Newest published call 2026-09-03 · derived from the graded close of 2026-09-02 · page generated 2026-09-04.
Since 2026-07-14 the published call has spent 32 of 47 days in Wait in Stables — no Bitcoin, no altcoins, all cash. Over that stretch Bitcoin went $62,127 to $77,189. This page is the accounting for that.
If you came here because the model looks broken, the fair version of that complaint is: it has been out of a rising market, and it cost you the rally. That is true, it is measured below, and it is not the part we would hide.
That is the Bitcoin move the model was not holding for. Compounding every day-to-day step from 2026-07-14 to 2026-09-03 on which the published call was Wait in Stables, Bitcoin moved +5.4%. It is forgone upside, in full, and nothing in this window offsets it — the steps in question were net up, not net down. The ledger is missing 6 days (2026-07-13, 2026-07-15, 2026-08-30, 2026-08-31, 2026-09-01, 2026-09-02), so a couple of those steps span two days rather than one.
Across that whole span the price went $62,127 to $77,189, +24.2% in 51 days. 32 of the 47 published calls in it were Wait in Stables; on the remainder the model was invested, and those steps came to +17.9%.
Read as an all-in / all-out rule — fully invested on every call that is not Wait in Stables, fully out on every call that is — following this ledger from 2026-07-10 would have returned +15.9% against +22.2% for doing nothing at all. That reading is deliberately crude: the invested phases are dollar-cost-averaging instructions rather than a switch, so take it as the shape of the gap and not as a portfolio result. The gap is real either way.
Wait in Stables is not re-decided each morning. It is a held state with an exit gate, and the gate — not a fresh daily opinion — is what you are looking at when the call repeats. The engine entered the current run on 2026-08-18; the site published it from 2026-08-20, after the display dwell that stops a one-day flicker becoming a headline. The ledger names how it was reached: de-risk from Hold BTC — risk-off, not the end of a cycle.
The arm of that gate which currently binds applies between 50% and 35% below the all-time high, in the era of US spot Bitcoin ETFs (from 2024-01-11). Inside that band the on-chain valuation measures read cheap while institutional selling can keep price grinding sideways for months, so "it looks cheap" is explicitly not sufficient on its own. Below 65% from the high the rule stops applying altogether and the model leaves this call automatically, because no cycle bottom in the record has formed above that depth.
A rally on its own IS an exit — above one specific level. This said the
opposite until 2026-08-28, and the code disagreed: cycleNavigator.ts:966,977
computes priceRecovering = priceVsATH > -35% and leaves WAIT when it is true,
with no score, buy-signal or regime condition attached. So price alone, above 35% below the
all-time high, ends this call. What a rally below that level does not do is end it early:
inside the band the gate needs the model actually buying, and that is the stickiness which
kept it defensive through the last drawdown instead of being talked out by a bounce. The
threshold is written down in advance so it cannot be argued with afterwards — in either
direction — and this page was wrong about it, which is exactly the failure the written
threshold exists to make visible.
This is a checkable claim rather than a posture. The sentence below is generated from the same constants the navigator evaluates at the gate, so this page cannot promise an unlock the engine will not honour.
The exit condition, as the engine holds it. This call is held, not re-decided daily. It ends on a recovery above ~$81,200 (+5.2% from here, 35% below the all-time high), or on the model turning buyer with the score above +0.30 — sustained five days, or with the regime no longer bearish. The score is −0.022 today. Neither is close, so this call will not change on a small move.
So there are two doors. Price can clear the recovery threshold, which is a fact about the tape and needs no interpretation from us. Or the model's own composite score can clear +0.30 — either sustained five consecutive days, or on a day when the regime classifier has stopped reading bearish. Outside that band the ordinary form of the gate is looser: an active buy signal alongside a score above +0.15. Check tomorrow's row on the ledger against those numbers yourself. If the condition is met and the call has not changed, this page is wrong, and that is exactly the kind of wrong we want to be caught being.