Funding stress, leverage build-up, stablecoin flows and cross-asset coupling across the top crypto markets — scored against what actually happened next, with the detectors that failed named rather than quietly dropped.
Crypto moves on flow, funding and leverage as much as on technical levels — and those show up in the plumbing before they show up in a candle. Fahali watches the plumbing: what it costs to hold the position, how much borrowed money is stacked behind it, where the stablecoins are going, and which other instruments have quietly started moving as one.
Both, clearly separated. The response labels a proxy as a proxy and never returns it as a measurement.
Funding-rate dynamics and perpetual premium divergence read from exchange feeds — the cost of holding the crowded side, before the crowded side is forced out of it.
Open-interest acceleration and leverage concentration, the conditions under which a move stops being orderly and starts being a cascade.
Mint, burn and exchange inflow/outflow patterns — where the dry powder is moving, or leaving.
Tail dependence measured pair by pair, so you can see when instruments that normally trade apart have started behaving as one bet.
Not all of the things Fahali computes, and the honest list is short enough to print. The microstructure-absorption proxy is inferred from public aggregate trades and order-book behavior; it does not identify participants or measure hidden executed volume. That proxy, along with volatility expansion and compression and the order-book imbalance reads, has been measured over large samples as no better than its own base rate on a forward question. All of them are gated out of making standalone published claims and kept as sensors feeding the committee read.
They are still computed and still visible; they are simply not sold to you as forecasts, because the record says they have not earned it. The four capabilities above are the lanes that make judged claims, and the judged record covers those.
Pull the record. app.fahaliai.com/api/track-record/lead-time is public and needs no key: one row per stratum, with the effective sample, the signal-conditioned base rate, precision, recall and the misses. Because every claim is registered before its outcome exists, none of it can be backfilled — which is the only reason a track record in this market is worth reading at all.
Read-only by design. Fahali does not route orders, manage capital, hold funds, or provide buy/sell recommendations. It surfaces structured observation so you and your systems can decide. Past outcomes are published including the misses; future outcomes are uncertain.
Liquid crypto pairs alongside US equities and ETFs — the judged record spans all of them, not crypto alone. Coverage per asset class, and how much of each universe was scanned on the last cycle, is reported by the platform rather than inferred from how many alerts happened to fire.
No. It is read-only by design: no order routing, no custody, no path to capital. It observes market structure and reports what it sees. Every execution decision is yours.
Funding rates, open interest, order-book depth and volume are read from exchange feeds. Anything derived rather than observed is labelled as a proxy in the response — the microstructure-absorption read, for instance, is inferred from public aggregate trades and order-book behavior and does not identify participants or measure hidden executed volume.
Because they are scored on claims registered before the outcome existed, which cannot be backfilled. The judged record publishes, per stratum, how often a call was right, how often the outcome was already going that way anyway, and how much of the qualifying population was caught — with the misses kept in. Detectors that have not earned a forward claim are gated out of making one rather than left in the brochure.
Nothing here is gated. Open today's read, or pull the judged record and check it.