What are confidence buckets (the W-framework), and why are there gaps between them?
Confidence buckets are four discrete weights — Speculative (0.15–0.25), Uncertain (0.30–0.40), Credible (0.50–0.60), and Strong (0.70–0.80) — that set how much probability the dual-scenario valuation puts on a company’s development case, with deliberate forbidden gaps between them so that a conviction level has to be defended rather than fudged to a comfortable middle.
The valuation runs two scenarios for each name, and one number decides how much weight the more forward-looking development case carries: the Scenario B probability weight, W. W does not float freely on a 0-to-1 dial — it must land in one of four discrete buckets, each tied to a category of evidence: Strong (0.70–0.80) — financing closed, permits in hand, construction started; Credible (0.50–0.60) — feasibility study complete, financing visible, no major blockers; Uncertain (0.30–0.40) — PEA stage, permitting unclear, financing unconfirmed; Speculative (0.15–0.25) — scoping only, with significant technical or regulatory risk.
The interesting part is what sits between the buckets. The ranges 0.26–0.29, 0.41–0.49, and 0.61–0.69 are forbidden — as is anything below 0.15 or above 0.80. A name cannot be assigned 0.45: if the evidence seems to sit on a boundary, the rule forces a definitive call — resolve down to the ceiling of the lower bucket or up to the floor of the higher one, and document why.
That prohibition is the whole point, and it is a deliberate epistemic constraint, not a stylistic one. A continuous dial invites false precision — a 0.45 looks rigorous but is really a way of declining to decide. The gaps remove the comfortable middle and make conviction earn its place: to claim Credible over Uncertain, the financing has to actually be visible, not “probably fine.” Discrete buckets turn a vibe into a defended position, and make two analysts — or the same analyst two months apart — far likelier to land in the same place for the same reasons.
The bucket structure — these thresholds, these gaps, this rule — is public, and so is the two-scenario architecture it feeds. What sits behind the login is the part that does the work for any specific name: which bucket a given company is assigned this month, and the evidence ladder that put it there. The thresholds are the ruler; the per-ticker reading off that ruler is the paid product.
Worked example
- VGZ — a live developer whose public profile — the nine category scores and the Re-Rate Score — is free to inspect; the page deliberately does not show its W-bucket or the evidence ladder behind it (the paid layer). The mechanism, in the general case: a developer with a completed study but financing only discussed cannot be split to 0.45 — it resolves down to the 0.40 Uncertain ceiling until financing becomes visible, then earns the 0.50 Credible floor. The jump from 0.40 to 0.50 is exactly where a real piece of evidence changes hands.