You run a backtest, and MT4 spits out a wall of numbers. Net profit looks decent. The equity curve goes up. Great — ship it to live?
Not so fast. We review backtest reports every single day at FXTool, and we've learned to be skeptical of the ones that look the best. A report showing 300% annual return with 5% drawdown isn't a goldmine — it's a warning sign. The report that tells you an EA earned $20,000 but had a $3,000 drawdown along the way is far more useful than one showing $50,000 with a $40,000 drawdown. Same positive ending, completely different stories.
The whole point of a backtest report is to answer three questions: does this strategy make money, is the ride tolerable, and how much can go wrong? Here's how to read those answers from the numbers.
The metrics that actually matter
Net profit (and why it's not enough)
Everyone looks at this first. Total gross profit minus total gross loss. Simple.
The problem is that net profit without context is meaningless. $7,000 profit on a $10,000 account is a 70% return. $7,000 on a $100,000 account is 7%. Same number, wildly different story. Always calculate the return as a percentage of starting capital.
Profit factor
Total gross profit divided by total gross loss. This single number tells you more about a strategy's viability than almost anything else in the report.
| Profit factor | What it means |
|---|---|
| Below 1.0 | Strategy loses money. Walk away. |
| 1.0–1.3 | Barely profitable. Transaction costs might eat the edge. |
| 1.3–2.0 | Solid range. Most strategies that work long-term live here. |
| 2.0–3.0 | Strong. Worth investigating further. |
| Above 3.0 | Suspicious. Probably overfitting or too few trades. |
We tested one of our gold EAs (GoldPulse) over 5 years of data and got a profit factor of 1.72. Nothing spectacular on paper. But it's been running live for months with consistent results. A 1.72 that holds up in live trading is worth more than a 4.5 that falls apart after two weeks.
Maximum drawdown
The biggest peak-to-trough drop in account equity during the test. This is the most important risk number in the entire report.
MT4 shows both the dollar amount and the percentage. Focus on the percentage. If max drawdown is 30%, that means at some point during the backtest, the account shrank by 30% from its highest point. Can you stomach watching your real money do that?
Here's the thing most people get wrong about drawdown: whatever the backtest shows, live trading is almost always worse. Slippage, spread widening, and execution delays all add up. If the backtest shows 20% max drawdown, plan for 25–30% in reality.
A practical rule from our testing: divide net profit by max drawdown. This gives you the recovery factor. Above 2.0 is acceptable. Above 3.0 is good. Below 1.0 means the strategy earns less than its worst drawdown — not worth the stress.
Total trades
This determines whether any of the other numbers mean anything at all.
30 trades over 5 years? The results are statistically worthless. You might as well flip a coin 30 times and conclude it lands heads 70% of the time. You need at least 200 trades for the numbers to start being meaningful, and 500+ is better. We won't draw conclusions about any of our EAs with fewer than 300 trades in the sample.
The numbers that lie to you
Win rate (by itself)
A 90% win rate sounds incredible. It's often terrible.
The win rate only tells you how often trades close in profit. It says nothing about how much they win versus how much they lose. A strategy that wins 90% of trades but loses $200 on each loser while making $10 on each winner is a net loser.
| Win rate | Avg win | Avg loss | Win/loss ratio | Result per trade |
|---|---|---|---|---|
| 70% | $50 | $100 | 0.5 | +$5 (barely alive) |
| 50% | $120 | $80 | 1.5 | +$20 |
| 35% | $300 | $80 | 3.75 | +$53 |
| 80% | $30 | $200 | 0.15 | -$16 (losing money) |
That last row is what we call the "high win rate trap." It's everywhere in Martingale and grid EAs. The curve looks smooth right up until the moment it falls off a cliff.
Always look at win rate together with the average win/loss ratio. Neither number means anything alone.
Net profit (again)
A strategy that made $50,000 but dropped $40,000 along the way isn't a $50,000 strategy. It's a strategy where you needed an iron stomach and a lot of luck. Net profit without drawdown context is marketing material, not analysis.
Reading the equity curve
After the numbers, look at the chart. A healthy equity curve goes from lower-left to upper-right with tolerable wobbles along the way.
Steady upward slope — Drawdowns are small and the curve recovers quickly. This is what you want. It means the strategy works across different market conditions, not just one lucky stretch.
Strong first half, declining second half — The strategy worked in one market regime and stopped working when conditions changed. We see this constantly with EAs optimized on trending data that get destroyed in ranging markets. Check whether the backtest period includes both — if it doesn't, the results are suspect.
Rollercoaster — Big swings up and down, eventually ending in profit. The final number might look OK, but the psychological experience of living through those swings with real money is brutal. Most traders would have turned off the EA during the first big dip and missed the recovery.
Staircase — Long flat periods punctuated by sudden jumps. This usually means the EA trades very rarely or relies on extreme market events. The stats might be fine, but you'll spend months watching nothing happen and wondering if it's broken.
A crude but effective test: show the equity curve to someone who doesn't trade. Ask them "does this look stable?" If they hesitate, you probably should too.
Metrics most people skip
Expected payoff — Average profit per trade. Net profit divided by total trades. This has to be large enough to survive real-world costs. If expected payoff is 2–3 pips, spreads and slippage in live trading will likely eat all of it. We generally want to see at least 5–8 pips of expected payoff for any strategy we'd consider running live.
