Crypto Market Data: A Complete Guide
Crypto market data is the information generated by trading activity in crypto assets, every trade, every order, every price, collected, organised and delivered so that people and systems can use it, and the raw material for trading, valuation, risk management, research, and the on-chain contracts that now settle billions of dollars against it.
Most guides to crypto market data are a list of data types. This one is too, in its first half. A list of types leaves out the question that actually matters for anyone using the data: where does the price come from, and how is its accuracy established? The second half answers that.
Types of crypto market data
Market data divides into a handful of categories. Each describes a different aspect of what is happening in a market, and each has different uses.
Trade data
The most fundamental category. Every executed trade generates a record: the price, the quantity, the time, the side (buy or sell), and usually a unique identifier. At the finest resolution this is tick data, every individual trade as it happened. On an active pair like Bitcoin against USD, a single exchange can generate over a thousand trade records a minute.
Trade data is the basis for everything derived from what actually happened: volume, volume-weighted prices, and any analysis of executed activity.
Order book data
An order book is the list of standing orders to buy and sell at various prices. Order book data captures the state of that book, and comes in levels of depth. Level 1 is the best bid and best ask, the top of the book, plus their sizes; it gives the spread and the mid price. Level 2 is the aggregated depth at each price level: how much is bid and offered at every price, without identifying individual orders. Level 3 is every individual order, including its size and position in the queue, and few crypto exchanges publish it in full.
Order book data describes intent, meaning what the market is willing to do, where trade data describes what it has done. Order book data is the basis for measuring liquidity, estimating the cost of a trade before making it, and pricing assets where trades are too sparse to rely on.
OHLCV and candles
Open, high, low, close, volume: the summary of trading over a fixed interval, whether a minute, an hour or a day. Candlestick charts are OHLCV data drawn visually. OHLCV is the most familiar form of market data and the most compressed: it reduces thousands of trades to five numbers per period, and loses everything about the sequence and distribution of those trades in the process.
OHLCV is well suited to charting, long-horizon analysis and anything where the summary is enough, and poorly suited to anything that depends on what happened within the interval.
Derived prices
Raw trades and quotes are inputs. Most uses need a single number, "the price," derived from them, and there are several ways to derive it: the last traded price, simply the most recent trade; the mid price, the average of best bid and best ask; the volume-weighted average price (VWAP), the average price of trades over a window weighted by size; the time-weighted average price (TWAP), the average of prices sampled at regular intervals; mark and index prices, derivatives-specific prices designed to resist manipulation; and state prices, read directly from decentralised exchange pool reserves instead of from trades.
Each measures something different, and each is right for some purposes and wrong for others. Every one of them is covered in Crypto Prices Explained.
Reference rates
A reference rate is a derived price constructed specifically to be used by others: to settle a contract, value a fund or benchmark performance, calculated by a published method from defined sources over a defined window, and designed to be representative, manipulation-resistant and reproducible. The CME CF Bitcoin Reference Rate, which settles CME's Bitcoin futures and most US spot Bitcoin ETFs, is the best-known example. Reference rates, and what regulation adds to them, are covered in Crypto Reference Rates and Benchmarks.
Derivatives data
Crypto derivatives, perpetual futures in particular, generate data of their own: funding rates (the periodic payment between longs and shorts that keeps a perpetual anchored to spot), open interest (the total value of outstanding contracts), liquidation data, and the mark and index prices mentioned above. Derivatives account for the majority of centralised exchange volume in CoinDesk Data's monthly exchange reviews, so this data describes the larger part of the market.
On-chain data
Everything above is generated by trading venues. On-chain data is generated by the blockchains themselves: transactions, wallet balances, smart contract state, token transfers, DEX pool reserves. On-chain data is a different category, closer to fundamentals than to market data, with one important overlap: prices on decentralised exchanges are on-chain data, and state prices are derived from it.
Metadata
The reference information that makes everything else usable: which assets exist, which pairs trade where, which symbols map to which assets, contract addresses, decimals, listing and delisting dates. Metadata is unglamorous, and it is where a large share of data errors originate, because the same asset has different symbols on different venues and the same symbol can mean different assets.
