Help & Docs

Hyperliquid is a non-custodial copytrading analytics platform for Hyperliquid perpetuals. Every position on Hyperliquid is public and on-chain, so any wallet's record can be audited by anyone. We do that audit for you: we check where each wallet's profit came from, score whether you could realistically follow it, and — if you choose to — mirror its trades from your own wallet. We never hold your funds.

Every wallet in the universe is processed through:

1. Audited risk metrics — Sharpe with real cadence (365×n/age, not √365 naive) + small-sample shrink, MDD over PnL (not AV), Calmar, monthly consistency
2. Skeleton check — distinguishes legitimate withdrawals from real losses, flags wallets whose PnL was ever negative
3. Structural classifier — MM / Grid / Scalp / Swing / Martingale / HFT on top of 8×8 cohorts (size × profitability)
4. Copyability score ★ 1-5 — can we mirror this wallet's entry AND exit before the edge is gone? Pure timing: hold duration, frequency, grid/MM concurrency, sub-30s scalp. NOT a quality verdict — Sharpe/ROI/MDD/ticket/recency are the leaderboard filtersof the leaderboard, applied separately.
5. Realistic backtest — Portfolio simulation enforces HL's $10 minimum notional. When proportional copy would put a sub- dollar trade, we force it to the floor.

• Free — full leaderboard, audited metrics, wallet detail (read-only deep tabs), portfolio backtest with your selected wallets
• Everything — Free, no fee — alerts, watchlists, side-by-side compare, advanced metrics, CSV/JSON export and (on Hyperliquid, when armed) auto-copy are all included. The platform is funded by voluntary donations.

Go to Leaderboard. Use the filter strip to set the quality bar — min profit, min capital, Sharpe, recency — then a minimum copyability (★) and a sort metric (Sharpe / PnL / MDD / Median AV). Losers and dust-capital wallets are filtered out by default.

Click any row to open the full Wallet Detail. Use the + Add button to put a wallet in your cart.

The Wallet Detail shows:

• Header with cohort badges + copyability stars + watchlist toggle (Activated)
• KPI grid with the most important metrics — PnL is Polymarket's venue figure
• Deep tabs: Capital / Trades / Coins / Temporal / Episodes / Structure / Skeleton

Some tabs reconstruct from the wallet's full activity. The first open downloads it; subsequent visits are incremental and fast.

From the Leaderboard or Wallet Detail, add wallets to your cart. Open Portfolio to:

• Adjust weights with sliders
• Set starting capital ($100-$1M)
• Toggle Realistic mode (default ON) — enforces HL's $10 min notional
• See aggregate KPIs (PnL, return %, final value)

The cart is persisted in your browser's localStorage and works across tabs.

Every feature is free — there is no subscription, no card and no fee. Connect your wallet and use everything.

The platform is funded by voluntary donations. Donations are optional, grant no service and are non-refundable — if it brings you value, support us from the Account page.

The headline differentiator — and it answers one question: given our signal→react latency, can we mirror this wallet's entry AND exit before the edge is gone? It is pure timing & structure, NOT a quality score.

Quality — Sharpe, ROI, MDD, win-rate, ticket size, recency — lives in the leaderboard filters, applied separately. A mediocre-but-slow trader is highly copyable; a brilliant scalper is not. Filter on quality, then copy what the stars say you can.

What it costs:

• ★★★★★ Copy at leisure — discrete entries held with plenty of time to mirror both legs
• ★★★★☆ Easy to copy — slightly brisk or a little concurrency
• ★★★☆☆ Workable — fast-ish or somewhat concurrent, copyable with effort
• ★★☆☆☆ Marginal — fast + concurrent, easy to miss fills
• ★☆☆☆☆ Uncopyable — scalp / HFT / heavy grid/MM: no time to react

Hard fails forcing 1★: pct_lt_30s > 30% (latency scalp), peak_simultaneous ≥ 30 (grid/MM), median hold < 1min, trades/day > 15 (churn).

On the leaderboard (no fills) we approximate timing with trade frequency; the wallet detail uses real per-position hold time + concurrency.