Maximum consecutive losses — How many losing trades happened back to back. This is a psychological metric more than a financial one. Can you watch 8 losses in a row without pulling the plug? We've seen traders manually disable an EA after 5 consecutive losses, only to miss the winning streak that started on trade 6. If you can't handle the maximum consecutive loss shown in the backtest, this EA isn't for you — regardless of the final profit number.
Sharpe ratio (MT5 only) — Risk-adjusted return. Below 0.5 means the strategy isn't earning enough for the risk it's taking. Between 0.5 and 1.0 is passable. Above 1.0 is good. Above 2.0, start asking questions — according to Investopedia's Sharpe ratio guide, readings that high sustained over long periods are rare for any asset class.
Modeling quality — This appears in MT4 reports as a percentage. Below 90% means the price data has gaps — ticks were reconstructed from candle data rather than being real recorded prices. For strategies on short timeframes (M1, M5) or with tight stop losses, low modeling quality can swing the results by 15–30%. We've run the same EA on 90% quality data and 99%+ real tick data and gotten completely different outcomes. MT5's "every tick based on real ticks" mode solves this — it downloads actual broker tick recordings.
Spotting overfitting in the report
Overfitting is the single biggest reason EAs look amazing in backtests and die in live trading. A few things to watch for:
Too many parameters. If the EA has 20+ adjustable settings, the developer can tune it to fit any historical data perfectly. A robust strategy usually has 3–5 core parameters. The more dials to turn, the easier it is to memorize history rather than learn patterns.
Hyper-specific values. An MA crossover of 5.37 and 21.89? That precision is a red flag. Robust parameters work across a range of nearby values. If changing a parameter from 14 to 15 flips the strategy from profitable to losing, it's memorizing noise.
Profit factor above 3.0 with low trade count. Incredible numbers on a small sample are almost always coincidence.
Perfect curve on in-sample, collapse on out-of-sample. If you can split the backtest period in half and the strategy only works on the period it was optimized for, it's overfit. Period. We write about this in depth in our overfitting guide.
Martingale reports: the beautiful lie
Martingale EAs produce the prettiest backtest curves you'll ever see. Nearly straight line up, 95%+ win rate, smooth as silk.
Don't be fooled.
Martingale works by doubling position size after a loss. Win once and you recover everything. The math works — until it doesn't. Seven consecutive losses means the next position is 128x the original size. The account can't handle it, and the curve that was going up for months drops to zero in a single afternoon.
When reading a Martingale report, ignore the win rate and the equity curve. Instead look at:
- Maximum position size — how heavy does it get at max layer count?
- Maximum floating loss — not realized drawdown, but unrealized loss while positions are open. This number is often far worse than the max drawdown statistic.
- Backtest timeframe — does it include extreme events? The 2015 Swiss franc shock, COVID crash, 2022 rate hike cycle? A Martingale that survived all three might be genuinely robust. One that only tested 2021–2023 has never been truly tested.
Martingale isn't inherently a scam — but using its smooth curve as proof of safety is dishonest. We have a full breakdown of the risks of Martingale and grid strategies.
Is this backtest even trustworthy?
Before analyzing the report, check whether the test itself was run properly.
Data quality. Tick-level data produces far more reliable results than 1-minute candle interpolation. The difference matters most for scalping strategies and EAs with tight stop losses. MT5's real tick mode is the gold standard. On MT4, tools like Tick Data Suite can push modeling quality above 99%.
Time span. Three years minimum. Five is better. The backtest needs to cover trending markets, ranging markets, and at least one major volatility event. A 6-month backtest that happened to coincide with a strong trend tells you nothing about how the EA handles consolidation.
Realistic costs. Was the spread set to something realistic? Most backtests default to a fixed spread that's lower than what you'll actually pay. We always test with at least 1.5x the broker's average spread and include commission if applicable. If the strategy stops working with realistic costs, the edge was too thin to begin with.
Out-of-sample validation. The single most important test: optimize the EA on one data range, then run it on a completely separate range it hasn't seen. If results hold up, the strategy likely has real edge. If they collapse, it's curve-fitted. We use this method on every EA we build, and it catches overfitting roughly 40% of the time.
What to do after reading the report
A good backtest report isn't the finish line. It's the starting line.
-
Run your own backtest. Don't trust the developer's screenshots. Download the EA, load it into your own MT4/MT5, and run the test yourself. Every EA in the FXTool marketplace comes with preset files so you can replicate our published results.
-
Demo first. Put the EA on a demo account for 1–3 months. Watch whether execution matches what the backtest predicted. Pay attention to slippage and spread differences. We wrote a full guide on the gap between backtests and live trading.
-
Start small. When you go live, use money you can genuinely afford to lose completely. $200, $500 — whatever won't affect your life if it goes to zero. Scale up only after 3–6 months of consistent results.
-
Manage your risk. No single trade should risk more than 1–2% of your account. Position sizing matters more than the EA itself. Our risk management guide and position size calculator can help with the math.
About the author: The FXTool team builds and tests MetaTrader trading tools daily. We run every EA we sell on live accounts and publish the results. This guide reflects what we've learned from building 50+ EAs and working with thousands of retail traders.
Forex trading involves significant risk and may result in total loss of capital. This article is for educational purposes only and is not investment advice. Understand the risks and consider your financial situation before trading.