Sonar Sciences
Start free
Start freeSign in
Research/Cross-venue market data & signals

Is one data feed enough to trust a backtest?

No. A single feed shows you one venue's prints, one venue's spreads and one venue's outages, and a backtest inherits every one of those quirks as if they were the market. The same instrument can print different highs and lows on two venues in the same minute, a candle store can quietly interpolate through a gap, and a broker feed can be filtered before it reaches you. A strategy fitted to one feed has partly learned that feed.

This pillar is about knowing what your sample actually contains. It covers what cross-venue data means concretely: the same instrument sampled from several venues and reconciled, so a result has to hold across sources before it counts. It also covers market regimes, because coverage in time matters as much as coverage across venues; a sample without a rate shock or a volatility spike has never tested the strategy in bad weather. Start with the cross-venue article, then the regimes piece, then the coverage comparison. If you take one habit away, take this one: before you trust a backtest, ask which feed produced the candles, and what that feed could not see.

One instrument sampled from several venues: where the feeds agree, where they diverge, and what a single source hides.

Suggested reading order

1How to align timestamps across data sourcesAligning timestamps across feeds takes explicit timezone offsets, careful daylight-saving handling, awareness of exchange clocks, and one declared bar-boundary convention. RFC 3339 gives unambiguous formatting and Unix time a common numeric base.5 min2How to build a point-in-time datasetA point-in-time dataset preserves each value as it was known at each timestamp, instead of overwriting history with later revisions. In backtesting, this prevents look-ahead bias because the simulator can query only the records whose observation and storage times are no later than the simulated decision time. An effective design uses append-only versioning, explicit time fields, time-aware queries, and audit checks that verify no future revisions enter the pipeline.6 min3How to choose candle timeframes for researchMatching candle timeframe to a strategy's signal horizon and holding period is supported as a research best practice because bar aggregation can distort cross-venue price information, while indiscriminate timeframe tuning increases multiple-testing and overfitting risk. The framework is supported, but it does not provide explicit matched-vs-mismatched numerical comparisons, so no quantified claim should be made beyond that.4 min4How to clean tick dataA cleaning workflow for cross-venue tick data: deduplication, event ordering, bad-print filtering, and a traceable record of every transformation. Each stage exists because a defect it removes would otherwise flow silently into every candle, indicator and backtest built on top.3 min5How to detect a regime change in dataCross-venue financial data can exhibit changing structure over time, and any regime-detection workflow should be checked for overfitting and multiple-testing bias.3 min6How to detect bad prints and outliersStatistical filters can flag suspicious ticks, and cross‑venue confirmation can help decide whether an extreme move is a real market event or a venue‑specific anomaly. Validation of downstream effects with an overfitting audit and deflated Sharpe ratio is recommended. Quantified reliability requires empirical evidence such as labeled bad‑print precision and recall or case studies of retained crash events.5 min7How to detect interpolated candlesThe sources support using cross-venue reconciliation and research-audit methods to investigate suspicious candles. The strongest supported conclusion is that patterns such as zero-range bars, repeated closes, and volume anomalies are useful anomaly flags, and validation should come from comparison with independent feeds and robustness checks on backtest results.4 min8How to detect stale quotes in a feedThe concept and mechanisms are explained.3 min9How to estimate slippage from dataSpread, depth, and volatility are conceptually relevant inputs for bounding execution cost at the decision timestamp, with spread representing immediate crossing cost, depth representing liquidity consumption, and volatility representing timing risk.4 min10How to find gaps in historical dataA careful, venue-aware approach to finding historical data gaps involves defining session calendars, comparing observed bar counts to expected session counts, and verifying sequential integrity of timestamps or sequence fields. Flawed inputs can distort research conclusions, so data-quality checks are essential. Concrete session-calendar examples, missing-bar case studies, and before-and-after statistical validation are not available.5 min11How to handle exchange outages in a datasetMethodology: venue outages should be preserved as unavailable intervals rather than filled, because venue-specific tradability matters and biased preprocessing can inflate apparent strategy quality. No outage timestamp list, side-by-side backtest metrics, or historical case studies are available, so those claims cannot be made.3 min12How to reconcile data from multiple venuesMulti venue