How SignalHub works

Where does the data come from?

Prices, volume, open interest, funding, and position figures are computed from the Hyperliquid public API. Polymarket data comes from public markets tagged Fed rates, macro indicators, economic policy, CPI, tariffs, jobs report, geopolitics, economy, finance and must meet the configured volume and liquidity minimums. Macro news comes from RSS feeds including Federal Reserve, ECB, BLS, BEA, Bloomberg, CNBC, MarketWatch, Investing.com, OilPrice.com, and FXStreet. Grok AI only selects, translates, and summarizes macro news and writes short cause-and-effect impact reasons; it does not produce numbers.

Which wallets are tracked, and how often are they scanned?

The whale pool uses Hyperliquid leaderboard addresses with account value above $100K, positive monthly PnL, and all-time PnL above $1M. It currently contains 1,446 addresses, including 237 that meet the gross-position threshold of $1M; the hot set is scanned about once per minute, while other addresses follow a cold cycle of about 10 minutes. Market makers are excluded when maker share reaches 95%, or when it reaches 80% with at least 300 fills per day.

What rules trigger each signal type?

Each signal is filtered by inspectable program rules using public market data and configured thresholds.

• Whale move: For qualified tracked wallets, the position is at least $1M; over 12 hours, the add reaches the greater of $1M or 3 times that wallet's typical add.
• Fake consensus: At least 6 wallets on one side each hold at least $50K, their total reaches $10M, and the largest wallet holds at least 60% of that total; profile coverage must reach 80%. It warns that one large wallet may dominate what looks like crowd consensus.
• Liquidation alert: Tracks positions whose liquidation price is within 10% of the market price; a position must be at least $5M, or at least $1M and 5% of open interest. Alert bands are 2%, 5%, 10%. The displayed distance is how far price would need to move for liquidation; it does not mean liquidation already happened.
• Whale-hunt risk: Flags tracked positions of at least $500K whose liquidation price is within 8% of mark, when the position is at least 10% of open interest or 0.5 times 24-hour volume. Distance bands are 3%, 5%, 8%. It describes public liquidation risk, not a directional recommendation.
• Extreme funding: When open interest is at least $5M, funding qualifies at an absolute annualized rate of 50% plus a z-score of 3 after 72 history samples; before that coverage, the fallback annualized threshold is 100%. Mainnet only; on HIP-3 markets funding signals showed no edge in backtests or live tracking, so they are no longer sent — see Crowded rally for HIP-3.
• Open interest anomaly: With open interest of at least $5M, it flags changes of 10% over 1 hour or 18% over 4 hours; a price move no greater than 1% is marked as a quiet build. It reports positioning and price changes without predicting direction.
• Smart-money move: Over a 4-hour window, qualified wallets show net flow of at least $5M across at least 3 wallets; the dominant side is at least 70% of total changes, with no single wallet above 60%.
• Polymarket odds: Tracks public Polymarket markets in configured topics (Fed rates, macro indicators, economic policy, CPI, tariffs, jobs report, geopolitics, economy, finance) with at least $20K in volume and $10K in liquidity; YES odds move by at least 8 percentage points over 1 hour or 12 points over 6 hours. It does not predict Hyperliquid price direction.
• Price move: Flags a price move of at least 3% over 1 hour with a z-score of 5; open interest and 1-hour volume must each be at least $5M. Its volatility baseline spans 72 hours with at least 80 samples.
• Crowded rally: HIP-3 markets only: open interest is at least $5M, the 24h gain is at least 5%, and annualized funding is at least 100% both now and one hour earlier (longs pay).

What role does AI play?

Software calculates prices, volume, open interest, funding, and position changes from public data; Grok AI is limited to macro-news selection, translation, summaries, and short cause-and-effect explanations, not market figures.

How do signal-time and current data differ on detail pages?

Detail pages separate the price and position data saved when a signal was emitted from currently available market data; current values should not be treated as information known at the earlier time.

How should position changes and timestamps be read?

Position changes are estimated from differences between scans, and timestamps identify scan times rather than executions; the high-activity wallet set is scanned about once per minute, while other wallets follow the cold-scan cycle.