Decision Framework

Should You Trust Gold Trading
Bot Backtests?

A structured decision process for evaluating any XAUUSD EA backtest. Four conditions make a backtest trustworthy. Four signals make it unreliable. Here is how to apply the framework to any result you receive.

Quick Answer

Trust a backtest if all four conditions are met: test period at least 3 years across multiple regimes, realistic ECN spread (8–15 pips), 99% modelling quality, and separate out-of-sample validation. Distrust it if any one of four red flags is present: period under 18 months, profit factor above 3.0 with under 5% drawdown, undisclosed spread, or no out-of-sample testing.

Trust if ALL 4 are true

Test period ≥ 3 years spanning multiple market regimes

Spread set to realistic ECN value (8–15 pips for XAUUSD)

Modelling quality ≥ 99% (shown in Tester report)

Out-of-sample validation on a separate date range

Distrust if ANY 1 is true

🚩

Test period < 18 months or only in one market direction

🚩

Profit factor > 3.0 with < 5% max drawdown

🚩

Spread set to 0 or not disclosed in the report

🚩

No out-of-sample test; single period optimised and tested

Run the Trust Check

Check each item that applies to the backtest you are evaluating. A single unchecked distrust signal overrides all trust signals.

Trust signals present

0 / 4 trust signals confirmed

Distrust signals present

0 / 4 distrust signals flagged

Why Even Experienced Traders Get Fooled

Confirmation bias is the most powerful force in backtest evaluation. When a trader already believes a strategy type makes sense — "breakout trading works during London session open, that is logical" — any backtest that confirms that belief bypasses critical scrutiny. The numbers look reasonable. The logic checks out conceptually. The trader proceeds.

The problem is that the numbers can look reasonable while being completely invalid. A 2.1 profit factor on 12 months of one-regime data with 0 spread looks just as convincing as a 1.6 profit factor on 5 years of multi-regime data with realistic spread — but only the second number means anything. The first is a product of favorable conditions and missing costs.

The solution is not becoming skeptical of all backtests — valid backtests are the best tool available for evaluating strategies before risking capital. The solution is a fixed checklist applied identically every time, before looking at any performance numbers. When you know the trust and distrust signals in advance, you evaluate the methodology before you evaluate the results.

What to Do with a Moderate Confidence Backtest

A backtest with three trust signals and zero distrust signals represents a common situation: the core data is valid but one validation piece is missing. Typically this means: good tick data, realistic spread, multi-year coverage — but no separate out-of-sample test. The developer ran the whole available data set for both optimisation and testing.

The appropriate response is not to demand a new backtest — request the full HTML report first, verify the three signals you believe are present are actually present (not just claimed), and then ask specifically: "Was this period also used for parameter optimisation?" If yes, request out-of-sample results or supply them yourself by rerunning the stated parameters on the most recent 6–12 months of data that were likely included in the optimisation period.

Then proceed to a 60-day demo forward test. A strategy that survives demo testing for 60 days, performing within a reasonable range of what the backtest predicted, has earned more confidence than any additional backtest analysis can provide. Live (but risk-free) data is the ultimate out-of-sample test.

The Same Framework for Your Own Backtests

This framework is not only for evaluating vendor-provided results. Apply it to your own backtests before trusting them enough to trade live. It is psychologically harder to apply to your own work — you know the effort you put in, you want the results to be valid, and you are predisposed to interpret ambiguity in your favor.

When backtesting any XAUUSD strategy, including evaluating the Goldie Razor V2.8.4, applying this same four-signal framework to any results you receive or produce is the baseline for responsible evaluation. The four trust signals are not onerous — every one of them is achievable within a standard MT5 backtesting setup. If any are missing, they are missing because someone chose not to include them, not because they were impossible to obtain.

The practical habit to build: before you look at any profit numbers in a Strategy Tester report, scroll to the header and check modelling quality, scroll to find the date range, and confirm you know whether the settings were optimised on this same data. Those three checks take thirty seconds and determine whether the rest of the report is worth reading.

Frequently Asked Questions

Goldie Razor V2.8.4

M15 breakout + H4 EMA filter — built for XAUUSD on MT5

View Goldie Razor →