REK-T reads the positions of the biggest Hyperliquid accounts every 5 minutes and lays them out. This guide explains every number on the site: where it comes from, how it is computed and how to read it. Nothing here is investment advice.
Hyperliquid is a perpetual futures exchange whose positions are recorded on-chain: every position of every account is public and verifiable. REK-T only uses the Hyperliquid public API.
Whale data can therefore be up to 5 minutes old. The scan time is shown at the top of the panel, and a ⚠ appears if the last scan is more than 12 minutes old. Every address links to Hypurrscan so you can check for yourself.
The sum of the positions of the 450 tracked accounts on the selected coin. Longs on the left, shorts on the right: total size in coin, number of accounts and dollar value (notional, i.e. size × price). The bar and percentages compare sizes.
Why isn’t it 50/50? Across the whole market there are always exactly as many longs as shorts. What matters here is which side the biggest accounts pick: if they are 70% long, the rest of the market (small accounts, market makers) is mostly short against them.
Read with care: one huge account can swing the ratio, and some positions hedge exposure held elsewhere that we cannot see.
Each whale is classified from its public Hyperliquid history, recomputed every 6 hours:
The two bars show the profitable side and the losing side, weighted by position size. The table lists profitable whales positioned on the coin, sorted by return.
How to read it: a gap between the two camps (profitable whales short, losing whales long) says more than the overall ratio. But having won in the past guarantees nothing about the next trade.
Compares the latest scan with an older one (the previous scan, or ~1h, ~4h or ~24h ago) and lists every account whose position changed by more than $1M. Before and After give the size in coin; Δ $ gives the change in dollars at the current price.
The server keeps 26 hours of history: the 24h comparison fills up after a day of running. A closed position may also be a liquidation, in which case it shows up in the Rekt panel.
Positions under $50k are hidden: they are often leftovers with absurd liquidation prices.
Between two scans, the server spots positions above $250k that disappeared or shrank while price touched their liquidation level (checked with 1-minute candles). It then reads the account’s fills: only those Hyperliquid marks as liquidations are kept. A manual close or a stop does not count.
Cascade pages: for every BTC cascade (defined in the next section), a page is published with who got liquidated and in what order, how whales and smart money were positioned before and after, the whales that held on by adding margin, the surrounding candles and the score shown at the start. It updates for 3 hours. Only the 450 tracked accounts are covered, not the whole market.
A cascade, here, is a 1-hour BTC candle whose range (high − low) exceeds 1.75% with volume at least 3 times the median of the previous 48 hours. Over 200 days (April to November 2026), 56 candles match this definition.
The score counts how many of these 5 conditions are true right now:
Each score maps to the historical frequency of a cascade within the next 12 hours: 0 → 2.4%, 1 → 6.7%, 2 → 10.8%, 3 → 14.3%, 4 → 21.2%, 5 → 29.9%. The average across all scores is 13.7%: the “× average” figure compares against it.
Hot window: half of all historical cascades started between 12:00 and 16:00 UTC, around the US market open. Bias: after a move of more than 1.3% over 24 hours, cascades tend to extend the trend.
Even at 5/5, nothing happens 7 times out of 10: it is a statistic, not a signal.
To check the score instead of trusting it: at the start of each hour, the score is frozen using completed candles. Twelve hours later, we check whether a cascade started in that window. Verdict: cascade, or nothing.
Reading the chart: for each score, the orange bar is the observed frequency and the black line the predicted probability. A reliable score has bars landing close to the lines, and bars that grow with the score.
Brier score: mean squared error between the predicted probability and the outcome (0 = perfect). It is compared with a naive forecast that always predicts the average frequency: if the score does worse, it adds nothing.
48h strip: one cell per hour, darker when the score is high; a red dot marks a cascade hour. The 12-hour windows overlap, so a single cascade validates several hours in a row: hours are not independent tests. The full history can be downloaded as CSV.
