Most crypto traders look at price charts. A smaller number look at derivatives data: open interest, funding rates, and long/short ratios. An even smaller number integrate macro and on-chain data: BTC dominance, fear and greed index, NUPL, stablecoin market cap, and exchange flows. The traders who consistently outperform tend to be in this last group, because they are trading on information that the majority of market participants ignore.
Coinglass is the leading provider of aggregated crypto derivatives and on-chain data. Their Hobbyist plan provides access to 24 data types with approximately 6 months of history at 4-hour granularity for derivatives, and years of history for macro and on-chain indicators. The data is comprehensive, well-structured, and covers 25 major symbols.
The challenge is turning that data into trading decisions. Coinglass provides the data. QuantForge provides the strategies, backtesting, and deployment infrastructure to act on it.
The Data Pipeline
Our integration with Coinglass fetches 24 distinct data types into 24 SQLite tables. The data is organized into five categories.
Per-symbol derivatives data covers 25 coins across 8 types: funding rate OHLC, top trader long/short ratios, top position ratios, liquidation history, taker buy/sell volume, cumulative volume delta (CVD), futures-to-spot volume ratio, and aggregated open interest.
BTC and ETH specific data includes options open interest and exchange balance (net inflows and outflows).
Global indices provide full historical depth: BTC dominance, fear and greed index, stablecoin market cap, altcoin season index, Coinbase premium, BTC correlation with traditional markets, options-futures ratio, BTC NUPL, BTC short-term holder SOPR, BTC vs M2 money supply, and BTC active addresses.
The sync runs every 4 hours via a background loop in the API server. Each sync fetches incremental data since the last timestamp, using INSERT OR IGNORE for deduplication. The rate limiter respects Coinglass API constraints with 2-second delays and exponential backoff on 429 responses.
The full data pipeline can also be triggered manually for backfill operations. The fetch script supports a registry pattern with 24 data types, and aliases for bulk operations: "derivatives" fetches all 8 per-symbol types, "macro" fetches all global indices, "order_flow" fetches CVD and taker volume, "onchain" fetches exchange balance and NUPL-related metrics.
Strategies That Use Coinglass Data
The Coinglass integration enables four categories of strategies that would be impossible with price data alone.
The leverage composite strategy combines three Coinglass-derived signals: open interest change rate, funding rate extremes, and long/short account ratios. When OI is rising abnormally (above the threshold), funding is extreme (indicating crowding), and the long/short ratio confirms the crowd direction, the strategy takes the contrarian side. This strategy is deployed on ARB, OP, and WIF with average Sharpe ratios of 1.89-3.02 across validation periods.
The NUPL cycle filter uses BTC's Net Unrealized Profit/Loss ratio, which measures whether the market as a whole is in profit or loss on their holdings. When NUPL enters the euphoria zone (above 0.75), the strategy reduces exposure, anticipating a correction. When NUPL drops to capitulation (near 0.0), it increases exposure, anticipating recovery. This strategy is ROBUST on 7 symbols with 4-5 positive regime periods out of 5.
The stablecoin supply momentum strategy tracks the rate of change in total stablecoin market cap. An accelerating stablecoin supply indicates fresh capital flowing into crypto, which historically precedes altcoin rallies. The strategy is ROBUST on 5 symbols (AVAX, NEAR, SOL, LTC, ATOM) with Sharpe ratios that strengthen during the 2023-2024 recovery period.
The correlation regime strategy uses BTC's correlation with traditional equity markets (S&P 500) to detect regime shifts. When BTC-equity correlation is high, crypto behaves like a risk asset and altcoins are vulnerable to equity selloffs. When correlation is low, crypto trades on its own fundamentals. The strategy adjusts exposure based on this regime detection and is ROBUST on 6 altcoins.
The Data-Strategy Connection
Each Coinglass data type maps to specific trading hypotheses.
Funding rates tell you who is paying whom. When longs pay shorts (positive funding), the market is net long. Extreme positive funding historically precedes long squeezes. Our leverage composite uses funding extremes as one of three composite signals.
Open interest tells you how much leverage is in the market. Rising OI means new leveraged positions are being opened. Falling OI means deleveraging. Combined with price direction, OI reveals whether a move is driven by new conviction (OI rising) or position unwinding (OI falling).
Long/short ratios tell you how retail accounts are positioned. When 65%+ of accounts are long, crowding risk is elevated. Our lsr_crowded_long threshold at 0.65 triggers a bearish signal component in the leverage composite.
CVD (cumulative volume delta) tells you whether buyers or sellers are more aggressive. Sustained negative CVD during a price decline confirms selling pressure. CVD divergence (price dropping but CVD rising) can signal exhaustion.
NUPL tells you where the market sits in the profit/loss cycle. It has predictive power because unrealized profit levels influence selling behavior: holders in deep profit are more likely to sell, creating supply pressure that caps prices.
Stablecoin market cap is a measure of sidelined capital. When stablecoin supply grows, money is entering the crypto ecosystem but sitting in stable assets. Historically, stablecoin supply growth precedes altcoin rallies as that capital rotates into risk assets.
What Coinglass Cannot Do Alone
Coinglass provides excellent data visualization on their website. You can see funding rate heatmaps, OI charts, liquidation distributions, and macro indicators with well-designed dashboards. For discretionary traders who want to check market conditions before placing a trade, the Coinglass dashboard is valuable.
But visualization is not automation. Seeing that funding rates are extreme does not generate a trading signal at a specific price with a specific confidence score. Noticing that NUPL is approaching euphoria does not automatically reduce your position sizes across a 45-bot portfolio. Observing that stablecoin supply is accelerating does not trigger a systematic entry with risk-managed position sizing.
The gap between observation and execution is where most traders lose their edge. They see the signal but act too late, too early, or with the wrong position size. Automated strategies eliminate this gap by executing the same logic at every opportunity, without delay and without deviation.
The Integration Architecture
From a technical perspective, the Coinglass integration follows the same pattern as our Binance data pipeline. Data is fetched via API, stored in SQLite tables with timestamp-based deduplication, and served to strategies through the DataStorage layer.
The backtest engine loads all available Coinglass data alongside OHLCV candles when setting up a backtest run. The data is passed to strategies through the candles dictionary with keys like "funding", "oi", "lsr", "basis", "cvd", "btc_dom", "fear_greed", and others. The engine slices the data to prevent lookahead bias, ensuring that strategies only see data that would have been available at each historical timestamp.
This architecture means that adding a new data type from Coinglass requires three steps: add a SQLite table, add a fetch function to the Coinglass script, and add a loader to the backtest engine. The strategy can then access the new data immediately through the standard interface. No changes to the strategy framework, the risk system, or the execution layer.
The combination of comprehensive data and automated strategy execution is what distinguishes a data-driven trading operation from a data-informed discretionary one. Coinglass provides the information. QuantForge provides the discipline to act on it consistently.