Skip to main content

We just pre-registered the test most likely to embarrass us

Swiss Whale Intelligence — Honest Research · 2026-06-09

Today we froze a public, tamper-proof test of our own most flattering claim — the one that, if it held, would retroactively justify everything we've built. We expect it to fail, or more likely to return a clean "we can't say." We are committing, in advance and on the record, to publish whichever answer the data gives. The answer arrives in roughly two years.

Here is exactly what we did, and why.

The claim on trial

The intuition every on-chain watcher wants to be true: when a long-dormant whale wakes up and moves old coins, it foreshadows a price drop. We track exactly these movements, so this is the claim our infrastructure most wants to validate. That is precisely why we are the wrong people to judge it after the fact — and precisely why we wrote the test down before looking.

Concretely, the registered hypothesis: a dormant wakeup — a single UTXO of ≥ 500 BTC, unmoved for ≥ 2 years, being spent — is followed by a meaningfully negative 7-day Bitcoin return, measured against a matched baseline. Pure on-chain trigger: coin size and coin age. No entity labels, no judgement calls, nothing we could quietly tune.

Why it can't be fudged

A pre-registration is only worth the hindsight it removes. So we removed it:

  • Frozen in git, commit fec427a, with an immutable timestamp. The hypothesis, the floor, the decision rule, and the stopping rule were all fixed before any forward data existed.
  • A frozen data cutoff (block 952966). Only events that happen after the cutoff count. We cannot reach back and cherry-pick history we've already seen.
  • An immutable instrument — a one-time snapshot of the 4,483 large coins eligible to "wake" (dormant_cohort_952966). It will never be expanded, so a more complete database two years from now cannot quietly inflate the result.
  • Three pre-committed verdicts — CONFIRMED, REFUTED, or INCONCLUSIVE — with the boundaries drawn in advance. The most likely outcome, "cannot say," is declared a first-class result, not a failure to spin.

The floor is set deliberately hard on ourselves: refuting the flattering claim is reachable; confirming it would require an effect so large it is almost certainly not real. The test is built to kill its own darling. Only overwhelming evidence saves it.

We've done this to ourselves before

This isn't our first self-inflicted null. For a year we ran on-chain whale signals as live trading rules on a real account. The win rate was 74.9% — three trades in four closed green — and the equity curve still fell to −96% before a single fee was charged, because the average loss was four times the average win. A 75% win rate that loses money is what an absent edge looks like dressed in a flattering metric. We published the curve, with a slider so you can watch fees deepen a hole the asymmetry already dug.

We also retired the more iconic version of today's test — "whales depositing to exchanges foreshadow drops" — on the record, because its event process turned out to be non-stationary: the large holders who once made those deposits are migrating to OTC desks and ETF custody, so the test could wait three years and still not separate "no effect" from "the population is gone." We chose the answerable hypothesis over the iconic one, and said why in public.

Why bother

In a field whose entire commercial gravity pulls toward overclaiming, the durable position is the one that still holds when the confidence interval is wide. We would rather be trusted than impressive. So we are not telling you which whale signals work. We are showing you, in advance and with the receipts, exactly how hard it is to establish that any of them do — and committing to the result before we know it.

If, in 2028, the verdict is "we can't say," that will be the honest headline, and we will run it.

The receipts. Full pre-registration (frozen at fec427a): prereg_dormant_wakeup_v2.md. The structural argument behind the choice — "The cleaner the signal, the less it's yours": essay_three_axes.md. The live trading exhibit with the fee slider: kraken_myth_buster.html. Descriptive research only — not financial advice.