Risk Disclosures & Methodology

This page collects Tape's full risk warning and explains, in plain language, how every number on the site is produced. The short version shown across the site links here so the terminal stays readable without hiding anything.

1. Full Risk Warning

Risk warning: trading perpetual futures is highly risky and involves the use of leverage, which can amplify losses and lead to the rapid, total loss of your funds through liquidation. Crypto prices are volatile. Nothing on Tape is financial, investment, tax, or trading advice, and no outcome is guaranteed. Markets shown (including whale prints and the liquidation heatmap) are estimates or raw data, not signals. You are solely responsible for your own decisions. Tape is non-custodial software provided "as is" and "as available", with no warranties of any kind, express or implied. To the maximum extent permitted by law, Tape and its operators, contributors, and affiliates disclaim and are released from all liability for any loss, damage, or claim — including lost funds, liquidations, missed or failed orders, downtime, data errors, or third-party (Hyperliquid, wallet, RPC) failures — arising from use of this site or any trade. By using Tape you accept these terms, assume all risk, and release Tape from any and all liability. Trade only with funds you can afford to lose. Not available where prohibited.

2. What Tape Is (and Is Not)

Tape is a free, read-only, non-custodial analytics interface for the Hyperliquid perpetual-futures protocol. It never holds funds or keys, cannot place trades, and is not a broker, dealer, exchange, custodian, or investment adviser. Full legal terms: Terms of Service · Privacy Policy · Refund Policy.

3. Data Sources

Market data (prices, funding, open interest, volume, order books, trades) comes live from the public Hyperliquid L1 API and is verifiable on-chain. TradFi quotes derive from public market-data endpoints; Treasury figures from official U.S. Treasury sources, cited in place. Data can be delayed, gapped, or wrong — treat every number as an input to your own research, not a conclusion.

4. Methodology, Panel by Panel

Technical levels & RSI scan

Support/resistance, moving averages (SMA 20/50/200, EMA 21), RSI(14), and ATR(14) are computed mechanically from Hyperliquid candles using standard Wilder formulas. Swing levels are fractal pivots clustered within ~1.2%, with touch counts shown. Breakout triggers are the 30-day range ± 0.25×ATR. Daily Fibonacci levels divide the prior day's high–low range at the classic 23.6/38.2/50/61.8/78.6% ratios. The RSI scanner applies the same RSI(14) to 1h closes (4h aggregated from the same series) across the top markets by volume. These are descriptive statistics, not signals or advice.

Liquidation heatmap (/liquidations)

Observed levels, not a forecast — but a sample, not a census. Hyperliquid publishes no aggregate liquidation-level feed, but each open position carries its own liquidation price in the exchange’s clearinghouse state. Tape reads that price directly for a sample of the largest wallets by account value, weights each level by that position’s notional, and buckets the result into a price grid. Bright zones are real liquidation prices belonging to real positions we could see.

Three limits are stated on the panel itself and in every API response. Coverage is partial: we publish the share of the market’s open interest the sample actually represents — the remainder is unobserved, not empty. Cross-margin levels are account-level: a cross position’s liquidation price reflects the whole account, so it can sit far from the mark; levels beyond ±25% of mark are excluded rather than smeared across the scale. The time axis is built from snapshots Tape records going forward, so a range shows only the columns captured so far and says as much when history is incomplete. A second bar plots realized liquidations from the live feed on the same price axis, so predicted and actual can be compared directly.

Liquidation feed (/liquidations)

A representative sample, not an exhaustive record. Hyperliquid’s public trades stream carries no liquidation flag and there is no global liquidation subscription. A liquidation is visible only on the fill of the wallet that absorbed it, which names the liquidated trader. Tape discovers those counterparty wallets from public data and polls their fills, so any liquidation absorbed by a wallet outside that set is not counted. The side shown is the side that was liquidated: a liquidated long is a forced sell, which prints as a buy for the counterparty.

Liquidation heatmap (terminal chart tab)

A model estimate, not a data feed. The terminal’s "Liq heatmap" chart tab is a different, older visual: it projects where liquidations would cluster if positions were opened at recent candle prices across common leverage tiers (5–100x), weighted and decayed over time. Bright zones there mean "model-implied density", not confirmed or observed levels. For observed levels use /liquidations.

Volume delta / CVD and depth heat

CVD accumulates taker buy minus taker sell notional from the live public trade feed since you opened the page — it resets when you switch markets and is session-relative, not exchange-official. The depth heatmap bins resting order-book notional within ±0.6% of mid, sampled about once per second; it shows displayed liquidity only, which can be spoofed or pulled at any time.

