to be ambiguous, the entry resolves as disputed: the ambiguity is recorded, the entry is not scored, and it is never resolved under a more flattering reading.