After a record wipeout in crypto derivatives, Hyperliquid’s Jeff Yan says Binance and other CEXs report just one liquidation per second during cascades — potentially understating events by up to 100x. Bybit previously acknowledged API throttles and has since upgraded. Here’s the state of transparency and how to trade around the data gaps

Days after crypto derivatives suffered a historic flush, the industry is in a bruising debate over what actually happened. Hyperliquid cofounder Jeff Yan publicly alleged that centralized exchanges (CEXs) — including Binance — vastly underreport user liquidations during volatility, sometimes exposing just one liquidation per second even when thousands are occurring in the same moment. The claims lit up trading desks because the selloff already produced the largest single-day liquidation tally on record, estimated near $19B.

What Hyperliquid says is wrong with the data

In a thread on X, Yan argued that some CEXs “report only the latest liquidation in each 1000-millisecond window,” meaning market-wide liquidation heatmaps and venue feeds can miss the true size and pace of a cascade. He contrasted that approach with Hyperliquid’s fully on-chain model where “every order, trade, and liquidation happens on-chain” and can be permissionlessly verified. Coverage from industry outlets summarized his post and pointed to exchange documentation describing this one-per-second behavior. See: Crypto.news.

Separately, research highlighted by market analysts — and mirrored in exchange community posts — documented that Binance and Bybit switched their WebSocket streams in 2021 to emit at most one settlement per second (and that OKX limits one order per second per contract). Whatever the motivation, the net effect is that public feeds can show a gentler picture than the reality traders experience on venue. Reference: Binance Square / K33 note.

Bybit’s admission and upgrade

There is precedent for these concerns. In early 2025, Bybit CEO Ben Zhou publicly questioned widely cited liquidation totals and estimated the industry-wide figure was far above aggregator prints. Bybit subsequently made liquidation data more comprehensive via API and increased its update cadence (for example, from once per symbol per second to every 500ms), acknowledging that prior limits likely understated events during cascades. See: Cointelegraph.

Binance’s position — and the weekend context

As criticism mounted, Binance also faced heat for synthetic-asset depegs during the crash (reports cited USDe, wBETH, BnSOL) that allegedly triggered forced liquidations for some users. Over the weekend the exchange said it is reviewing affected accounts and announced a compensation plan that multiple outlets summarized at roughly $283M. Whatever one’s view of liquidation feeds, the company signaled it would make impacted accounts whole where platform issues contributed to losses. Example statement: Binance Square.

All this comes in the wake of a record liquidation day (~$19B) tied to an abrupt macro shock that swung risk assets lower and detonated crowded longs across venues. That total — already debated — is precisely why the quality of liquidation reporting matters: if feeds throttle, models that rely on those feeds may misjudge the speed and depth of deleveraging.

Why a one-per-second throttle matters in practice

Illusion of control: A feed that emits one liquidation per second during a 500-per-second cascade can create a lagging illusion of smaller stress, influencing risk controls, market making, and trader behavior.
Downstream distortion: Analytics dashboards (liquidation heatmaps, “squeeze” meters) often mirror exchange streams. If the upstream firehose is throttled, the downstream picture is necessarily blurry.
Cross-venue contagion: Underreported wipes on one large venue can still drive price on the composite order book; dashboards reading throttled feeds may attribute the break to “sentiment” rather than forced flows.

What transparency looks like (minimum bar)

Documented API cadence (and per-product exceptions) with public changelogs when throttles are tightened or relaxed.
Venue-level liquidation counters that publish aggregated per-interval totals even if per-event streams are rate-limited.
Post-mortem bundles after major cascades: instrument-level counts by hour, not just marketing statements. (Bybit’s upgrade and public framing is a useful reference point.)
Independent attestations (hashes, Merkle proofs, or third-party audits) showing that off-chain liquidation logs reconcile with internal ledgers.

How traders can navigate the data gap right now

Until standards improve, assume public liquidation numbers are a floor, not the whole story. Practical steps:

Triangulate sources: Compare aggregator prints with venue notices, research posts, and any available API counters (for example, K33 and CoinGlass notes on throttles).
Watch basis and funding, not just wipes: If perp funding dives and spot-perp basis compresses while reported liquidations stay oddly muted, the flow is likely bigger than the feed shows.
Stress-test leverage: Model scenarios where liquidations are 10–100× the visible stream during bursts — especially around macro catalysts — and size risk accordingly.
Prefer transparent rails for critical hedges: On-chain venues like Hyperliquid claim verifiable, constant-cadence logs; even if you trade CEX perps, hedge or sanity-check with DEX/on-chain data as a second opinion.

The bigger picture: after the record flush

The wipe was historically large. Venue recaps and community dashboards converge on a single-day record near $19B in liquidations after macro headlines ricocheted through markets.
Some data feeds are throttled by design. The one-per-second behavior is documented and materially affects real-time analytics during cascades.
Exchanges are responding — unevenly. Bybit raised its cadence and opened APIs after acknowledging undercounts; Binance has promised compensation related to depeg-linked liquidations and says it will strengthen risk controls, but has not publicly committed to changing feed throttles.

What to watch next

Binance’s post-incident report: Scope of compensation and any operational changes (risk engines, cross-margin halts, feed cadence).
Standard-setting moves: Whether major venues coalesce around interval counters or attestations so dashboards can reconcile bursts even if per-event streams stay rate-limited.
Regulatory heat: Whether supervisors add transparency of liquidation reporting to market-structure expectations.

Bottom line

The fight isn’t just Hyperliquid vs. Binance — it’s on-chain verifiability vs. off-chain opacity. Rate-limited liquidation streams make the biggest blowups look smaller in real time, and that can warp risk signals for everyone from retail to market makers. Bybit has shown one path forward by acknowledging past constraints and lifting its data cadence. Binance, for its part, is moving on user restitution after depeg-linked liquidations — the open question is whether it will also open up its liquidation telemetry. Until then, treat public liquidation prints as an underestimate by design, hedge with multiple data sources, and assume the next macro shock will again test both leverage — and transparency.