1 · Venue and counterparty
The largest losses in crypto funds were venue events, not signal failures. Where the money sits, against limits that shrink as the venue score falls.
2 · Book risk
What the book is exposed to, how few independent bets it really holds, and how much can be lost before positions can be reduced.
3 · Market state
The fragility index measures the size of the pile of dry material, not the timing of the spark: levered longs paying to hold, no protection bought, price at the highs, no new stablecoin money arriving.
4 · Active triggers
Pre-committed responses to configurations known to precede forced flow. Thresholds are frozen; hit rates come with their (small) sample sizes.
5 · Trade structures
Where durable return in crypto funds has come from: basis, funding carry, volatility premium, unlock shorts. Gross, cost, net, and the venue whose failure would matter most.
6 · Event strip · next four weeks
Scheduled supply and dates that move positioning: cliffs in days of real volume, governance, upgrades, large options expiries.
7 · Tier 2/3 screens and liquidity gate
Screens condition attention; they are not forecasts. Each carries its trailing information coefficient with a standard error of the same order.
8 · Exchange tokens
The one sector with equity-like cash flows: fee revenue, a shareholder yield from burns, an idiosyncratic tail (exchange solvency) that is orthogonal to BTC, and a market-priced solvency signal in the token itself. Ranking here is within one sector with real fundamentals. On-chain fees are verified; CEX revenue is estimated and flagged.
9 · Resistance / support tests
Φ is not a direction forecaster. The published quantity is a competing-risks cumulative incidence: the probability of the level breaking before it is rejected (holds), by horizon — "chop" is censoring, not an outcome, so it no longer flatters either side. Every number is recalibrated walk-forward (a model that says 95% must be right 95% of the time) and carries a block-bootstrap band; the headline uses the consensus of the three level definitions when the data say consensus predicts better. The crowding/Φ overlay stays preliminary on its short sample and never drives a published number.