Check every red flag present in the backtest you're evaluating
1. Backtest covers only 1–2 years of data
Cherry-picked profitable period
Why It's Suspicious
Any random strategy will find a 12–24 month window where it appears profitable. XAUUSD went through multiple distinct regime types between 2018 and 2024 — trending up (2019–2020), extreme volatility (2020 COVID), a sideways range (2021), a commodity surge (2022), and mixed conditions (2023–2024). A backtest that only includes the 2020 bull run will show almost any gold strategy looking excellent.
What to Ask the Vendor
"Can you show me 5+ year backtest results? What does performance look like before 2020?"
2. Profit factor above 3.0
Suspiciously smooth — likely over-fitted
Why It's Suspicious
A profit factor above 2.5 on a multi-year backtest is possible but rare and should be scrutinised carefully. Profit factors above 3.0 for strategies trading volatile instruments like XAUUSD are almost always the result of curve-fitting to historical data. Real-world XAUUSD strategies typically show profit factors between 1.3 and 2.0 across extended test periods.
What to Ask the Vendor
"What does the profit factor look like on out-of-sample data? What about a walk-forward test?"
3. Maximum drawdown under 5%
Real EAs have larger drawdowns — zero means something is wrong
Why It's Suspicious
XAUUSD is a volatile instrument. Any strategy trading it will experience drawdowns. A backtest showing a maximum drawdown of 3–5% across 3+ years either has an unrealistically small lot size (making the strategy impractical), a very short test period where luck kept drawdown minimal, or spread/slippage set to 0 so losses are artificially smaller.
What to Ask the Vendor
"What lot size was used in this backtest? What was the spread setting?"
4. Spreads set to 0 in backtest
Fake results that hide the true execution cost
Why It's Suspicious
This is the most common form of backtest manipulation — often unintentional. MT5's Strategy Tester defaults to a fixed spread, and some users set it to 0 to simplify testing. On XAUUSD with a real ECN spread of 10–15 pips, running a backtest with 0 spread artificially inflates every winning trade and reduces the loss on every losing trade. The result can be the difference between a profitable and unprofitable strategy.
What to Ask the Vendor
"What spread value was used in this backtest? Can you re-run it with a realistic 12–15 pip spread?"
5. Modelling quality below 90%
Tick data not used — results are unreliable for active EAs
Why It's Suspicious
MT5's Strategy Tester shows a "Modelling Quality" percentage. Below 90% means the EA was tested on reconstructed price data (1-minute bars interpolated as tick data) rather than real tick data. For any EA that trades within a candle — uses intra-bar logic, trailing stops, or is sensitive to spread timing — low modelling quality produces results that cannot be trusted.
What to Ask the Vendor
"What is the modelling quality percentage shown in your Strategy Tester report?"
6. Same EA profitable on 10+ different symbols
Optimised for everything = optimised for nothing
Why It's Suspicious
When a vendor shows an EA performing well on XAUUSD, EURUSD, GBPJPY, and five crypto pairs simultaneously, the settings have been independently optimised for each. This means the EA has no consistent underlying logic — it has been fitted to each instrument's historical data separately. The probability that all of these over-fitted configurations continue to work going forward is essentially zero.
What to Ask the Vendor
"Was the same parameter set used on all instruments, or were settings optimised separately for each?"
7. No losing months shown
12+ months with zero losing months is statistically impossible for most strategies
Why It's Suspicious
If an EA vendor shows a 12-month backtest or live track record with no losing months at all, something is wrong. XAUUSD has periods of low-quality ranging behaviour, news-event losses, and false breakout streaks that affect every strategy. An EA that shows only green months is either: cherry-picked, run at an unrealistically low lot size, or the losing months have been excluded from what was shared.
What to Ask the Vendor
"Can you share the full month-by-month breakdown including losing months? What was the worst single month?"
8. Results exclusively from 2015–2018
Gold trended strongly for 3 years — trend-following EAs look amazing
Why It's Suspicious
Between 2015 and 2018, gold had an extended range from roughly $1,050 to $1,350 with a clear eventual upward trend. Any trend-following or breakout EA that runs during a prolonged trend will generate impressive backtests for that period. The test of whether an EA is genuinely robust is how it performed during the 2019–2021 choppy period, the 2020 COVID spike-and-crash, and the 2022–2023 mixed conditions.
What to Ask the Vendor
"Why does the backtest only cover this specific period? Can you show pre-2019 and post-2022 results?"
Backtest Quality Score
0 / 8
red flags detected
No red flags — backtest looks clean
How to Backtest XAUUSD EAs
Without Getting Fooled
Published 15 June 2026 · 12 min read
Misleading XAUUSD EA backtests share 8 common red flags: too short a test period, unrealistically high profit factors, no losing months, zero-spread settings, low modelling quality, and cherry-picked date ranges. Use the fraud detector above to score any backtest before trusting it. Three or more red flags means you should request significantly more evidence.
Why Most Backtest Manipulation Is Unintentional
When you see a suspiciously perfect backtest, the most likely explanation is not that the vendor is deliberately deceiving you. Most EA sellers genuinely believe their results — they optimised their EA on available data, saw impressive numbers, and convinced themselves the strategy is robust.
The problem is the process itself. Optimising settings on historical data until the results look good, then presenting those results as evidence of a good strategy, is a circular argument. The settings look good on that data because they were selected to look good on that data. This is not fraud — it is a methodology error that most EA developers make because they were never taught about curve-fitting and out-of-sample validation.