How crypto market data is delivered: REST, WebSocket, flat files
Three delivery methods cover nearly all market data access. REST APIs return data on request: ask for the current price or the last hour of trades, get an answer. They are simple, stateless and rate-limited, and every update costs a request. WebSocket streams push data continuously over a persistent connection; subscribe to a pair, and the exchange or provider sends every trade and every order book change as it happens, which is what real-time means in practice. Flat files are bulk historical data, days or years of trades and quotes, delivered as downloadable files for backtesting and research.
Institutional providers also offer FIX, the protocol traditional finance uses for order and market data messaging, for firms whose systems are already built around it. What real-time delivery involves once many feeds are combined is set out in Real-Time Crypto Data Aggregation: How It Works.
Where crypto market data differs from traditional market data
Anyone arriving from equities or FX expects market data to work a certain way, and crypto breaks the expectation in four places.
There is no consolidated tape. In US equities, quotes and trades from every venue are consolidated into a single official feed with a national best bid and offer. Crypto has nothing equivalent. Each exchange publishes its own data, in its own format, with its own timestamps, and there is no authority that could consolidate it and no single official price. Why that matters so much is set out in Rethinking Crypto Data.
The market never closes. There is no opening auction, no closing price, no end-of-day. Anything that traditional finance anchors to the close, valuation, settlement, reporting, has to be anchored to a chosen time instead.
Venues vary enormously in quality. Regulated equity exchanges are held to common standards. Crypto exchanges range from large, well-run and monitored to small, thin and, in documented cases, reporting volume that never happened. Bitwise's 2019 analysis presented to the SEC found that a large majority of reported Bitcoin volume at the time was fake or non-economic. The share has fallen since; the problem has not disappeared.
And the same asset has many prices. Because venues are independent and nothing forces them to agree, the same asset trades at different prices on different exchanges at the same moment, usually close, sometimes not, and furthest apart when it matters most.
Where a crypto price comes from: five questions
Everything above describes what market data is. The harder question is what a given piece of it is worth, and for the single most-used piece, the price, that depends entirely on how it was produced.
A price from one exchange shows what that exchange's order book looked like. A price averaged naively across ten exchanges shows what those ten books looked like, including whichever of them was broken at the time. A price aggregated with filtering and volume-weighting across seventy venues shows something closer to where the whole market was trading.
These are three different products, and they are routinely sold under the same word. Five questions separate them.
Where does the price come from? One venue, several, or many? Centralised only, or decentralised too? A price that draws from one exchange inherits that exchange's outages, thin moments and regional flow.
How is it aggregated? A simple average lets a small venue move the result as much as a large one. Volume weighting gives influence in proportion to real trading. Median-based methods resist single outliers. The method determines what the number means.
How is it filtered? Anomalous trades should be removed, and venues that drift from the market excluded. A source with no filtering will occasionally publish a flash crash on one exchange as if it were the market.
Is the methodology published? The calculation should be readable in full: which sources, what window, what filters, what happens when data is missing. A number whose method is unpublished cannot be audited.
How is it validated? A single-vendor feed is checked by the vendor. A price published into a decentralised oracle network is checked by independent operators with no incentive to agree with the publisher, every time it is submitted. How that works is covered in How Oracle Price Feeds Work.
Most price sources answer the first two questions and go quiet on the rest. The ones worth relying on for anything with consequences, valuation, execution, settlement, answer all five. The framework behind these questions, breadth, depth and latency, is set out in What Is Price Accuracy in Crypto.
Who uses crypto market data
The same data serves very different purposes, and each purpose has its own requirements.
Trading firms and brokers need real-time prices for execution, a market-wide reference to benchmark fills against, and historical data to test strategies. Their question is whether the price they got was fair, covered in Execution Quality in Crypto and Backtesting in Crypto.
