Cyclesolved

Market & technical data

• CoinGecko (Pro API) — the universe backbone. Coin selection, live price, OHLC candles (the basis for RSI, MACD, and the 50/200-day moving averages), circulating supply, drawdown from all-time high, developer activity (commits, stars, forks, merged pull requests), global-market figures including Bitcoin dominance, Bitcoin perpetual-futures funding rates, and 7- and 30-day price changes (the basis for the relative-strength-vs-Bitcoin macro signal).

On-chain, fundamentals & sentiment

• Santiment — the primary on-chain source for major assets: MVRV, active-address trend, whale-transaction counts, social volume, development activity, and supply held on exchanges.

DeFi protocol metrics

• DefiLlama — total value locked, 24-hour fees, and 24-hour revenue for DeFi protocols (the fundamentals score on DeFi coins); and token-unlock / emissions schedules (the percentage of circulating supply unlocking in the next 30–90 days, plus upcoming unlock events), from DefiLlama's public emissions data — used for the dilution signal and the unlock alert.

Exchange-supply & on-chain fallbacks

Where Santiment does not cover a particular coin or chain, these sources fill in supply-on-exchanges, active-address, and whale-transaction signals:

• CoinStats — exchange-wallet balances on non-EVM chains (e.g. Bitcoin, Solana, Tron).
• OKX Proof of Reserves — published custody attestations, used as a single-exchange floor for supply on exchanges.
• Binance Proof of Reserves — Binance's published custody address set across many chains, used to attribute exchange-held supply.
• Koios — Cardano (ADA) exchange custody via stake-pool attribution.
• Stellar.Expert and Stellar Horizon — exchange-tagged Stellar (XLM) accounts and on-chain balances.
• Nodely (AlgoNode) — large-transaction counts for Algorand (ALGO), via the public Algorand Indexer.
• Subscan — active-address trend for Polkadot (DOT).
• Dune Analytics — active-address trend for major EVM chains (Ethereum, BNB Chain, Arbitrum, Polygon, Optimism) plus native-L1 chains (Avalanche, Sei, Flare, TON, Bitcoin, Sui, TRON, XRPL, Cronos, Stellar); supply-in-profit for Ethereum and Arbitrum ERC-20 tokens; and, for ERC-20s on Ethereum, Arbitrum, and Polygon, exchange-held supply (from CEX-labeled wallet balances) and large-transfer counts (network-wide transfers of $1M or more).

Market sentiment

• Alternative.me — the Crypto Fear & Greed Index, used as a contrarian market-wide sentiment input.

Traditional & macro markets

• Alpha Vantage — stock and ETF data for the Macro page: the S&P 500, NASDAQ-100, Russell 2000, a software-sector ETF, Coinbase, gold, silver, the US Dollar Index, and the 10-year Treasury yield.
• CoinGecko — Bitcoin's price on the Macro page (Pro plan — the same integration behind the scanner).
• Alternative.me — the Crypto Fear & Greed Index on the Macro page.

Macro page substitutes ETFs for the underlying indices (SPY for S&P 500, QQQ for NASDAQ-100, IWM for Russell 2000, GLD for gold, SLV for silver, UUP for the U.S. Dollar Index). These ETFs track their underlying with negligible error for display purposes. VIX is not included due to CBOE licensing constraints; we use the Crypto Fear & Greed Index as our primary sentiment signal. Data sourced from Alpha Vantage, CoinGecko Pro, and Alternative.me.

Methodology & limitations

We would rather disclose a limitation than paper over it. A few worth knowing:

Scanner universe. Sometimes coins outside the top 100 are showing real promise. We score developer activity across the top 500 coins by market cap — a wider net than the scanner itself — so the promotion rule has room to fire on projects long before they climb the rankings. When our methodology flags a coin for sustained high developer activity, we add it to the scanner so users get the full data — it appears in the Top 100+ view, scored on the same pipeline as any top-100 coin. Once promoted, a coin stays for at least two weeks so the scanner doesn't flicker day to day; after that it drops out only on a sustained decline. It's a mechanical rule, not an editorial pick.

Supply on exchanges. Our supply_on_exchanges metric reflects labeled-address totals from on-chain data — it counts only coins sitting in exchange wallets we can positively identify. On chains where address labeling is incomplete (notably XRP and SOL), it can understate true exchange custody by a wide margin — potentially several-fold, not a rounding error. On those chains the figure can read well under 1–2% when real custody is materially higher, so a low reading there usually reflects thin labeling rather than genuinely little supply on exchanges. Read it as a floor on those coins, not a precise figure. We disclose this rather than silently estimate the gap.

Drawdown from all-time high. We compare current price to the all-time high and treat depth as a probability tilt — deeper drawdowns lower the odds of recovery rather than being rewarded as "cheap." For the rare coin whose ATH is a corrupted print — a manipulated launch wick on near-zero real liquidity (e.g. ICP's FTX-era ~$630–750 top) — measuring from that peak is meaningless. For those we measure drawdown from a post-stabilization recovery high instead (ICP: its March 2024 ~$21 high) and disclose the adjustment on the coin's drill-down. The list is kept short and is for genuine manipulation only — not for low-float launches, whose decline is real, or dead projects, which the quality gates handle.

Token unlocks. Upcoming-unlock coverage comes from DefiLlama's emissions dataset, which tracks most low-float / high-FDV coins (where pending dilution matters most) but not every coin — fully-circulating assets have no schedule, and some smaller coins aren't tracked. Where we have no schedule the signal simply drops and the other Fundamentals signals carry the category. It refreshes on a slow cycle from a public source; if that's ever unavailable, the signal and the unlock alert degrade silently rather than affecting anything else.

