The Backtest Fraud Crisis: What the Numbers Actually Say

87% of retail Expert Advisors blow accounts within 90 days. Yet on MQL5, thousands claim 90%+ win rates in their backtests. The disconnect is not a mystery — it's evidence of systematic fraud.

Most EA creators are not trying to scam you. They just have no idea how to properly verify their own code. The result: garbage strategies that look pristine in backtests but collapse the moment real money touches them.

This article shows you exactly how the manipulation happens, the eight tactics backtest fraudsters use, and how professionals verify before they touch a single live trade.

How Backtest Fraud Actually Works: The 8 Manipulation Tactics

Here are the exact methods traders and EA creators use to artificially inflate backtest performance. Some are accidental. Most are deliberate.

1. Curve-Fitting (Overfitting Parameters)

The creator runs the strategy against historical data, tweaking parameters (stop loss, take profit, period values) until the backtest sings. They find the exact settings that work on 2020-2023 data. Then they go live with those settings and get slaughtered.

Why it fails: The market changes. The exact parameters that made 90% win trades in 2022 are worthless in 2026. You're buying a solution to a problem that no longer exists.

Red flag: Any EA that claims to have "optimal parameters" without explicit statements about robustness testing across multiple market regimes is curve-fitted. Period.

2. Look-Ahead Bias

The EA uses data it shouldn't have access to when making trades. Example: the creator codes a rule that triggers on "the lowest price of today" — but the backtest has access to the entire day's data when the trade is executed. In live trading, you don't know what the low will be until the day ends.

This is often accidental. MQL5 testing is done on a single timeframe (like 1H), but the EA references data from multiple timeframes without proper synchronization. The backtest uses future data; live trading doesn't.

Red flag: If the backtest results are too consistent (almost every trade wins), look-ahead bias is likely. Real trading is ragged and uncertain. Smooth curves are suspicious.

3. Survivorship Bias

The backtest runs only on currency pairs or timeframes where the strategy works. If EURUSD, GBPUSD, and USDJPY are tested, but AUDUSD (where the strategy loses) is not reported, the win rate is artificially high.

This is pervasive. Creators backtest across 28 pairs, find the 7 where the strategy wins, and only report those. They claim "tested on major pairs" without mentioning the ones that got destroyed.

Red flag: Backtest reports that show results on only 3-5 pairs, or show "average returns" across pairs with zero variation (all winners), are hiding losers.

4. Hidden Losses / Partial Reporting

The backtest includes winning periods (e.g., Jan-Feb 2022, Jun-Aug 2023) but omits losing periods (Mar-May 2022, when the strategy was underwater). The report says "tested on 5 years of data" but really shows cherry-picked windows.

Legitimate reports show continuous backtests across the entire period, with drawdowns and losing streaks visible. Fraudulent ones show fragmented windows or "best performance periods."

Red flag: Any report that doesn't show continuous equity curve from start to end. Gaps mean hidden losses.

5. Unrealistic Slippage / Spread Assumptions

The backtest assumes 0 slippage and 0 spread — meaning every trade fills at the exact price quoted, with no dealer markup. In reality, brokers add 1-2 pips (or more) on every trade. Over 100 trades, this wipes out the margin.

Professional backtests use realistic slippage (1-5 pips per trade, depending on the pair and broker) and include spread costs. A 90% win rate with 0 slippage becomes a 67% win rate with realistic costs.

Red flag: If the backtest report says "Slippage: 0 pips" or doesn't mention slippage at all, the results are fake.

6. Commission & Fee Omission

The backtest doesn't account for broker commissions, account management fees, or VPS costs. A $100 EA that costs $15/month to run (commission + VPS) has a different ROI than a $100 EA that costs $80/month.

This especially torpedoes scalping and high-frequency strategies. A scalper that makes 50 pips on 100 trades (5,000 pips total) sounds great until you realize 3,000 of those pips went to the broker in commissions.

Red flag: Backtest reports that don't itemize costs (commission, swap, VPS) or show the profit after all fees are stripped out.

7. Backtesting on the Wrong Data / Old Models

The creator backtests using the MetaTrader 4 built-in tester with default price data (often low-resolution, with gaps, or using ask/bid data inconsistently). The results are not comparable to real live fills because the price bars themselves are inaccurate.