Liquidation prices of whale positions within 35% of the current price (the 24 largest), with their size. Red above: shorts, liquidated if price rises. Green below: longs, liquidated if price falls. The dashed line is the current price.
A liquidation forces a market buy (for a short) or a market sell (for a long): a large cluster close to price can speed up a move if it gets hit. These levels move if the trader adds margin, reduces the position or, with cross margin, if their other positions win or lose. A blue outline marks a position defended in the last 7 days.
The question a stop hunter asks: how much do you have to sell (or buy) to trigger a cascade, and how far would it go? On most exchanges nobody knows the real liquidation prices: paid “liquidation maps” estimate them from assumed leverage. On Hyperliquid they are public. The simulator combines them with the real order book.
Domino track: The chart at the top of the panel maps the ground within ±20% of price. Each bar is a cluster of whale liquidations: red below price are longs (liquidated if price drops); green above are shorts (liquidated if it rises). Taller bars mean bigger clusters. The solid bar on each side is the first domino, the shaded area is what has to be pushed to reach it, and the dotted line is the simulated landing.
The path of least resistance is the more fragile side. The server recomputes it every scan, archives it in /archive/domino_BTC.csv (and the other coins), and sends a Telegram alert when a side exceeds ×5 with at least $10M of liquidations and a first domino within 3% of price.
Limits: this is a static simulation. In reality the book refills (or thins) during the move, and Hyperliquid liquidates large positions in pieces. Only scanned accounts are counted: smaller accounts, invisible here, add fuel to the chain. Cross-margin liquidation prices move with the account’s other positions. Read it as a fragility gauge, not a price forecast.
The domino simulator answers “what does it take to trigger a cascade?”. Simulate your crash answers the reverse: if someone fired an X-dollar market order right now, where would price go?
The Share my simulation button posts the result on X with a link that reopens the exact same simulation (coin, direction and amount) for whoever clicks, recomputed on the current book.
Limits: the same as the domino simulator. The book is frozen (in reality it refills or thins out during the move), only the 450 largest accounts are counted and Hyperliquid liquidates large positions in pieces. It is an order of magnitude of market fragility, not a forecast or financial advice.
To join the leaderboard, pick a nickname: no account, no e-mail. Your browser keeps a secret token that identifies you (clear the site data or switch devices and the nickname is lost).
A defense is a whale close to liquidation that adds money instead of cutting. The server detects it when, between two scans:
Only positions above $1M and additions above $50k are kept. The margin added is estimated (liquidation shift × size, or money arriving on the account).
Status: “still holding”, “closed” or “liquidated later”. Survival profiles (90 days): “defends” (never liquidated), “gets liquidated” (liquidated without ever defending) or “mixed”. A whale that keeps defending a level makes that level worth watching; a whale that lets itself get liquidated is potential fuel for a cascade.
The server listens live to every BTC, ETH, SOL and HYPE trade. When an account trades big, it checks the account’s age: if its first deposit was less than 7 days ago and it holds a position above $3M, it is flagged.
A brand-new account opening a large position straight away may be an incoming fund, someone well informed, or simply a trader who switched wallets: interpret with care.
Paste any Hyperliquid address (0x followed by 40 characters): your browser reads its positions live, on every coin, with account value and total unrealized PnL. The address is only remembered in your browser. Data refreshes every 5 minutes.
Every scan is archived and free to download. Times are in UTC.
.gz files open with any decompression tool, or directly with Python and pandas.
The Telegram channel receives whale moves above $10M (with 💀 when it is a liquidation), defenses on positions above $10M, every BTC cascade with a link to its page, the cascade score when it reaches 5/5 (at most once every 6 hours), new wallets and a daily summary. The X account only posts the biggest events, in English, a few times a day.
Data: Hyperliquid public API. Price, risk score and tracked whale are read live by your browser; the biggest accounts’ positions are scanned by the server every 5 min. The risk score is a historical statistic, not a trading signal. Nothing here is investment advice.