Every time I evaluate a new Expert Advisor, the MT4 Strategy Tester results are the first thing I examine. A backtest report contains dozens of numbers, but most traders look at the wrong ones. They fixate on total profit while ignoring drawdown, or they celebrate a 95% win rate without realizing it signals over-fitting. Below, I walk through each metric, the values worth targeting, and the red flags that separate real performance from a curve-fitted illusion.

Step 1: Check Modeling Quality First

Before looking at any performance number, scroll to the bottom of the backtest report and find the Modeling Quality line. This single number tells you whether the rest of the report is worth reading.

• 90% -- Maximum for MT4 using "Every Tick" mode with quality M1 data. This is the minimum standard for reliable results.
• 49-89% -- Data quality issues. Results may be inaccurate, especially for scalping EAs or tight stop losses.
• 25% or "n/a" -- The test used "Control Points" or "Open Prices Only" mode. Not reliable for evaluating real performance.

If modeling quality is below 90%, I do not trust the results. I re-download history data from my broker or use the MetaQuotes history center, then re-run the test. For even higher accuracy, consider testing on MT5, which offers 99.9% modeling quality with real tick data.

How MT4 Actually Generates Its Tick Data

Unless you are feeding the tester real historical ticks, MT4 does not replay the market the way it actually happened. It reconstructs a synthetic tick stream from whatever bar data you have loaded, then fills the gap between the open, high, low, and close of each candle with an interpolated price path. Which reconstruction method it uses depends on the mode you pick before running the test, and that choice affects result accuracy more than most traders realize. The official MQL5 documentation covers the technical details of how each mode is built if you want to go deeper.

When we validate Golden Viper EA settings internally, we run every-tick mode exclusively and discard anything that comes back below 90% modeling quality, for the same reason you should: a control-points test is a connect-the-dots sketch of price action, not a simulation of it. It can be useful for a first-pass screen of dozens of parameter combinations, but never as the basis for a final decision.

Step 2: Read the Summary Metrics

Once modeling quality checks out at 90%, move on to the core performance metrics. Here is how I evaluate each one:

Total Net Profit

The bottom-line number: total gains minus total losses. It obviously needs to be positive, but a large profit figure on its own means nothing without context. A $10,000 profit with $9,000 drawdown is far worse than a $3,000 profit with $500 drawdown.

Profit Factor

Calculated as Gross Profit divided by Gross Loss. This is the single most useful metric in the entire report because it measures how much you earn per dollar risked:

• Below 1.0 -- Losing strategy. The EA loses more than it makes.
• 1.0 to 1.3 -- Marginal. Barely profitable after spreads and slippage in live trading.
• 1.3 to 2.0 -- Good range. Realistic and likely sustainable.
• 2.0 to 3.0 -- Excellent. Strong edge with good risk management.
• Above 5.0 -- Suspicious. Could indicate over-fitting or too few trades.

Maximum Drawdown

The largest peak-to-trough equity decline during the test. This tells you the worst-case scenario you should expect (and in live trading, it will usually be worse):

• Under 15% -- Conservative. Comfortable for most traders.
• 15% to 25% -- Moderate. Acceptable if returns justify it.
• 25% to 40% -- Aggressive. Requires strong conviction and proper position sizing.
• Above 40% -- Dangerous. Most traders will panic and disable the EA during a drawdown this deep.

Total Trades

The number of trades in the test period. The more of them, the more confidence you can place in the statistics:

• Under 50 -- Results are meaningless. Could be pure luck.
• 50 to 100 -- Barely acceptable. Treat with skepticism.
• 100 to 300 -- Decent sample size for initial evaluation.
• 300+ -- Strong statistical significance. Results are likely repeatable.

Expectancy and Average Trade

Win rate gets all the attention, but expectancy is the number that actually tells you whether a strategy is worth trading. Expectancy combines win rate with the size of your average win and average loss into a single figure: the average amount you can expect to make or lose per trade. A strategy that wins 40% of the time with a 2.5:1 reward-to-risk ratio has positive expectancy and can comfortably outperform a strategy that wins 70% of the time but risks three times what it makes on each winner. MT4's report calls this "Expected Payoff" and shows it as a currency amount per trade -- look for a figure that stays positive and reasonably stable when you re-run the test over different date ranges. Investopedia's overview of backtesting fundamentals covers why relying on a single summary statistic like win rate can be misleading.

Recovery Factor and Risk-Adjusted Return

Recovery Factor divides Net Profit by Maximum Drawdown, answering a simple question: how much did the strategy earn relative to the worst pain it put you through to get there? A recovery factor above 3 means the strategy made three times more than its deepest drawdown, which is a reasonable bar for something you would trust with real capital. Below 1, the strategy lost more at its worst point than it ultimately earned, which is a strategy you should not run live regardless of how the equity curve looks on the last day of the test. For a more academic take on risk-adjusted performance, the Sharpe ratio measures return per unit of volatility rather than return per unit of drawdown, and some traders like to compute both since they can disagree on borderline strategies.

Step 3: Analyze the Equity Curve

The equity curve is the graph at the top of the backtest report, showing how your account balance changed over time. I consider it the most honest part of the report, since it reveals patterns that raw numbers hide.

What a Good Equity Curve Looks Like

• Smooth, consistent upward slope with no major gaps
• Small, regular drawdowns that recover quickly
• Profitability across the entire test period, not just one lucky stretch
• Similar angle throughout -- not flat for months then suddenly profitable

What a Bad Equity Curve Looks Like

• Large vertical jumps -- most profit came from one or two trades
• Extended flat or declining periods
• Deep drawdowns that take months to recover
• Staircase pattern -- indicates optimization for specific historical events

Reading the Detailed Trade List

Below the equity curve, MT4 gives you a row-by-row list of every trade: order number, open time, type, size, entry price, stop loss, take profit, close time, and profit. Most traders skim past this table straight to the summary, but it is where over-fitting hides. Scan for a handful of trades that make up a disproportionate share of total profit -- if 5 trades out of 300 account for half the net gain, the strategy's real edge is thinner than the headline number suggests, and one missed signal in live trading could erase a meaningful chunk of expected return.