reconciliation means aligning timestamps and comparing prints across sources so only cross‑confirmed events influence a signal. This reflects a broader methodological standard of cross‑source validation and robustness testing.3 min13How to store market data for researchStable, versioned historical data and detailed experiment logs enable quantitative research to remain auditable and repeatable. Making dimensions such as venue and instrument explicit supports cross-venue analysis.4 min14How to verify a data vendor's historyBefore trusting a vendor dataset, quantitative teams should verify sampled records against exchange archives, run a full gap census, and maintain revision logs.5 min15Why a single data feed biases a backtestSingle-venue backtests can absorb venue-specific microstructure quirks, and cross-venue validation helps reveal whether a strategy is robust or merely fitted to one feed. Sonar’s overfitting audit framework and its explanation of the deflated Sharpe ratio both reinforce the need for stricter validation when many variations have been tried. Exact comparative statistics, numerical performance decay, or specific audit outputs should not be claimed without additional cited evidence.4 min16Why broker feeds differ from exchange feedsBroker-visible market data should not be assumed to match raw exchange data. Venue selection and data construction alter observed spreads, prices, and liquidity, and those choices can distort backtest conclusions or inflate confidence if treated as fixed truths.5 min17Why the same candle differs across venuesCandles are built from executed trades within a time interval, so their high, low, close, and volume depend on the specific trade stream used. Because each venue reports only its own trades, the same instrument can show different candles across venues for the same time bucket. Backtests that use a single venue’s candle data therefore inherit that venue-specific view of price extremes, closing prints, and volume.4 min18Why consolidated prices hide venue behaviourConsolidated price feeds merge multiple venues into a smoother series that can hide venue-specific spreads, quote gaps, and outages. That smoothing matters because orders execute on individual venues, not on an average market view. Using consolidated prices alone can therefore understate spread variability, mask interruptions, and make backtests appear more stable than live execution conditions.3 min19Why execution research needs tick dataA cautious methodological argument: execution research is microstructure‑sensitive, and Sonar’s cross‑venue data and validation tooling are aligned with that reality. Specific empirical tests, backtests, or queue‑reconstruction case studies required to prove that candle‑based execution research mis‑estimates performance versus tick‑level analysis are not included.4 min20Why funding rates matter in perpetual backtestsFunding payments in perpetual futures are real trade level cash flows that can materially change realized PnL. Because funding varies across venues, contracts, and time, backtests that ignore it can misstate performance, especially for longer holding periods.3 min21Why liquidity varies by session and venueLiquidity changes with participation and venue characteristics. Major sessions and primary venues usually support more nearby resting interest, so the same order can require less price movement than during quieter hours or on thinner venues.3 min22Why perpetual and spot prices divergeThe infrastructure and validation framework for studying perpetual-versus-spot divergence includes normalized cross-venue data to measure price gaps alongside funding, open interest, liquidations, and venue-level stress, while overfitting audits and the Deflated Sharpe Ratio help test whether any apparent predictive relationship is real. Direct empirical proof that the gap predicts future spot moves is not provided. Based on this information alone, the divergence is best treated as a measurable research feature and hypothesis, not a proven forecasting signal.6 min23Why market regimes belong in your sampleSample completeness should include both venue coverage and regime coverage. Cross‑venue research shows market behavior differs across venues; the Backtest Overfitting Audit detects fit‑to‑sample artifacts; and the Deflated Sharpe Ratio provides a stricter lens on risk‑adjusted performance after multiple testing.4 min24Why reported volume can misleadReported volume can mislead when datasets contain duplicated records or economically non‑distinct activity, so volume should be cross‑checked rather than accepted at face value. Researchers should perform cross‑venue reconciliation, deduplication, and audit research inputs.3 min25Why timezone errors corrupt daily candlesDaily candles are session-defined aggregates, not universal objects. Cross‑venue timestamp normalization and session conventions are critical; misalignment can change which observations fall into a given day, corrupting OHLC values. Researchers should use audit methods to test whether backtest behavior may reflect data‑construction errors.4 min
Tool · free, no signupBacktest Overfitting AuditEight questions about your sample, your process, and your cost model. Answer them and you get a written verdict you can keep.

Terms used in this pillar