Funding & premium scans and top-wallet panels

Scans are mechanical threshold crossings on public data (funding extremes, OI shifts, momentum). Leaderboard and top-wallet rankings are past-performance sorts of public wallets; past performance does not predict future results. Whale prints are large public fills, shown raw.

Wallet activity labels

Labels are mechanical descriptions of published Hyperliquid figures, computed by the rules stated at /methodology — not ratings, endorsements, recommendations, or predictions of future results.

Rules version 2026-08-29.1. Every label below is arithmetic over figures printed on the same row, so any chip can be recomputed from the table by hand. Each chip also carries the computed input in its tooltip. Labels describe past published periods and nothing else. Inputs are the four windows Hyperliquid publishes per wallet — 24h, 7d, 30d and all-time — each with realized PnL, ROI and traded volume, plus current account value. Note that window PnL includes unrealized mark-to-market on open positions, which is why a wallet can show a large PnL move with no trading volume.

When no label is shown. A wallet with no lifetime traded volume gets no labels at all, because there is no trading for a trading label to describe — roughly 2,500 wallets in the upstream feed report a PnL with zero volume, and a confident label beside such a number would be worse than a blank. Labels are also withheld when the feed does not supply all four windows.

Activity (first chip). First matching rule wins; a row shows at most one.

• No equity reported — account value is 0. ROI is not meaningful for this row.
• Equity under $1k — account value under $1,000. ROI is not meaningful for this row.
• No volume 30d — no traded volume in 30 days, but lifetime volume above zero; the ranking reflects older activity.
• No volume 7d — no traded volume in 7 days, though the wallet traded earlier in the 30-day window.
• Turnover under 5x / 5–50x / 50–500x / 500x+ — 30-day traded volume divided by current account value, for accounts holding at least $10,000. How the bands fill depends on which sort you are looking at, so both figures are given rather than the flattering one: across every wallet reachable from these pages (all four sort metrics combined, 556 wallets on 29 Aug 2026) the bands hold about 17% / 28% / 35% / 20% of eligible rows; on the default view alone — 30-day window sorted by PnL — they hold about 20% / 46% / 18% / 2%, so on that view the middle band is the most common label and tells you least.

Windows up (dots). Four dots, in order 24h · 7d · 30d · all-time, filled where the published PnL for that window is above zero. This is a restatement of four signs the row already carries. It is rendered identically whether four dots are filled or none, and it is deliberately not sortable or filterable — a screen built on it would turn a description into a recommendation.

Context (second chip). These fire only when something should reduce the weight you place on the headline number. There is deliberately no complimentary counterpart. First matching rule wins.

• Short history — 30-day volume is 80% or more of lifetime volume.
• Day-concentrated — the 24h PnL is at least half the size of the 30-day move, in either direction, and at least $1,000.
• Week-concentrated — the 7d PnL is at least 60% of the size of the 30-day move, in either direction, and at least $1,000.
• Swing 50%+ of equity — the 30-day PnL, in either direction, is at least half of current account value.

What Tape deliberately does not label. No skill, quality or grade words, no size nicknames, no percentile or letter ratings, no streak or momentum framing, and no inference about who or what a wallet is — high turnover is produced by market making, arbitrage, hedging and volume farming alike, and Tape does not assert which. Nothing is derived from a wallet's display name or any off-chain source. ROI is not used as a rule input anywhere: Hyperliquid computes it against a moving equity base affected by deposits and withdrawals, so a rule built on it would not be reproducible from the figures shown.

A separate threshold, not to be confused with the bands above. The top-wallet panel served by /api/analytics excludes accounts whose 30-day volume exceeds 80x account value. That is an exclusion filter for one panel; the turnover bands above are a four-way description used on the leaderboard and directory. The two use different cut points because they do different jobs, and neither asserts that any wallet is a market maker. The leaderboard ranking itself applies no turnover exclusion.

AI Market Read

The AI read is generated by a large language model (Google Gemini) from the mechanical technical snapshot above — nothing else. It is cached for several minutes, can be inaccurate, incomplete, or stale, and is commentary on data, never a recommendation. The model name in use is shown in the API response at /api/btc-ai-read.

5. Paid Features

Tape Pro and the x402 pay-per-call API are optional. Crypto payments settle on-chain and are final; see the Refund Policy. Prices are shown before you pay.

6. Contact

Questions about anything on this page: support@tapeperp.com.