Same pair, two data sources, two different histories
Posted: Mon Nov 02, 2026 9:50 am
Same pair, two data sources, two different histories
Open a EURUSD one minute chart on two different brokers and zoom in on the same hour. The highs and lows will not match. Sometimes the difference is a fraction of a pip, sometimes more.
Each broker builds its charts from its own quote stream, with its own liquidity providers and its own filters. There is no single true price in spot forex, only a collection of quotes from various places.
For a long term trader the differences wash out. For a scalper, they decide whether a strategy works. A setup that triggers on a 0.3 pip break above a high can fire on one feed and never on another.
Strategies tuned on one data source can underperform when run live on a different broker. The problem is not overfitting in the usual sense, but data mismatch.
Best practice is testing on data from the broker actually used for trading, and ideally recording that broker's live ticks for a few months before relying on a backtest.
A short experiment: export one hour from each source, line up the ticks by timestamp, and plot the difference in bid. A flat line near zero is a good sign. A noisy band with occasional large excursions explains why the same setup behaves differently on each.
Some traders go further and keep recordings from two or three brokers at once. The cost is some storage and a bit of scripting, and the reward is a clear sense of how much the data source affects the strategy.
Because of this, a backtest is best seen as a result for one broker's feed, not for the pair in general. The honest label on a test result includes the data source, the period, and the spread model used.
How different are the highs and lows between the data source used for testing and the live broker?
Open a EURUSD one minute chart on two different brokers and zoom in on the same hour. The highs and lows will not match. Sometimes the difference is a fraction of a pip, sometimes more.
Each broker builds its charts from its own quote stream, with its own liquidity providers and its own filters. There is no single true price in spot forex, only a collection of quotes from various places.
For a long term trader the differences wash out. For a scalper, they decide whether a strategy works. A setup that triggers on a 0.3 pip break above a high can fire on one feed and never on another.
Strategies tuned on one data source can underperform when run live on a different broker. The problem is not overfitting in the usual sense, but data mismatch.
Best practice is testing on data from the broker actually used for trading, and ideally recording that broker's live ticks for a few months before relying on a backtest.
A short experiment: export one hour from each source, line up the ticks by timestamp, and plot the difference in bid. A flat line near zero is a good sign. A noisy band with occasional large excursions explains why the same setup behaves differently on each.
Some traders go further and keep recordings from two or three brokers at once. The cost is some storage and a bit of scripting, and the reward is a clear sense of how much the data source affects the strategy.
Because of this, a backtest is best seen as a result for one broker's feed, not for the pair in general. The honest label on a test result includes the data source, the period, and the spread model used.
How different are the highs and lows between the data source used for testing and the live broker?