coinglass alternative
two liquidation trackers can watch the same cascade and print different totals. the comparison worth making is not the dashboard, it is what each one publishes about the prints it cannot see. this page states the constraint every liquidation feed sits under, says where it bites us, and lists five things you can check on this site — with the two sources it rests on linked under the table.
the constraint every liquidation tracker sits under
binance's all-market force-order stream is the feed most liquidation dashboards are built on, and binance documents what it drops: "For each symbol，only the latest one liquidation order within 1000ms will be pushed as the snapshot. If no liquidation happens in the interval of 1000ms, no stream will be pushed." — latest, not largest. whichever print the feed happens to sample at the end of that second is the one that survives, a bigger print inside the same second is gone, and nothing in the data says it was ever there.
the throttle sits on the exchange's own feed, not on any one tracker, so it truncates our tape exactly the way it truncates anybody else's. every liquidation total on this site is a lower bound for that reason, the page that prints the total says so, and no tracker reading the same public stream — this one included — can claim otherwise.
coinglass flags push-frequency limits on its own liquidations page: there is a notice headed "About Liquidation Order Push Frequency Limits" with a link to their explainer. the explainer body did not render when that page was opened on 2026-09-22, so the heading and the link are all this page will point at — we are not going to paraphrase text we did not read.
that is the state of it. two trackers print different totals because they read different venue sets over different windows through feeds that thin under load, and none of them is reporting the market.
what each side publishes about its own numbers
one column is filled from this site's own pages and code. the other is filled only where a dated source we opened says something, which is one row of the five.
| what to check | liqsiren | coinglass |
|---|---|---|
| does the page say the exchange feed drops liquidations? | yes — /liquidations-today/ quotes binance's stream documentation on the page and calls its own total a lower bound | yes — the liquidations page carries a notice headed "About Liquidation Order Push Frequency Limits", linking its own explainer (checked 2026-09-22) |
| is the venue list published, with what each feed can and cannot carry? | yes — binance, hyperliquid, okx and bybit, named on /liquidations-today/, with the 2026-09-23 start date okx and bybit share, bybit's own line saying why it starts that late, and binance's reduction of coin-margined contracts to the base coin stated beside them | not checked here |
| are reconstructed rows flagged rather than counted silently? | yes — hyperliquid publishes no liquidation stream, so those rows are rebuilt from the public trades feed and carry an estimate flag in the database column they are stored in | not checked here |
| is there a public record of calls made before their outcomes were known? | yes — /audit/ writes down every cluster call before it resolves, resolves it against this site's own tape by a fixed rule 24 hours later, and prints the misses in the same table as the hits | not checked here |
| do the liquidation-price numbers come from the venues' own maintenance-margin brackets? | yes — brackets are fetched from binance's and bybit's own endpoints, and each symbol carries a provenance record saying whether it was refreshed automatically or checked by hand, with the date | not checked here |
"not checked here" is a statement about our checking, not about their product: it means no dated source we opened settles the row either way. we did not install, sign into or measure anything to fill this table, and a blank we guessed at would be worth less than a blank we admit.
sources: binance docs — all-market liquidation order streams (checked 2026-09-22) · coinglass liquidations page — "About Liquidation Order Push Frequency Limits" notice (heading only) (checked 2026-09-22)
what this page does not claim
it does not claim anyone undercounts by an amount. measuring that would need both raw tapes over the same window and the same venue set, and we have one of the two. no number on this page compares one tracker's coverage to another's, because no such measurement exists here.
it does not claim a feature is missing from another product. four of the five rows above say we did not check, and that is what they mean.
it does not claim completeness for our own tape either. there is no row count on this page and no coverage percentage, because neither has been measured. what is published is what was recorded.
it does not claim okx history it does not have. the okx collector was deployed on 2026-09-23 and is recording, so okx prints are in totals from that date and from no earlier one — there is no okx history behind it and none was backfilled. okx's liquidation endpoint answers — it was probed on 2026-09-23 — so the short history is ours, not theirs, and it is written here rather than left for someone to find.
what is on this site
the tape is recorded here rather than bought: binance's all-market force-order stream, bybit's per-symbol liquidation topics, and hyperliquid rebuilt from public trades. the rolling 24-hour total, the hourly shape, the long/short split and the per-venue breakdown are on /liquidations-today/, with the caveats attached to the number rather than filed on a separate page.
/audit/ is the part that is hard to copy. the cluster model writes its calls down before the outcome exists, a fixed rule resolves each one against this site's own liquidation tape 24 hours later, and the misses are printed next to the hits. a hit rate whose denominator and method are not published is a brochure; that is why the denominator is on the page.
the calculators run on the venues' own maintenance-margin brackets, not on a flat rate — liquidation price, position size and funding, per coin and per venue, with the estimate disclosure on every figure because the exchange computes the real one from live margin.
the alerts are the point of the site: telegram, web push or email when price closes on your liquidation, on funding, or on position size. two active liq-distance alerts are free with no signup; unlimited is $9/mo.
where to check the claims above
the rolling 24h tally and its venue list · the scorecard, misses included · where the clusters sit · arm an alert on your own position
faq
is there a free alternative to coinglass?
this site is free to read and free to arm two liq-distance alerts on, with no signup: a rolling 24-hour liquidation tally with its venue list and caveats attached, per-coin liquidation heatmaps, and liquidation-price calculators built on the venues' own maintenance-margin brackets. unlimited alerts are $9/mo.
does this site record more liquidations than coinglass?
we do not know, and this page will not say. comparing coverage honestly would need both raw tapes over the same window and the same venue set; we have ours and not theirs. what we can tell you is which venues are in our number and why it is a floor rather than a census.
why do two liquidation trackers print different totals for the same hour?
three reasons before anything else: different venue sets, different windows, and a feed that thins under load. binance's public stream pushes at most one liquidation order per symbol per 1000ms, and the one it keeps is the latest, not largest. any total drawn from that feed is a lower bound — ours included.
are liquidation totals from any tracker the real market figure?
no. every public liquidation figure is built from exchange feeds that throttle themselves, over whichever venues the tracker happens to listen to. treat the number as a floor under the day and read the venue list next to it before comparing it with anything.
which venues are in this site's liquidation tape?
binance, hyperliquid, okx and bybit. binance's stream reduces coin-margined contracts to the base coin; okx has been recording since 2026-09-23 and has no history before that date, and its prints are bankruptcy-priced because that feed publishes no fill price; bybit has been recording since 2026-09-23 too: the collector read bybit's allLiquidation frames against the wrong schema until then, so every earlier frame was dropped and there is no bybit history before that date to backfill. hyperliquid publishes no liquidation stream at all, so those rows are reconstructed from public trades and flagged as estimates.
what does this site have that a liquidation chart does not?
a published scorecard and an alarm. /audit/ logs each cluster call before its outcome exists and resolves it against this site's own tape by a fixed rule, misses included. the alerts fire on your own position's distance to liquidation rather than on someone else's, over telegram, web push or email.