HyperSonnar guide

The product

HyperSonnar only reads public, on-chain Hyperliquid data — positions, trades, funding, everything already visible on the blockchain. Nothing is collected elsewhere, and no private access is needed to analyze a wallet.

The product's job is to sort through thousands of wallets and surface the ones that are actually copyable — not just the ones with the best win rate or the biggest balance. A copiability score (§2) does that sorting automatically, and it's the backbone of Copy trading, Coins and the wallet page.

• A transparent ranking of Hyperliquid wallets by copiability, not raw win rate.
• A deep per-wallet breakdown: PnL, trades, risk, funding.
• A continuous market-measuring engine (Scanner).

Copiability score

The six components: expectancy (average gain per trade, ROI-based — what the wallet earns relative to its size), risk (maximum drawdown — the worst equity drop from a peak), consistency (the share of time spent in progression, i.e. making new equity highs), quality (win rate), returns (profit factor — gross wins ÷ gross losses — on a log scale, so one extreme trade can't dominate), and activity (trade count). The product is geometric: one weak component drags the whole score down — none can be compensated by the others.

Quality and consistency are shrunk by the Wilson interval — put simply: a win rate is only believed as far as the trade count proves it. 90% wins over 10 trades weighs far less than 90% over 1,000. It's also why nobody scores 100: a perfect score would demand proof no finite track record can supply — a short history cannot reach a high score, whatever its win rate.

Before the score is even computed, a filter drops wallets whose median hold (median time from opening fill to closing fill) falls outside 0.5 hour – 2 weeks. Below that: scalpers — the window to copy the trade is too short to follow by hand. Above it: positions opened and forgotten — technically a trade, but nothing to copy, no identifiable entry or exit within a reasonable window.

The score is a heuristic, not a guaranteed prediction: it summarizes public history — it promises nothing about future trades.

Copy trading

Copy trading is the main ranking. Three tabs: Top copyable (sorted by copiability score), Today's winners (today's best perp performers), and All (every tracked wallet, no score filter).

The 1D / 1W / 1M columns come from each wallet's Hyperliquid portfolio and refresh roughly every 30 minutes. A "—" doesn't mean zero — it means the worker hasn't analyzed this wallet yet, not that its performance is nil.

On Today's winners, two quality signals: a greyed "Pending" chip means this wallet isn't in the rankings table yet — there's no way yet to tell if it's a genuine copyable perp trader or a spot-only wallet that gets excluded later. A ★ badge with a number means the opposite: this wallet is already ranked, its copiability score is proven.

Coins

Coins aggregates, per coin, the performance of every tracked wallet: total PnL, active traders, trade count, average win rate, and the top wallet on that coin — over a chosen period (today, yesterday, 7 days, 30 days, all-time, or a specific date).

Click a coin to see its top traders over that window, then Analyze a wallet to open its dedicated page. It's the fastest path to "who's trading this coin well right now."

Wallet page

Four KPIs at the top of the page: perp PnL all-time, win rate, profit factor, median hold. Win rate and profit factor feed the Quality and Returns components of the score (§2) directly, PnL feeds expectancy, and the median hold is the ranking's entry filter — enough to see immediately why a wallet is ranked where it is.

The cumulative PnL chart plots perp equity over time; hover a point to see the exact value and date. Below it: Win/Loss, PnL by coin, and PnL by month.

The aggregated trades table summarizes activity per coin/day. Duration = cumulated TIME IN POSITION (the sum of every closed round-trip's hold time), not the span between the first and last fill of the day — two one-hour trades inside an 8-hour day show 2h of Duration, not 8h. Fees = fees paid. Funding = net funding over the period (positive = received, negative = paid). Net = PnL + funding − fees.

Click a trade row to open its TradingView candlestick chart: ▲ (buy) and ▼ (sell) markers at each fill, plus two dashed horizontal lines for the entry and exit price.

Scanner

The Scanner is the measuring machine underneath: it continuously computes per-coin, per-minute metrics across every tracked coin (1m / 5m / 15m windows).

• z = this window's volume vs its own rolling average, in standard deviations (normal < 1.5 · spike ≥ 1.5 · strong ≥ 3).
• Spike× = the same volume as a plain multiple of its usual level — the raw number behind the z-score.
• ATR× = the size of the price move in multiples of the coin's average true range — a normalized way to say "is this a big move, for THIS coin."
• Flow labels (NEW MONEY / SHORT COVER / CROWDED LONG) come from combining the open-interest change with price: OI rising with price = fresh capital coming in (NEW MONEY); OI falling while price rallies = shorts capitulating (SHORT COVER, a fragile move); OI rising while price stalls = longs piling in without follow-through (CROWDED LONG, reversal risk building).

Market

Market ranks every tracked coin by % change vs its 24h reference open — gainers, losers, or all — with mark price and 24h volume.

Analytics & liquidation walls

The home page (Analytics) shows whole-market derivatives — open interest, funding, volume — and the REAL positioning of tracked wallets. “Real” means every position comes from an on-chain read of the exact wallet (clearinghouseState), never from a model or an estimate. The honest trade-off: coverage is partial, and it is always displayed (“X% of OI covered”).

• The tracked universe = every address ever seen on the Hyperliquid leaderboard (~40,000). Assumed bias: wallets that never hit the leaderboard are invisible — a selection bias, not a random sample.
• Three-speed freshness: ranked wallets are re-checked every ~20 min; positions within 2% of their liquidation price every ~3-5 min; the rest of the universe on a continuous rotation (a few hours). The “Checked” column shows it per row, and entries older than 45 min are excluded from the at-risk counter.
• “Smart money” = tracked leaderboard wallets, winners AND losers — the term describes provenance (accounts trading real size), not a guarantee of skill. The Net-loser badge on their pages exists for a reason.
• The liquidation wall aggregates real liquidation prices by price level (±40% of the mark). The heatmap = the HOURLY snapshots of that wall, stacked over time.
• Δ OI is computed in BASE units (position flow, price-neutral); displayed OI is notional USD. Funding is the HOURLY Hyperliquid rate — flat at 0.0013% = the base rate, normal in a calm market.

Economic calendar, earnings & indicators

Three connected pages: the Economic calendar (US statistics — CPI, jobs, GDP…), Earnings (company releases) and Company pages (audited SEC accounts). One shared rule: no in-house estimates — a value only appears once its source (FRED, BLS, SEC) has published it, and anything computed rather than published says so (a “derived” badge, an “announced — not yet filed” banner).

• Clicking a calendar row opens the indicator's page (/macro): the FULL published history — CPI goes back to 1914 — as a chart and a table, with the next release date.
• The CPI chart plots the year-over-year rate (“inflation”) computed between two PUBLISHED values — arithmetic, not an estimate; the raw index is one click away, with its base (1982-84 = 100).
• A hole in a series stays a hole: the October 2025 CPI was never published (federal shutdown) — the table skips that month instead of inventing a value, and the percentages stay correct.
• The CPI shown is the NON-seasonally-adjusted series — the headline release figure. The page and the calendar row show the same number.
• On earnings, two natures are kept apart: the figure ANNOUNCED in the press release (available within the hour, via Finnhub) and the figure FILED with the SEC (~30 days later, audited). The former shows on an amber-hatched banner until the latter lands — it never enters the charts.
• Macro series source: FRED (Federal Reserve Bank of St. Louis), linked at the bottom of every page.

FAQ