The Difference Between Genuine and Manipulated Backtests
Manipulated / Curve-Fitted
- ✗Settings optimised on the same data being tested
- ✗Spread set to 0 or unrealistically low
- ✗Test period selected for good results
- ✗No out-of-sample validation period
- ✗Shown only on one instrument where it works
Genuine / Robust
- ✓Default or walk-forward validated settings
- ✓Realistic ECN spread included
- ✓5+ year period spanning multiple regime types
- ✓Explicit out-of-sample period shown separately
- ✓Performance consistent across different market conditions
What a Trustworthy XAUUSD EA Backtest Actually Looks Like
For Goldie Razor V2.8.4 specifically: when evaluating it, look for backtest results that span 2018–2024 to capture the pre-COVID consolidation (2018–2019), the COVID spike and recovery (2020), the 2021 drift, the 2022 commodity surge, and the 2023–2024 mixed period. Results that hold up across all five of those distinct regime types are meaningful evidence. Results from any one sub-period are not.
The gold standard for any XAUUSD EA backtest evidence:
5+ years of data
Spanning at least one trending year, one range-bound year, and one high-volatility event year
Modelling quality 99%
Tick data used, not reconstructed from OHLC bars
Realistic spread included
Minimum 12 pips fixed for XAUUSD — ideally variable spread matching your ECN broker
Out-of-sample period labelled
At least 12 months of data explicitly held back and tested only once after optimisation
Month-by-month breakdown
Including losing months — any presentation that only shows profitable months is incomplete
Monte Carlo analysis
Shows range of possible maximum drawdowns under different loss sequence orderings
The Expected Gap Between Backtest and Live Performance
Even a genuinely well-constructed backtest will not match live trading exactly. This is normal and expected. The question is how large a gap to expect.
| Metric | Backtest | Realistic Live Expectation |
|---|---|---|
| Profit Factor | 1.8 | 1.3–1.5 (30% reduction typical) |
| Maximum Drawdown | 15% | 18–22% (30–50% higher) |
| Win Rate | 62% | 57–60% (5% reduction typical) |
| Avg Trade Duration | Same | Same — not affected by live conditions |
| Trade Count per Month | Same | 5–10% fewer (spread filter blocks more in live) |
A 30% reduction in profit factor from backtest to live is normal and does not mean the backtest was misleading. If your EA shows a backtest profit factor of 1.8 and lives at 1.3 — that is expected and acceptable. If it shows 1.8 in backtest and 0.7 live, that is a red flag indicating either curve-fitting or unrealistic backtest assumptions.
Related Reading
How to Backtest a Gold EA Properly
The correct end-to-end methodology — what a good backtest looks like.
Backtesting a Gold Scalping Bot
Scalping-specific backtesting with tick data and realistic spread.
Questions to Ask Before Buying a Gold EA
How to interrogate a vendor's backtest claims effectively.
What Documentation to Demand Before Buying
The complete list of backtest documentation to request from any vendor.
Why EAs Stop Working After You Buy Them
When a backtest-passing EA fails live — and why it happens.
Frequently Asked Questions
A minimum of 3 years, and 5+ years is strongly preferred for XAUUSD strategies. The reason is regime diversity: XAUUSD has gone through meaningfully different market conditions in each of the years between 2018 and 2024. A 3-year backtest that covers 2019–2021 captures the COVID spike, the 2019 consolidation, and the post-COVID drift. A 5-year backtest from 2019–2024 adds the 2022 commodity surge and the 2023–2024 mixed period. Each additional regime is evidence that the strategy is not just optimised for one type of market.
Yes, and this happens regularly. A backtest profit factor of 2.5 with zero spread and 1-minute OHLC bars might correspond to a live profit factor of 1.2 once real spread (10–15 pips on XAUUSD), slippage (1–3 pips per trade), and commission are factored in. Additionally, backtest equity curves are deterministic — they cannot replicate the real emotional pressure of a live drawdown, which often causes traders to interfere with the EA at precisely the wrong moment. A high backtest profit factor is a necessary but not sufficient condition for a live-profitable strategy.
A trustworthy XAUUSD EA backtest should include: (1) 5+ years of data, (2) real tick data (modelling quality 99%), (3) realistic spread (at minimum 12 pips fixed, ideally variable), (4) slippage set to 2–3 points, (5) an out-of-sample validation period explicitly labelled as such, (6) full month-by-month breakdown including losing months, (7) results at the specific lot size and risk settings you would actually use, and (8) a Sharpe ratio or similar risk-adjusted return metric alongside raw profit.
Three main reasons. First, sellers optimise settings on historical data before release — this process naturally produces a configuration that performs well on that data but may not generalise. Second, transaction costs in backtests are often lower than real trading conditions, especially on XAUUSD. Third, market regimes change. An EA optimised on 2019–2022 data faces a different XAUUSD environment in 2024–2026. The best sellers acknowledge this and demonstrate out-of-sample validation and live forward testing alongside their backtests.
A Monte Carlo test takes the sequence of individual trade results from a backtest and randomly shuffles them hundreds of times to model different possible orderings of wins and losses. This matters because the order in which wins and losses occur affects drawdown significantly. A strategy that had 10 consecutive losses in one specific month might have had those losses spread across the year in a different scenario — producing a dramatically different equity curve. Monte Carlo testing shows the range of possible outcomes given the same trade results, providing a more realistic picture of maximum possible drawdown.
Goldie Razor V2.8.4
M15 breakout + H4 EMA filter — built for XAUUSD on MT5