Methodology
Heat Score (异动分)
The composite score that combines OI Accel + CVD + liquidations + funding into a single anomaly number.
Heat Score is a per-symbol per-hour composite that combines the component signals (OI Accel, CVD Accel, Volume Accel, recent liquidations, funding skew) into a single number you can sort by.
What it does
- Z-scores each component signal against the asset's own 30-day baseline so cross-symbol comparison is meaningful.
- Sums the component z-scores with weights that reflect how much each signal has historically led price action on this market.
- Clamps the result into a bounded range so a single outlier component can't dominate.
We don't publish the exact weights because they evolve — the
implementation lives in src/lib/heat-score.ts and is the single
source of truth.
What it's used for
- The Rankings default sort when the time window is 1H.
- Coin Radar picks (top-N by Heat Score with diversity filter).
- Signals screener — when you add
Heat Score > Xas a condition, this is what gets evaluated hourly.
What it doesn't do
- It doesn't tell you direction. A high Heat Score means something is unusual; not which way the move resolves. Cross-read with CVD + funding for direction.
- It's not predictive. The score is computed on current data. Past Heat Score did not "predict" a move; it described a state.
Limits
- The 30-day baseline means newly-listed contracts get unstable Heat Scores for the first ~30 days of trading until baseline warms up.
- During market-wide events (BTC dump, FOMC, ETF flows) almost every contract spikes Heat Score together; rank ordering becomes noisy. The 24H window helps but doesn't fully decorrelate.