Crypto Chart Data Across Exchanges: Compare Without Mixing Markets

Two charts for the same asset can both be accurate and still disagree because they may not describe the same market. Crypto chart data is defined by a venue, trading pair, instrument, price field, time-bucketing rule and aggregation method—not by a ticker alone. Use the series that matches the research question, keep those definitions fixed across the sample, and record enough provenance for another person to reproduce the chart. Consistency makes comparisons defensible; it does not make one dataset universally correct or predictive.

Why crypto chart data can disagree

A displayed price is an observation from a particular market. Each exchange has its own order book, participants, access rules, fees, inventory and latency. Arbitrage connects venues, but it is neither costless nor instantaneous. A temporary price difference may therefore be genuine rather than a bad print.

The quote asset matters just as much as the base asset. BTC/USD, BTC/USDT and BTC/EUR are three different pairs. Treating one USDT as one US dollar hides any stablecoin premium or discount; comparing a euro quote with a dollar quote adds foreign-exchange movement. A 2026 BIS working paper, using four USD-pegged stablecoins against 27 fiat currencies across 64 exchanges from 2021 through 2025, documented persistent gaps between the cost of dollar exposure through stablecoins and through conventional FX. That study is about stablecoin-based FX, not every crypto pair, but it shows why quote conversion is an economic choice rather than cosmetic relabeling.

Asset identity can also be ambiguous. Tickers may be reused; wrapped assets can exist on several networks; a token may migrate to a new contract; and derivatives can share a familiar symbol with the underlying asset. Record a stable identifier where possible: network and contract address for a token, plus the exchange's exact product or contract ID.

Liquidity explains many short-lived differences. A deep market may update frequently around a narrow spread, while a thin market can remain stale until its next trade and then print a larger move. Highs and lows are especially sensitive because one small execution can set an extreme. A close is simply the last qualifying observation in a bucket; it is not an official universal value for the asset.

Separate spot, perpetual, last, mark and index prices

Before comparing candles, classify the instrument. A spot trade exchanges the base asset for the quote asset. A perpetual contract is a derivative without a scheduled expiry; it has its own order book and can trade above or below spot. Funding, leverage, collateral, participant positioning and contract design all affect that market. A spot versus perpetual chart is therefore not an apples-to-apples price comparison, even when both labels contain the same asset and quote symbol.

Then identify the price field:

• Last price is the price of the most recent execution on that specific order book. Ordinary trade candles are usually constructed from these executions.
• Index price is a rules-based estimate derived from specified underlying markets. Its components, weights, conversion rules and fallbacks belong to the index methodology.
• Mark price is a platform-defined fair-price measure commonly used for risk calculations. It is not necessarily an executable price or the last trade.

Definitions are venue-specific. For a current official example, Bybit's mark-price documentation, updated in March 2026, describes its perpetual mark as a spot index plus a funding-basis component and says the platform uses it for liquidation triggers and unrealized profit-and-loss measurement. Its documented formula and fallbacks should not be assumed to apply elsewhere. A platform may let users switch one chart among last, mark and index series, so even two screenshots from the same contract can show different wicks.

Volume needs the same discipline. Spot base-asset volume, quote-currency turnover, contract count and derivative notional are not interchangeable. Never splice spot price with perpetual volume—or last-price candles with mark-price candles—without a defined research reason and an explicit label.

How OHLCV candle rules change the picture

OHLCV summarizes observations inside a time bucket: first price, highest price, lowest price, last price and volume. Every term depends on implementation.

• Boundary: a daily candle closing at 00:00 UTC groups trades differently from a candle based on another timezone. Weekly start days and daylight-saving treatment can also differ.
• Timestamp convention: an API may label a candle by its opening time, closing time or an interval endpoint. Align the meaning, not just the displayed timestamp.
• Input: trades, mid-quotes, mark prices and index observations produce different OHLC values.
• Precision: timestamp resolution and rules for trades exactly on a boundary can move an execution from one candle to the next.
• Volume unit: base volume, quote volume and contracts can show different magnitudes and require different conversions.

The open candle is provisional. New observations can change its high, low, close and volume until the interval ends. Kraken's current OHLC endpoint documentation explicitly says its last array entry is the current, not-yet-committed timeframe and is always present. A backtest that repeatedly saved that row before completion would not have the same data as a later historical download.

Missing crypto candles do not always mean a failed request. Coinbase's Exchange candle documentation warns that historical rates may be incomplete and that no data is published for intervals with no ticks. Another provider may insert a zero-volume candle and carry the previous close forward. Both representations can be internally coherent, but they differ in row count and in calculations that assume equally spaced observations.

Choose a gap policy before analysis. Preserve a no-trade interval as missing when studying trade formation; insert a flagged zero-volume carry-forward only when a regular grid is required and the method permits it. Do not turn an outage or unknown interval into an observed flat market. Also exclude the active candle, preserve raw downloads and note whether the provider can revise history.

Exchange price versus aggregate price

An exchange candle answers, “What traded on this product here?” An aggregate answers a different question: “What value results after applying this provider's venue selection and calculation rules?” The provider may convert quotes, screen outliers, weight venues by volume or liquidity, carry prices through gaps, and change constituents. Two aggregates with the same ticker can therefore disagree.

