Every heat score on this site is produced by one public, deterministic formula. No language model touches the scores. Given the same inputs, anyone recomputes the same number.
Displayed on a 0–100 scale. The four asset-level factors are combined with a geometric mean, then gated by a single market-wide regime multiplier.
An arithmetic average lets one strong factor paper over a weak one: a coin with explosive momentum but no liquidity, no on-chain use, and heavy unlock overhang would still score highly. A geometric mean does the opposite — if any single factor is weak, it drags the whole score down, and no single factor can carry the rest. Heat is meant to fire only when several conditions line up at once, so the aggregation is chosen to enforce exactly that.
Regime is not averaged in with the others — it multiplies the result. It is a market-wide gate derived from Bitcoin's 30-day trend and dominance. Altcoins tend to ignite only in a risk-on market, so when the regime gate is low it pulls every coin's score down together, regardless of how strong that coin looks in isolation. The same asset-level factors mean different things in a risk-on versus a risk-off market, and the multiplier encodes that.
Each factor is defined, sourced, and normalized on the Factors page. In short: regime (market gate), momentum (the coin's own price trend), liquidity (traded volume depth), fundamentals (TVL relative to market cap), and supply lightness (how much is already circulating).
The heat score is a relative, momentum-oriented indicator. It is not a valuation model and it does not tell you whether an asset is cheap or expensive — we tested a cheap/expensive axis and found it has no predictive power for crypto, so we do not publish one. It does not forecast price. It measures whether an asset is running hotter or colder than its own recent baseline, within the current market regime.
The formula above is the whole model — there is nothing hidden behind it. Every scored asset ships a
machine-readable .md mirror and a /api/*.json endpoint carrying the same numbers,
so the score can be independently recomputed rather than taken on trust. See
Reproduce. Language models are used only to translate written
explanations; they never generate or adjust a single score.
Updated 2026-08-14 · rebuilt daily
This site provides information based on public data. It is not investment advice or a solicitation to buy or sell. The heat score is a deterministic, relative indicator computed by a fixed formula; it does not predict future prices. Trading crypto assets involves risk, including price volatility. Make your own decisions.
Data provided by CoinGecko · TVL / fees / stablecoins: DefiLlama · Fear & Greed: alternative.me
Operated by InSync Inc.