Every prediction we publish in The Pattern Brief carries two things most forecasting doesn't: a stated confidence level, written down before the outcome is known, and a verification date, after which we go back and grade it in public. We've done this quietly since our first edition. Starting today, we're showing the work.

Back in Edition 018, published as new jobs numbers were landing in June, we made a specific, falsifiable claim: the Federal Open Market Committee would leave interest rates unchanged at its June 16-17, 2026 meeting. We stated that at 90% confidence — high, but not certain, because a new Fed chair's first meeting carried genuine uncertainty about direction.

The FOMC held its target rate at 3.50-3.75%, by a 12-0 unanimous vote, on June 17, 2026. The call was correct.

That's one resolved prediction. We currently have 51 more sitting open across 38 editions, with dated verification points running from this September through 2028. Ten of those resolve on September 30, 2026 — a real batch, not a cherry-picked one, covering transformer lead times, Florida condo price divergence, a Tennessee sewer moratorium, and algorithmic rent-pricing legislation, among others.

Most economic and housing commentary never gets graded. A newsletter, a talking head, or an analyst can make ten confident calls, get eight wrong, and simply not mention the eight again. Nothing forces the accounting. We looked at what actually builds trust in newsletters that have grown past where we are — chart-forward data journalism, named authorities with track records, proprietary research behind a paywall — and the one thing genuinely nobody in this specific category does is publish a real-time, resolved, public accuracy record.

That's not a marketing decision. It's the same discipline this publication is built to apply to everyone else's numbers, pointed back at our own. If we're going to call out a builder's price comp, an insurer's non-renewal math, or a Fed governor's inflation forecast, we don't get to exempt our own predictions from the same scrutiny.

One correct call is not a track record. It's a starting point, and treating it as more than that would be exactly the kind of number-dressing we'd flag in anyone else's work. A single 90%-confidence hit tells you almost nothing about whether we're well-calibrated — a coin that comes up heads once isn't proven fair or weighted.

The real test starts in thirty days, when ten predictions resolve at once. That sample is still small by any statistical standard, but it's the first point where a hit rate starts to mean something more than a coin flip. We'll report it plainly either way: if we go 6-for-10, we'll say 6-for-10, not round up or bury the four misses in a footnote.

Every tracked prediction gets logged at publication with four fixed fields: the exact claim, a stated confidence percentage, a resolution date, and the evidence source we'll check it against. None of those four fields can be edited after the fact. When the resolution date arrives, we check the claim against primary sources — a Federal Reserve press release, a Census Bureau report, a state regulatory filing — not against a summary or a secondary writeup, and we log the source alongside the verdict.

A claim is scored HIT, MISS, or PARTIAL if the real-world outcome only partially matches the stated terms. We do not delete or quietly revise predictions that look likely to miss as their date approaches. The list you'll be able to see is the same list that existed the day each prediction published.

This matters more than it sounds like it should. The easiest way to fake a good track record is to quietly retire the calls that were clearly heading toward a miss and only ever discuss the ones that landed. We're building the scorecard specifically so that shortcut isn't available to us — every open prediction stays visible and dated whether or not it's trending toward looking smart.

If you're a new reader deciding whether to trust this newsletter: don't take our word for anything — check the scorecard yourself once it's live, and judge us on the graded record, not the confident tone of any single edition.

If you've been reading since the early editions: the 51 open predictions include several from your first months as a subscriber. Several verification dates are coming up over the next year, and you'll see exactly how those specific calls held up.

If you're skeptical of forecasting in general: you should be. Most of it is unaccountable by design. The point of this exercise is to make ours the opposite — accountable by design, including when that's uncomfortable.

If you write about housing, economics, or real estate yourself: we'd genuinely like to see more publications do this. A graded record is a harder standard to hold yourself to, but it's the only kind of credibility that can't be faked with confident writing alone.

Confidence is cheap. Everyone writing about the economy sounds sure of themselves — that's table stakes for getting read. What's rare is writing down a number before you know the answer and then publishing the answer either way. That gap between "sounding right" and "being provably right or wrong" is where most economic commentary quietly lives, and it's the gap we're trying to close in our own work first.

The 90-day marker (tracked)

By September 30, 2026, all ten predictions scheduled to resolve on that date will be graded and published within 72 hours of their verification date — testing whether we can actually run this process on a real deadline, not just describe it in the abstract.

Verification date: October 3, 2026 · Status: OPEN · Stated confidence: 85%

Every prediction in The Pattern Brief carries a verification date and is revisited in a future edition — including this one.