Don’t trust. Verify.
Every public claim we make links to an artifact you can check yourself: the live walk-forward accuracy ledger, the before-the-cascade autopsy maps, and the per-venue feed reconciliation audits. No log-in, no cookies, no sales call. This page loads every figure live from the same public endpoints you can query directly — nothing on it is typed in by hand.
Machine-readable version for agents and press: /receipts.json · narrative version: /accuracy-vs · every figure is a descriptive measurement of past performance, never a forecast.
This week in liquidations
The rolling last-7-days cut of the public record, recomputed live from the same published artifacts as everything below — the cascade autopsies (our own multi-venue tape) and the level ledger. Every figure carries its count; nothing is estimated. Window shown in UTC.
across autopsied cascades
strictly-timestamped snapshots
on this week’s snapshotted autopsies
Loading the weekly cut from /receipts.json…
The measured track record
Walk-forward, out-of-sample grading of the published liquidation levels — every figure carries its sample size and a 95% confidence lower bound, including the symbols we switched off. Loaded live from /proof/stats; the full per-symbol ledger is on /proof.
all tracked symbols · weighted
BTC / ETH / SOL / XRP · weighted
conservative lower bound
the sample size behind the scorecard above
incl. auto-disabled shown honestly
out-of-sample measurement days
Loading live data from /proof/stats…
Proof-of-foresight: the autopsy library
After each major liquidation cascade we publish a frozen artifact: what the liquidation map showed before the cascade (a snapshot strictly timestamped ahead of the move, when one exists on disk), what then happened on our tape, and which of the published levels were reached — misses shown at equal weight, and artifacts without a pre-cascade snapshot labelled as tape records with no before-claim. The before-map timestamp on a snapshotted artifact is the part that cannot be faked retroactively. Full library: /autopsies.
Loading the autopsy index from /receipts.json…
The feed trust board
A liquidation map is only as honest as its tape. Every few minutes each venue’s ingested events are reconciled against an independent check, and we publish the verdicts with their sample sizes — including the bad days. Live source: /proof/feed_trust. As of the research date cited below, we found no other liquidation-data platform publishing an independent reconciliation of its own feed at all.
| Venue | Verdict | Match rate (n runs) | Coverage |
|---|---|---|---|
| Loading from /receipts.json… | |||
Method, disclosed per venue: OKX — direct REST liquidation-orders reconciliation; Bybit/Binance — open-interest-signature confidence-interval checks (no clean public REST liquidation endpoint exists); Deribit — expected-zero structural check. A descriptive audit of data quality, not a trading signal.
The honest comparison
What we publish, with the evidence link for each row, beside what is commonly available elsewhere in the category. The right-hand column reflects what the named platforms’ own public product pages and documentation state — we make no claim beyond what those materials show, and the quotes below are their own words about their own data.
| Artifact | Hunter Killer (us) | Category norm (sourced below) |
|---|---|---|
| Walk-forward accuracy scorecard | Published — /proof | Not published (checked Coinglass, Hyblock, Velo) |
| Sample size + confidence bound on every accuracy figure | Yes — n and a 95% confidence lower bound on every published figure (/proof/stats) | No public scorecard to carry them |
| Proof-of-foresight artifacts (before-map timestamps) | Published — /autopsies | None found |
| Independent reconciliation of the venue feeds | Published — /proof/feed_trust | Not published |
| Downloadable raw evidence file | Published — /proof/evidence | Not published |
| Cryptographic anchor chain over the ledger | Published — /proof/anchors | Not published |
| Frozen metric-definition contract | Published — /methodology | Not published |
| Machine-readable evidence index for agents + press | This page — /receipts.json | None found |
"The data is generated by simulating liquidation events based on various leverage multiples and open interest data."Coinglass FAQ / product documentation (publicly accessible, 2026-07-11 research)
"The data shown is an estimate derived from open interest data and is not an actual leaked order book."Hyblock disclaimer / product documentation (publicly accessible, 2026-07-11 research)
Competitor characterizations sourced from publicly accessible product pages and FAQs as researched 2026-07-11 (full citations on /accuracy-vs). Their products change; if any platform has since published any of these artifacts, please let us know and we will update this page promptly. Not affiliated with, endorsed by, or sponsored by any platform named. No claim beyond what those materials show.
How to check us in three steps
Query the live stats yourself
curl https://hunterkiller.io/proof/stats — the same JSON this
page renders. Every accuracy figure in it carries its sample size and a
95% confidence lower bound.
Check a before-timestamp
Open any artifact in /autopsies: its map snapshot is timestamped strictly before the cascade it anticipated. Compare the timestamp against the public tape of the event.
Recompute the anchor chain
/proof/anchors ships the hash chain plus the verification recipe. A mismatch means retroactive edits; a match means every row is exactly as written.
The whole point is that you don’t have to take our word
Every figure on this page came from a public endpoint you can query right now. The machine-readable index of all of it is /receipts.json.
All figures on this page are walk-forward, out-of-sample, loaded live from public endpoints — no hand-typed numbers · sample size and a 95% confidence lower bound shown with every accuracy figure · auto-disabled symbols counted and disclosed · no comparison-to-chance framing anywhere · no predicted-profit claims · descriptive measurement of past performance, never a forecast, never financial advice · competitor characterizations from their own public materials only, as researched 2026-07-11 · full ledger · methodology · narrative comparison