Yearly Breakdown: Spotting EAs Reliant on One Lucky Year

Conclusion: Full-period numbers alone won't reveal a "lucky year"

The backtest equity curve rises steadily and the PF is high. Yet it is not unusual for most of the profit to come from one big market year, with the remaining years flat or negative. Full-period totals average this skew away.

That is why the first thing we do when evaluating an EA is a yearly breakdown: splitting profit and loss by year. Does performance survive once you remove the best year? We consider this the simplest and most effective check of robustness.

Why one lucky year can make up the whole result

Markets have "regimes"

Price action in forex and gold moves through regimes: periods of strong sustained trends, periods of tight ranges, and periods when volatility suddenly spikes. Every EA has regimes it handles well and regimes it struggles in.

A trend-following EA typically earns in bulk during a year with a big trend and piles up small losses during range-bound years. Breakout systems with a lower win rate, in particular, tend to show this pattern in our experience.

Compounding magnifies the skew

In a compounded backtest, years with big wins late in the test swell in monetary terms. The larger the cumulative return figure, the more important it is to check which year produced it.

Optimization can fit itself to one specific year

Parameter optimization tends to select settings fitted to the most volatile stretch of the test period. If the market doesn't reproduce that stretch, the results won't be reproduced either.

Practical steps for a yearly breakdown

After running a backtest in the MT5 Strategy Tester, check in this order:

1. Export the trade history from the report: Right-click the "Backtest" tab in the tester results and save the report. The trade list contains close times and profit/loss.
2. Sum profit and loss by close year: Add a year column in a spreadsheet and aggregate. To avoid the effect of compounding, we recommend running the test again with a fixed lot size.
3. Calculate PF and trade count per year too: Don't stop at profit and loss amounts; list each year's gross profit ÷ gross loss alongside its number of trades.
4. Recalculate the total without the best year: Whether performance collapses here is the heart of the check.
5. Look inside the losing years: Whether the losses are small or a deep drawdown changes the assessment significantly.

Rules of thumb

These thresholds are only guidelines, and we don't decide pass or fail on them alone.

Applying this to our own EAs

Most of our EAs are tested over long periods. For example, GOLD NEURON (XAUUSD M30) has a 22.5-year test period. The longer the test period, the more regimes it contains, and the more meaningful a yearly breakdown becomes.

On the other hand, trend-following EAs with low win rates are structurally prone to uneven profit years. MEGAMAX DONCHIAN (USDJPY H1) has a win rate of 34.98% and a maximum consecutive loss of 11. DONCHIAN ENGINE (USDJPY H1) has a win rate of 34.5% and a maximum consecutive loss of 11. For an EA of this type, we cannot rule out stretches of consecutive years with no profit. Even if the full test period is in the black, if you start using it in a year that suits its weak regime, it's entirely possible profits won't grow for a while.

Note that this article does not include a year-by-year breakdown for each EA. Measured full-period figures for all our EAs are collected in the measured ranking.

Limits and cautions

• A calendar year has no meaning to the market: Regime shifts happen mid-year. Looking at quarters or half-years can change how the skew appears.
• Not suited to EAs with few trades: If a year has only a handful of trades, that year's PF is determined almost entirely by chance.
• Past regime diversification doesn't guarantee the future: If a type of market arrives that was absent from the test period, even an EA that looked stable year by year may break down.
• Demo-account forward tests cover a short period: We publish them too, but the time since launch is still short, so a year-level assessment isn't possible. Please avoid judging regime resilience from short-term results.

Other validation perspectives are collected in EA & MT5 Know-How.

Summary

• Full-period PF and cumulative return average away the effect of one lucky year
• A yearly breakdown checks whether performance survives after removing the best year
• The steps: export trade history → aggregate profit/loss, PF, and trade count by year → recalculate without the best year → check how the losing years lose
• Trend-following EAs with low win rates are prone to uneven profit years, and ours are no exception
• Limits include the arbitrariness of calendar-year boundaries, small trade counts, and unseen regimes