Developer activity — tracked orgs, not the whole ecosystem. Our dev-activity metric counts GitHub activity across the organization(s) our data provider maps to each project — typically the core protocol org. For a project whose development lives in one place, that's the full picture. For a large, decentralized ecosystem it is not: Ethereum's figure reflects the core Ethereum org, not the independent execution/consensus client teams (Nethermind, Besu, Erigon, Prysm, Lighthouse…), the L2s, or the wider tooling — so a coin like ETH is undercounted relative to its true ecosystem, and the active-contributor count in particular reads far lower than reality. We treat this as a floor on activity, not a ceiling, and read dev activity as one signal among several rather than a precise ecosystem-wide census. The Dev Activity page ranks coins on this same core-org basis.

Supply in profit. Coverage is currently limited to 47 coins — 45 ERC-20 tokens on Ethereum plus 2 on Arbitrum, where Dune exposes a transfers table with pre-joined per-block USD pricing so weighted-average cost basis is a single-table computation. Native L1 assets including BTC, ETH, SOL, ADA, DOT, NEAR, and the other ~58 native-chain coins in our universe are not yet covered. Adding them requires per-chain methodology (UTXO cost-basis math for the Bitcoin family, account-level accounting for ETH including staking and beacon withdrawals, program-vs-wallet labeling for Solana) that established providers like Coin Metrics and Glassnode have spent years solving. Our plan post-launch is to ingest from those providers rather than rebuild the methodology from scratch on Dune. Until then, coins without coverage show "—" for this metric and the other Fundamentals signals carry the score for that category.

BTC dominance. Our dominance figure is CoinGecko's global market-cap percentage for BTC — the same source as our prices and market caps, so the number stays internally consistent. It can read a point or two below TradingView / CoinMarketCap, which compute dominance over a different asset universe (notably how stablecoins and wrapped assets are counted). The small gap is a methodology difference between references, not an error.

Capitulation gate. During a confirmed acute market capitulation, the composite temporarily shifts weight away from price momentum and broad-market trend — which are mostly noise during a forced-liquidation flush — toward holder economics and contrarian positioning, so a fundamentally sound asset being sold off indiscriminately isn't dragged down purely by the crash. The gate keys on the acceleration of fear (the speed of the Fear & Greed drop) together with price, not the standing fear level — so it engages on genuine events and stays dormant the overwhelming majority of the time. Below a small floor the adjustment switches off entirely, so calm and minor-chop periods leave every score exactly unchanged, not merely approximately so. It also keeps engaging smoothly through multi-day liquidation grinds and the bottom, rather than switching on and off. The lift is reserved for genuine quality at a discount: a coin only receives it to the extent its on-chain activity holds up and its fundamentals are more than just a deep drawdown, so an asset that has merely fallen hard does not get rewarded for falling.

Funding rates — tested and set aside for capitulation. We investigated perpetual-futures funding as a contrarian bottom signal, on the intuition that deeply negative funding marks crowded shorts near a capitulation low. Across four historical drawdown events it did not hold up: by the once-daily snapshot we record, funding had already reset to neutral at the actual bottoms, so it carried no usable signal at our sampling resolution. We therefore do not use funding in the capitulation gate. It does appear on the euphoria side, where persistently positive funding is one corroborating marker of a leveraged top — a slower-moving condition that survives daily sampling where the capitulation spike does not.

Euphoria defense — conservative cap live, full reweight in shadow. The mirror of the capitulation gate, for market tops: when greed is sustained, leverage-long funding is persistently positive, and price is extended near its highs, the methodology recognizes a euphoria regime. The detector fires across five historical tops spanning two cycles. What is not yet validated is the precise magnitude of the response — because our score history has been recorded entirely within a drawdown, there is no live top to confirm a reweighting's effect on actual scores.

We do not treat "we can't perfectly tune it" as a reason to apply zero defense — the asymmetry forbids that. Too-cautious at a top costs a little late upside (small, recoverable); leaning bullish into a top endorses the peak (catastrophic). So at a detected euphoria we apply a conservative, live cap: each momentum/trend signal (the technical trend components, the BTC-uptrend macro read) is capped at neutral, scaled by how strongly the detector is firing — so the model cannot add bullish points from momentum at a top. This is direction-only and needs no tuning: the cap is neutral, nothing more. Contrarian signals are untouched — an overbought RSI still reads as caution and can pull a score below neutral. When this is active you'll see a "capped in euphoria" note on the affected signals in the drill-down; the raw reading is still shown, because we're choosing not to bank it, not hiding it.

The fuller response — aggressively up-weighting contrarian categories at a top — remains a shadow: computed and logged for our own validation, never changing a user-facing score, until a real euphoria event lets us confirm its magnitude. The conservative cap is the floor that holds the line until then. We would rather be slightly too cautious at a top than pro-cyclical at one.

Capitulation-gate validation coverage. In the interest of not claiming more than we've verified: the gate's firing behavior — when it engages and how strongly — was validated against four real drawdown events spanning 2024–2026. Its full effect on composite scores and rankings, however, was only validated on the June 2026 event, because we do not retain the per-coin category scores needed to replay the earlier events. We treat the gate's broader composite-level behavior as reasoned-and-spot-checked rather than exhaustively backtested, and we'll revisit it as more events accumulate under live scoring.