A copy is sized proportionally: scale = capital × weight ÷ the wallet's typical ticket. With a small bankroll copying a big wallet, the proportional copy can fall below HL's $10 minimum — which the venue would reject.

Realistic mode detects this and forces the trade to the $10 floor instead. Each trade then becomes a much larger fraction of your bankroll than it was of the leader's, amplifying both wins and losses — so the backtest reflects what you'd ACTUALLY fill, not fictional sub-minimum trades.

The minimum: HL documents a “minimum order value $10” (per-coin floor varies $10.00-$14.07 with size decimals × mark price; we use a flat $10 for ~95% accuracy).

Returns come from the TRADER'S PnL over a constant capital base, annualized with the wallet's actual snapshot cadence: mean(ΔPnL / base) / std × √(365 × n_snapshots / age_days).

Most platforms compute mean/std × √365 regardless of actual sample frequency. With weekly snapshots over a 2-year wallet, that's a ~17× inflation. Ours matches the wallet's real trading cadence. We also do NOT derive returns from the account value: it moves with deposits and withdrawals, not only with trading — a wallet that had LOST $11.2M once ranked #5 that way (audit #180).

We compute max drawdown on the cumulative PnL series, not the AV (account value) history. Why: a wallet that retires $500k of profits shows a 99% AV drawdown — but the trader didn’t lose anything, they just took money home. MDD measures only what the trader could control: the PnL accumulation.

Note: most wallets ride a near-monotonic PnL curve, so a near-zero MDD is normal — we show Calmar as "—" below ~1% drawdown rather than a fake astronomical ratio.

Six ratios that summarise the economics of a wallet's wins vs losses. All show "—" when mathematically undefined (e.g. no losses yet) instead of a fake 0 or ∞:

• Sortino — like Sharpe, but only downside swings count as risk. High Sharpe + low Sortino = the volatility is mostly losses.
• Profit factor — total winning $ ÷ total losing $ — the unit is the MONTH on Hyperliquid and the DECIDED MARKET on Polymarket, never the trade. Above 1 makes money; 2 = wins twice what it loses. Display capped at "50+".
• Payoff ratio — average win ÷ average loss, independent of how often each hits. ⚠️ The unit is NOT the trade: it is the MONTH on Hyperliquid and the DECIDED MARKET on Polymarket.
• Expectancy — the average $ THIS WALLET nets on a closed trade, at ITS size — not yours. Your copy is sized to your own bankroll and your per-order cap, so read it as the shape of their edge, never as your expected profit.
• Recovery factor — net profit ÷ worst drawdown — how many worst-case holes the total gain is worth.
• Kelly size — suggested share of YOUR capital for this wallet (half-Kelly, capped at 50%). A sizing guide derived from win rate + payoff — not a guarantee.

Heads up: an extremely high Sharpe/Sortino (8+) is almost always a high-frequency market-maker — hyper-predictable PnL but uncopyable. Always read these next to the copyability stars.

Cross-checks AV timeline vs PnL timeline to distinguish:

• ▼ Withdrawals — AV drops, PnL flat → legitimate cash-out
• ✗ Real losses — AV drops AND PnL drops → trading loss
• ? Mixed events — both change in different proportions

Plus: detects if cumulative PnL was ever negative, counts consecutive negative months, checks recent activity. It outputs one of three severities:

clean attention skeleton

Every wallet carries three labels: how BIG it is, how much it has WON or LOST, and HOW it trades. None of them is a recommendation — they are a summary of what the account has actually done, computed from official exchange data. A big wallet can be a terrible trader, and a small one can be excellent.

Based on the account's TYPICAL capital (the median account value over its life), not on a lucky peak. It tells you whether a wallet's results are reproducible at your size: a strategy that works with $50M may be impossible with $500.

Lifetime realised profit or loss, read from the venue's official API — never reconstructed by us. The red tags are as important as the green ones: seeing what a losing account looks like is half of the lesson.

Measured from the REAL median hold time (entry → exit) taken from the wallet's own fills. That single number decides the style, because it is also what decides whether YOU could copy it in time.

