Tick Data Quality and Its Impact on EA Backtest Reliability: Three Factors That Determine Verification Accuracy

Many traders have experienced EA backtest results that look nothing like live performance. The primary cause is the quality of tick data used in backtesting. Three variables — modeling quality, spread handling, and the source of historical data — have a major impact on verification reliability. This article breaks down each of these factors from a quantitative and structural perspective.

What Is Tick Data, and Why Does It Determine Backtest Reliability?

In FX backtesting, an EA’s logic is verified by replaying historical price data. The granularity of price data used at this stage determines the accuracy of the results.

Tick data is the raw Bid/Ask data recorded each time a trade is executed. By contrast, MT4’s default backtesting uses ‘1-minute bar interpolation,’ which artificially reconstructs ticks from the four OHLC values of each 1-minute bar. The coarser the data granularity, the less accurately ‘spread widening, slippage, and order rejection’ — the real-world conditions an EA faces — are reproduced, causing backtest results to appear artificially favorable.

MT4’s Strategy Tester offers three modeling methods:

However, when relying solely on MT4’s built-in historical data, even selecting ‘Every Tick’ mode does not use actual ticks — it still uses M1-interpolated data. To achieve true tick-level accuracy, you must separately obtain CSV data from a tool such as Tick Data Suite or a provider such as Dukascopy.

Backtest reports display a metric called Modeling Quality (%). This is calculated as: reproduced tick count / theoretical tick count x 100. A common benchmark is 90% or above as a passing threshold, but it is important to note that even a high figure does not eliminate divergence from the live market if external tick data is not being used.

A Concrete Example: How Much Can Modeling Method Differences Affect Results?

The following is a simplified comparison from our internal verification samples, showing results for the same EA tested over the same period using all three methods.

With Open Prices Only, the total number of trades drops sharply and both PF and DD show artificially favorable values. This is not overfitting — it is a ‘numerical distortion’ caused by low price resolution that fails to reproduce spread widening and slippage. For a proper understanding of PF, see also Profit Factor (PF): Benchmarks and How to Read the Metric Correctly.

Spread settings also matter. MT4 backtests use a fixed spread, but in the live market spreads can widen to several times — or even ten or more times — their normal level during news events. When an EA designed around a fixed 2-pip spread incurs unexpected losses during live spread widening, this gap is the cause. For Gold and certain JPY cross pairs, where volatility is high, this risk is even greater.

Slippage is another easily overlooked factor. For scalping EAs in particular, even a few pips of slippage can shift the break-even point significantly. Excluding slippage from a backtest tends to overstate PF. Additionally, ECN accounts charge a commission on every executed trade, and this cost must also be incorporated into the backtest.

Common Pitfalls for Beginners

• Mistaking MT4 built-in Every Tick for real ticks: Selecting ‘Every Tick’ in MT4’s Strategy Tester is still nothing more than M1 interpolation unless you have loaded external tick data. Even if the modeling quality displays as 99%, divergence from actual market behavior remains. Obtaining and configuring external data is an essential step.
• Drawing conclusions from fixed spreads alone: Testing with a fixed spread of 2-3 pips serves as a reference for normal market conditions, but it cannot reproduce spread widening during key speeches, economic releases, or periods of low liquidity. For Gold and certain JPY cross pairs, spreads can widen by 20-50 pips around data releases. As noted in Key Metrics to Examine When Verifying a Gold EA, the more volatile the currency pair, the greater this risk.
• Backtest period too short or skewed toward one market environment: An EA verified only over a 3-month strong-trend period may fail in ranging markets. The standard approach is to test over at least 3 to 5 years, covering multiple market cycles. Including the 2020 COVID shock and the sharp rate hike environment of 2022 helps reveal whether an EA can withstand stress conditions.
• Omitting commissions: ECN accounts charge a commission on every executed trade. Excluding commissions from a backtest can easily inflate reported PF by 10-20% above the actual figure. The impact of omitting commissions is especially pronounced for scalping EAs due to their high trade frequency.
• Evaluating martingale/grid EAs by drawdown alone: Strategies that carry floating losses before ultimately closing in profit have a structural tendency to underreport DD in backtests. Always verify the maximum number of open positions, maximum floating loss, and margin buffer before a stop-out. For the risk structure of martingale-type EAs, see Why Martingale EA Drawdowns Become So Deep.

FX AI Lab’s Assessment

Our lab uses ‘multi-period verification with external tick data’ as the standard for assessing EA reliability. We apply this same standard to our own EAs currently under verification, and we plan to record and publicly disclose the sources of divergence between backtest conditions and live trading conditions. Verification reports are published on an ongoing basis in the Resource Library. To join the copy trade verification using an HFM demo account, register via the HFM Registration Form. For inquiries or feedback, use the Contact Form.

This article is provided for informational purposes only and does not constitute a solicitation to purchase or use any specific financial product or EA. FX trading involves the risk of loss, including loss of principal. For details, please review our Risk Disclosure.