Why Broker Feeds Differ — and What It Does to Your Backtest

Where the Differences Come From

A retail broker’s price engine aggregates quotes from its liquidity providers (banks, non-bank market makers), applies filtering, and adds a spread/markup model. Different LP panels + different engines = different feeds. The result isn’t “wrong” prices anywhere — it’s plural prices, which broker education will quietly admit: Admiral Markets’ backtesting guide notes OTC data divergence across brokers in passing (two sentences — close to the industry’s total public treatment of the subject).

What Actually Differs, Concretely

Dimension	Typical divergence	Backtest consequence
Candle highs/lows	a few tenths of a pip normally; several pips in fast markets	Wick-precise levels (equal highs, liquidity sweeps) don’t replicate exactly across feeds
Spread model	structural — raw vs marked-up, different widening behavior	Cost lines and stop-trigger distances differ (spread mechanics)
Weekend/holiday handling	some feeds trim Sunday stub candles, some don’t	Daily candle counts and indicator values diverge (gaps & rollovers)
Server clock	GMT+2/+3 standard, but not universal	Session logic and daily-candle shape shift (GMT & DST)
Stop/limit trigger side	bid vs ask conventions per instrument	Same chart, different fill moments

Which Strategies Care

High sensitivity: scalping with sub-10-pip stops; strategies entering on wick-precision levels (stop hunts, equal-high sweeps); spread-cost-dominated systems; anything trading the rollover hour. Low sensitivity: swing entries on closes; structure zones drawn in tens of pips; higher timeframe trend systems. The dividing line is the same one as in tick vs OHLC: how much of your edge lives at sub-pip precision?

The Defense: Robustness Over Precision

You cannot backtest on every broker’s history, and you don’t need to:

1. Build rules around zones, not exact prints. A level that only works to the tenth of a pip is a feed artifact, not an edge.
2. Match the data class to the account class. Testing on raw-spread-style data and trading a high-markup account (or vice versa) bakes in a systematic cost error — align them, or correct in the cost audit.
3. Stress the result. If doubling the spread or shifting every fill 0.2 pips against you kills the edge, the edge was feed-specific noise.
4. For multi-source perspective: replaying the same period on a different feed is a quick robustness probe — tools with multi-broker data make this cheap (StrategyTune streams data from multiple broker sources, so you can re-run a strategy against a second feed without any data pipeline).
5. Forward test on the actual broker before sizing up — the live-vs-backtest audit catches residual feed mismatch.

Related: tick vs OHLC · free data sources · GMT & DST

Frequently Asked Questions

Which broker has the 'correct' price feed?

None — correctness isn't defined in OTC markets. Institutional references like EBS or aggregated composites exist, but your fills happen on your broker's feed, so for backtesting purposes the most 'correct' data is whatever most resembles where you'll trade.

How big are the differences in practice?

In calm liquid majors, quotes across reputable brokers track within fractions of a pip. The divergence concentrates exactly where it hurts: news spikes, rollover minutes and flash moves, where highs/lows can differ by several pips — the same moments that decide tight stops.

Is this why my indicator values differ between platforms?

Often, yes — indicator math is deterministic, but it runs on different candles: different server clocks change where daily candles split, Sunday-stub handling changes candle counts, and feed differences nudge highs/lows. Timezone and candle-count checks usually explain it before suspecting the indicator.

Should I avoid strategies based on liquidity sweeps because of feed divergence?

Not avoid — but design them in zones rather than exact ticks, and validate on tick-level data from more than one source if possible. A sweep concept that only triggers on one broker's wick print is fragile; one defined as a zone violation with confirmation tends to survive feed differences.