The base-rate panel, explained

Every regime state shows a “what followed” panel: forward BTC returns at 30, 90, and 180 days, drawn from prior periods when the market was confirmed in that same state. This describes what happened before, in this classifier’s own history - it is not a prediction of what happens next.

N by periods, not N by days

The panel’s primary count, N, is the number of distinct confirmed periods in a state - each one counted once, no matter how many days it lasted. A 90-day period and a 3-day period each contribute exactly one outcome to N, not one-outcome-per-day.

Alongside that primary count, the panel also shows a secondary, much larger number: the count of individual days whose confirmed state matched, each treated as its own forward-return sample. This by-day view has far more data points, but consecutive days within the same period are not independent observations - they’re the same market condition sampled repeatedly, day after day. Blending that autocorrelated day-count into the primary N would make the sample look far more independent than it really is, so the two are always shown separately, never combined into one number.

Why the window starts August 2020

Comparisons are limited to the institutional-era market, from August 2020 onward - not this classifier’s full available history. This is a deliberate scope choice, not a data-availability limit: some of this classifier’s own inputs (MVRV, flows) have real history reaching back to 2010-2012, but the market those early years describe - pre-ETF, largely pre-institutional, far thinner in both liquidity and participants - behaves differently enough from today’s market that treating it as an equally-relevant analog would overstate how much those early periods can tell you about current conditions.

This window choice applies only to base-rate comparisons, never to a metric’s own percentile. Every panel carries this exact distinction, verbatim, as its own window note:

Comparisons use the institutional-era market from August 2020 onward; a metric’s own percentile-in-history uses full available history.

A metric’s percentile answers “how does today’s reading compare to everything this metric has ever done”; the base-rate panel answers “what happened after periods like this one, in the market era most comparable to today’s.”

The small-sample rule

With fewer than 5 prior periods, the panel is marked small-sample rather than presented as if it carries the same weight as a larger comparison. The panel is never suppressed or hidden for having too few periods - even a state with zero prior periods still shows a real panel, explicitly marked as having no prior periods to compare against, rather than disappearing or silently showing nothing. What’s shown is always honest about how much evidence sits behind it.

What this panel is conditioned on

Every base-rate comparison is conditioned on the core state only - valuation, flows, leverage, and halving position, the same four input families that determine state_id. Sentiment and any short-history modifier are excluded from the match key entirely. Two periods land in the same base-rate bucket because their core state matched, not because sentiment or any other adjacent signal also agreed.

A note on survivorship

This panel’s sample isn’t cross-asset, so it doesn’t carry the classic survivorship bias of a universe where failed assets quietly disappear from the record - bitcoin has neither gone to zero nor been delisted anywhere in this window, so every period that occurred is represented. The honest caveat runs the other way: the entire institutional-era window was reconstructed after the fact, by running today’s pinned classifier rules back over historical data - it was not computed and observed live, day by day, as those market conditions actually unfolded. And a period only enters the sample once it has actually finished (exited) and had enough time to reach its forward return horizons; the currently open, still-ongoing period is deliberately excluded from N until it resolves, rather than folded in with an incomplete outcome.

Where this came from

This panel generalizes an earlier, narrower idea: comparing similar-depth price drawdowns against what followed them. The base-rate panel keeps that same “what happened after conditions like this one” logic, but widens the match key from “similar drawdown depth” to “same confirmed core regime state” - a broader, more structured version of the same underlying question.