Two figures worth tracking manually if your report includes them (or if you export to a spreadsheet) are Maximum Adverse Excursion (MAE) and Maximum Favorable Excursion (MFE) -- how far a trade moved against you and in your favor, respectively, before it closed. A strategy where winning trades regularly show large MAE before turning profitable is riding stop losses close to their limit, which tends to produce more losers in live spread and slippage conditions than the backtest implied. Slippage itself is worth understanding on its own terms; Investopedia's explanation of slippage is a useful primer if the concept is new to you.

Step 4: Spot Red Flags

These are the warning signs I watch for in an EA backtest. If I see more than two of these in a single report, I move on to the next EA:

• Win rate above 90%: Often achieved through wide stop losses or martingale. The EA looks great until one catastrophic loss wipes the account.
• Profit factor above 5.0: Usually means too few trades or extreme curve-fitting to historical data.
• Max drawdown under 5%: Unrealistic unless the EA trades extremely small positions. In live markets, drawdown is always higher than backtests show.
• Fewer than 100 trades: Not enough data for statistical significance.
• Modeling quality below 90%: The tick data was interpolated, meaning stop losses and take profits may not have been triggered accurately.
• No losing streaks: Every real strategy has losing streaks. If the backtest shows none, the EA was likely optimized to avoid historical losing periods.

Common Backtesting Mistakes That Inflate Results

Beyond the individual red flags in a single report, there are process mistakes that quietly inflate almost every metric at once. These are worth understanding even if you never build or optimize an EA yourself, because they explain why a backtest that looks flawless can still fail in live trading.

Curve-Fitting Through Over-Optimization

MT4's built-in optimizer can test thousands of parameter combinations against the same historical data and hand you the single best-performing set. The problem is that with enough combinations, some set of parameters will always fit the noise in that specific data window, whether or not it captures a genuine market inefficiency. This is the classic case of overfitting: the strategy looks exceptional on the data it was tuned on and mediocre or worse on any data it has not seen. A rule of thumb: be suspicious of any parameter set with more than 4-5 tunable inputs that were all optimized simultaneously on one dataset.

Testing Only the Best-Case Historical Window

It is tempting to run a backtest over the specific date range where a strategy happened to perform best, especially when a vendor is trying to sell you on it. Always check the date range at the top of the report and ask what happens outside it. A strategy that only gets shown across a single strong trending year, without a ranging or high-volatility period included, has not really been tested.

Ignoring Spread, Commission, and Broker-Specific Costs

The default Strategy Tester settings often use a fixed, generous spread that does not reflect real trading conditions. Gold in particular can see spreads widen sharply around news events and session opens. Before trusting a backtest, set the spread and commission fields to match your actual broker's typical values, ideally using variable spread modeling rather than a fixed number, and re-run the test.

Watch for Vendor Overclaiming

Automated trading is a space where exaggerated marketing is common. Both the CFTC and the FTC publish consumer guidance warning against "guaranteed returns" and unverifiable performance claims in trading products, and the same caution applies to EA vendors who show only backtests and never publish verified live results. A hypothetical or simulated performance disclosure is standard in this industry for good reason: past backtested performance does not guarantee future results, live or simulated.

Step 5: Compare Against Live Results

The final and most important step: compare the backtest against actual live trading performance. Backtests should be treated as a first-pass filter, not as proof of profitability. Here is what to check:

• Does the live profit factor match the backtest? A small drop (10-20%) is normal. A large drop means the backtest was over-optimized.
• Is live drawdown similar? Live drawdown is typically 20-50% higher than backtest drawdown due to slippage and real market conditions.
• Are trade frequencies consistent? If the EA takes 5 trades per day in backtests but only 1 per day live, something is misconfigured.

This is exactly why we publish our Golden Viper EA results on Myfxbook -- verified live trading data that you can compare against any backtest. Our XAUUSD trading guide covers the fundamentals of the market we trade, and our broker comparison helps you choose the right execution environment.

Walk-Forward and Out-of-Sample Testing

If you optimize parameters yourself, a single backtest over the full history is not enough validation on its own. Walk-forward testing splits your historical data into sequential chunks: you optimize on the first chunk (the in-sample window), then test the resulting parameters, unchanged, on the next chunk the optimizer never saw (the out-of-sample window). You then roll both windows forward and repeat. A strategy with a genuine edge tends to hold up reasonably well out-of-sample; a curve-fit strategy usually falls apart the moment it hits data it was not tuned on. It takes longer than a single optimization run, but it is the closest thing to a live-trading dress rehearsal you can do without risking real money.

MT4 vs MT5 Strategy Tester: Key Differences

Both platforms run Golden Viper EA, and both include a Strategy Tester, but the testing engines are not equivalent. If you are deciding which platform to validate a strategy on, or wondering why the same EA logic produces slightly different numbers between the two, the differences below explain most of the gap. Our full MT4 vs MT5 comparison covers the broader platform differences beyond backtesting.

None of this means an MT4 backtest is worthless. Every Tick mode at 90% modeling quality is still a reasonable standard for evaluating an EA's logic. It simply means that when the two platforms disagree slightly on a metric, the MT5 number is generally the more trustworthy one, and it is worth re-validating a promising MT4 backtest on MT5 before committing capital.

Key Metrics Reference Table