When we have no fills for a wallet we fall back to its turnover — how many times a day it rotates its own capital (daily volume ÷ typical capital). It is a proxy, so the confidence drops. If there is not even enough capital to make turnover meaningful, the style is shown as Unknown: we never invent one.

Confidence: HIGH needs at least 30 days of life, 50 recorded changes and 3 months with activity. LOW means the speed signal is not reliable (no fills and no usable capital) or there is too little history. Treat a low-confidence style as a hint, not a fact.

Next to the style you may see one or more traits. They are the things that a single label hides, and several of them are warnings:

• laddered — opens many positions at once (a ladder or a grid) — hard to mirror, and it is why a slow trader can still score low on copyability
• high win-rate grinder — wins often but small: it lives on hit rate, so a single bad streak hurts more than it looks
• asymmetric — lets winners run — lets winners run — few, large wins pay for many small losses
• consistent month-to-month — green month after month, with profit spread across many trades instead of one
• streaky / few green months — few green months — the result depends on when you got in
• one-bet dependent — one single market made 60% or more of the profit: that is luck until proven otherwise
• deep drawdowns — has lost 60% or more from a peak at some point
• inactive >30d — has not traded in over 30 days

A tag is a summary of the past, computed from public data. It is not a prediction, not advice, and not a promise that the wallet will keep doing what it did. We publish the exact thresholds precisely so you can disagree with our reading and form your own — that is the whole point of this site.

Yes — completely free. The full leaderboard, audited metrics, wallet detail (read-only deep tabs), portfolio backtest, alerts, exports, watchlists and (when armed) auto-execution are all included. No subscription, no card, no fee — the platform is funded by voluntary donations.

Directly from Hyperliquid's public API (/info endpoint). No insider data. Anyone can access the same.

It is only timing: can we mirror the entry AND exit with our latency? Signals are median hold duration (entry→exit), trade frequency, grid/MM concurrency, and sub-30s scalp share. That's it. Sharpe, ROI, MDD, win-rate, ticket size and recency are quality filtersyou apply on the leaderboard — they are NOT part of the star score. See Metrics → Copyability score.

Never. Hyperliquid is non-custodial by design. For analytics we only read public on-chain data, and you only sign a login message (SIWE) that proves you control the address — it moves nothing. When you turn on auto-copy: you authorize an HL agent wallet (a limited signing key that cannot withdraw, transfer, or modify staking) — you keep full control of your keys and can revoke any time, and we cannot block that.

Because of Realistic mode. Below a certain bankroll, the proportional copy of a big wallet's trades falls below HL's $10 minimum — so we force the trade to the floor instead. Each trade then becomes a much larger fraction of your small bankroll, amplifying both wins and losses. See Metrics → Realistic backtest mode.

Sharpe (a quality filter) rates how good the wallet is in absolute terms (risk-adjusted return). It can be stellar and the wallet still uncopyable —— because it runs sub-second HFT or market-makes dozens of positions concurrently, so there is no time for us to mirror the trade.

Copyability ★ answers the orthogonal question: can we replicate the entries in time? They measure different things — quality vs timing. The actual signal is both: a high-Sharpe, real-capital wallet that is ALSO 4-5★ copyable.

• HL / Hyperliquid — high-performance perp DEX on its own L1 blockchain
• AV — account value (collateral + unrealized PnL) at a given timestamp
• Agent wallet — HL native secondary signing key with restricted permissions (no withdraw)
• PnL — profit-and-loss, cumulative or realized
• MDD — max drawdown, the largest peak-to-trough decline
• Sharpe — risk-adjusted return: mean / std × √annualization
• Calmar — annual return / max drawdown
• Gini — concentration coefficient on monthly PnL distribution; 0=even, 1=concentrated
• mPos% — share of months with positive PnL delta
• Copyability ★ — pure timing verdict: can we mirror entry→exit before the edge is gone (NOT a quality score)
• SIWE — Sign-In With Ethereum (EIP-4361), wallet-based auth without passwords
• Non-custodial — platform never holds user funds; user retains full control via their wallet keys