Asset managers, funds and family offices need a defensible price at a fixed time to value holdings, calculate net asset value, and report to investors, auditors and tax authorities. Their question is whether their assets are valued fairly, covered in Net Asset Value in Crypto, Crypto Asset Valuation for Audit and Reporting and Fair Market Value in Crypto.
Oracle networks and DeFi protocols need prices that can be trusted by smart contracts acting autonomously, where an error triggers a liquidation or an exploit. Their question is whether the price is safe to settle against.
Developers and startups need one integration instead of seventy, with the normalisation, filtering and uptime handled for them. And AI agents need data they can discover and consume programmatically, with accuracy standards higher than a human user's, because no human is checking the number, covered in Crypto Market Data for AI Agents.
How to evaluate a crypto market data provider
Pulling the above together, evaluating a crypto market data provider comes down to a short list.
Coverage comes first: how many venues, which ones, and whether decentralised exchanges are included. More is not automatically better, since a provider covering 500 exchanges without filtering is no more accurate than one covering 70 with it, but too few means missing where the market actually trades. Then data types: trades, order books, derived prices, reference rates, derivatives, on-chain, and whether each is taken from its own source or inferred from something else. Then the aggregation and filtering methodology, which should be published, readable, and specific about weighting, outlier handling and contingency rules.
Delivery matters next: WebSocket for real-time, REST for snapshots, flat files for history, and latency figures that state where they are measured. So does validation, meaning whether anyone independent of the provider checks the data, and reliability, meaning uptime guarantees and what happens when a source goes down. Transparency of pricing, whether the cost is published or requires a sales conversation, is not a data quality question, but it says something about how a provider expects to be evaluated. A side-by-side of the main providers on these dimensions is in Crypto Market Data Providers Compared.
Blocksize crypto market data
Blocksize is a crypto market data provider built around the five questions above. It aggregates trades, order book quotes and DEX pool state across more than 70 centralised and decentralised exchanges into one reference price per pair, for 9,000+ instruments, with each data type taken from its own source. Anomalous trades are filtered by a statistical confidence interval; venues that drift from the market are excluded until they return; results are volume-weighted. The methodology for every product, real-time, 24-hour and 30-minute VWAP, consolidated bid/ask, AMM state prices and trade events, is published in full.
The same aggregated feed is submitted to eight decentralised oracle networks, where independent node operators validate it against consensus before it is accepted on-chain. Delivery is over WebSocket at sub-250 millisecond latency, with a 99.9% uptime guarantee, published pricing from €49 per month, and an MCP server for AI agents.
Frequently asked questions
What is crypto market data? The information generated by trading in crypto assets, trades, order books, prices, volumes and derivatives activity, collected and delivered for use in trading, valuation, risk, research and on-chain applications.
What are the main types of crypto market data? Trade data (executed transactions), order book data (standing bids and asks at up to three levels of depth), OHLCV candles, derived prices such as VWAP and mid, reference rates, derivatives data such as funding rates and open interest, on-chain data, and the metadata that ties it together.
How is crypto market data different from stock market data? Crypto has no consolidated tape, no official price, no market close, and a wide range of venue quality. The same asset trades at different prices on different exchanges at the same time, so any market-wide price has to be constructed instead of looked up.
What is the difference between Level 1, Level 2 and Level 3 order book data? Level 1 is the best bid and ask. Level 2 is aggregated depth at every price level. Level 3 is every individual order. Most crypto exchanges provide Level 1 and 2; full Level 3 is rare.
What is a crypto market data API? An interface through which a provider delivers market data to applications, typically REST for on-request queries and WebSocket for real-time streaming. Most providers offer both.
How is the accuracy of a crypto price established? By knowing where it comes from, how it is aggregated, how it is filtered, whether the methodology is published, and whether anyone independent validates it. A price that answers all five can be defended. One that answers none is simply trusted.
What is real-time crypto data? Data delivered as it is generated, typically over a WebSocket connection, instead of on request or in batches. For aggregated data, real-time means recomputed continuously as new inputs arrive, as distinct from periodic aggregates recalculated every few seconds.