A 2025 paper hosted on the SEC website as a written submission, Aggregate Confusion in Crypto Market Data, examined eight widely used providers over data from November 2018 to October 2024. Its authors found substantial labeling and cross-provider measurement problems; reported metrics commonly differed by more than 5%, with volume discrepancies particularly frequent. The paper also shows why “aggregate” is not a complete provenance description. It is an authors' research submission hosted by the SEC, not a declaration that every provider or every record is wrong.

Transparent benchmarks demonstrate how much specification a meaningful aggregate requires. The CME CF reference-rate FAQ says its once-daily rates aggregate executed spot trades during a one-hour window, divide that window into twelve five-minute partitions, calculate a volume-weighted median for each partition and equally weight the twelve results. That output is designed for a benchmark purpose. It is not the same object as a continuous last price on any constituent exchange.

Choose a single-venue series when the question concerns executions, liquidity or a strategy on that venue. Choose a documented multi-venue benchmark when the question needs a broader reference value. For cross-provider robustness, compare results across preselected series; do not switch to whichever source makes a pattern look strongest.

A decision framework for selecting a series

The preferred series is the one whose measurement target matches the question. It may still contain errors, so selection and quality control are separate steps.

Worked example: a controlled four-chart comparison

Suppose four completed hourly charts show these illustrative closes for the same interval: Venue A spot BTC/USD at USD 100,000; Venue B spot BTC/USDT at USDT 100,250; a BTC/USDT perpetual at a 100,420 last price and a 100,180 mark; and a provider's USD aggregate at 100,140. These numbers are hypothetical, not current market quotations.

1. Freeze the interval. Confirm all candles cover exactly 12:00:00 through 12:59:59 UTC and exclude the still-forming 13:00 candle.
2. Classify each series. Keep Venue A spot, Venue B spot, perpetual last, perpetual mark and provider aggregate in separate columns. Do not average them.
3. Normalize the quote deliberately. If an independently selected, timestamp-matched USDT/USD rate is 0.9990, Venue B's spot close converts to USD 100,149.75. Record the conversion source and formula. The initial USD-versus-USDT visual gap is now much smaller, but the conversion itself adds a data dependency.
4. Measure like with like. Within the USDT group, the perpetual last is about 0.17% above Venue B spot, while the mark is about 0.07% below it. Those are two legitimate comparisons with different meanings: traded derivative basis versus the platform's risk reference.
5. Interrogate the aggregate. Check constituents, weighting, outlier rules, timestamp and revision policy. Without them, 100,140 is a displayed result, not a reproducible method.
6. Repeat over a fixed window. Pull the same completed intervals for a predeclared period, flag gaps and compare distributions of differences rather than selecting one dramatic candle.

The exercise does not declare a winner. It identifies which differences disappear after alignment and which remain because the markets or price definitions truly differ.

Source-provenance template and checklist

Attach this compact record to every research dataset:

• Asset identity: name, canonical symbol, network and contract address or other stable identifier.
• Market: exchange, product ID, base/quote pair, spot or derivative, linear or inverse contract, settlement and collateral asset.
• Price basis: last trade, bid/ask midpoint, mark, index or named benchmark; include a link to the methodology.
• Candles: interval, timezone, boundary convention, timestamp meaning and completed-candle rule.
• Volume: base, quote, contracts or notional, with the conversion formula if used.
• Acquisition: API endpoint or vendor dataset, retrieval time in UTC, requested coverage, pagination procedure and raw-file checksum.
• Quality policy: duplicate handling, missing-interval rule, outage flags, outlier tests, revisions and known coverage limits.
• Transformations: quote conversion source, resampling method, rounding and any constituent or identifier mapping.

Before publishing a result, ask: Are all rows the same asset and market definition? Are quotes and volume units consistent? Are timestamps aligned? Is the active candle removed? Are gaps distinguishable from zero trading? Can raw inputs be recovered? Would the conclusion survive one reasonable, preselected alternative source? If any answer is no, qualify the result or repair the dataset before interpreting the chart.

Consistency is not universal correctness

There is no single crypto price detached from venue, instrument and methodology. A local last price may be correct for execution analysis and unsuitable for portfolio valuation; a benchmark may be appropriate for broad comparison and unsuitable for reconstructing a fill. The research objective determines the measurement target.

Keep one definition fixed, document it, and run sensitivity checks instead of silently mixing series. If a finding changes materially when a defensible alternative dataset is used, that sensitivity is part of the result. Clean provenance improves comparability and error detection, but it cannot turn historical candles into a forecast or remove the risks of trading.

Sources and further reading

• Aggregate Confusion In Crypto Market Data — U.S. Securities and Exchange Commission, 2025-10-26
• Stablecoin flows and spillovers to FX markets — Bank for International Settlements, 2026-03-27
• Get product candles — Coinbase Developer Documentation, Accessed 2026-08-01; no update date displayed
• Get OHLC Data — Kraken Developers, Accessed 2026-08-01; no update date displayed
• Mark Price (Perpetual and Expiry Contracts) — Bybit Help Center, 2026-03-24
• FAQ: CME CF Single-Asset Cryptocurrency Benchmarks — CME Group, Accessed 2026-08-01; no update date displayed

Editorial note: This article is general educational information, not personalized financial, accounting, tax, or legal advice. Product capabilities and obligations can change; verify current facts and consult a qualified professional where needed.