Professional backtesting uses tick-level data from a data provider (like TickData or FXCM), imported into a proper backtester. This gives you real fills at real prices with realistic slippage. The results are worse than the built-in tester — but they're real.

Red flag: Backtest results from the default MT4 tester without mention of data source. Also, reports that claim "99.9% modeling quality" (MT4 max) instead of 99.99%+ (tick-level).

8. Time-Window Bias

The backtest covers years when the strategy happened to work (bull market, low volatility period) but ignores crash periods (2020 March crash, 2022 energy crisis, 2026 AI bubble pop). A strategy that worked from 2019-2023 might fail catastrophically in the next correction.

Real validation requires testing across multiple market regimes: trending markets, choppy markets, high-volatility shock events, and slow-grind declines. If a backtest shows profits across all four, it's more trustworthy.

Red flag: Backtest windows that coincide with one of the longest bull runs in history (2009-2021) without stress-testing in 2008, 2000, or 1998-style conditions.

How Professionals Verify Backtest Results: The 5-Step Proof Checklist

You don't need to be a programmer to audit an EA's backtest. Use this checklist when someone claims their EA has a 90%+ win rate.

Step 1: Ask for Tick-Level Data

Demand that the backtest was run on tick data, not bar data. This is the non-negotiable baseline. If they give you an MT4 backtest with "99.9% modeling quality," they're cutting corners.

Professional EA developers use external tick data providers and run backtests in dedicated backtesting software. They can prove it. If they can't, the results are questionable.

Step 2: Verify the Test Period and Stress-Test

Look at the exact backtest window. 2020-2025? That's most of a bull run. Ask: "Did you test on 2008? 2001? 2015? Have you stress-tested through a 20%+ drawdown?"

A legitimate backtest shows continuous results across multiple market regimes. If the creator only shows 2022-2025 results, they're hiding losses from 2020-2022 when their strategy got crushed.

Request a stress test report that shows the EA's behavior during historical crashes (March 2020, March 2015, Sep 2008). If they can't provide it, the backtest is incomplete.

Step 3: Demand Real Slippage and Spread Costs

The backtest must explicitly state slippage (minimum 1 pip per trade, realistic to your broker) and spread costs (2-3 pips for majors, 5-10 for minors/exotics). The profit report should show two numbers: "gross profit" and "net profit after slippage & spreads."

If the report only shows gross profit, the net profit is probably 50% or less. The creators are hiding this because it kills the narrative.

Step 4: Check the Equity Curve for Red Flags

A real equity curve is ragged. It goes up, it goes down, it has drawdowns, it has losing streaks. If the equity curve looks like a smooth exponential function (straight up), it's fake.

Look for:

Step 5: Test It Live on a Demo Account

This is the ultimate proof. Run the EA on a demo account for 30 days with real market conditions and real fill slippage. Compare the results to the backtest.

If the backtest shows 90% win rate and the demo shows 65% win rate, the backtest was overfitted. If the demo shows similar results, the backtest might be legitimate (though 30 days is a small sample).

Alorny includes a full backtest report with every EA we build, and we provide a 30-day demo period so you can verify our claims before you pay. If we can't back up our backtest with live results, we don't take your money.

The Red Flags: What Every Backtest Fraud Has in Common

You don't need to understand the math. Just look for these patterns, and you'll catch 90% of fraudsters:

Case Study: How Real Backtests Look (vs. Fraudulent Ones)

Let's compare two backtests side by side. One is legitimate; one is fake.

FRAUDULENT Backtest Report

"Tested on EURUSD 2023-2025. 92% win rate. Total profit: $45,000 on $10,000 account. Average trade: +450 pips. Slippage: 0. Modeling quality: 99.9%."

Why it's fake:

LEGITIMATE Backtest Report

"Tested on EURUSD, GBPUSD, USDJPY, AUDUSD on 1H timeframe, 2015-2026 (11 years). Data: Dukascopy tick-level. Modeling quality: 99.99%. Win rate: 58%. Profit factor: 1.87. Max drawdown: 18%. Gross profit: $23,400. Net profit after realistic slippage (1.5 pips per trade): $18,600. Tested during 2015 Swiss franc crisis, 2020 March crash, 2022 rate hikes. 30-day live demo results: 59% win rate, $1,200 profit on $10,000 account."

Why it's legitimate:

The second one is less exciting. But it's the only one you should trade.

The Alorny Standard: How Real EA Developers Handle Backtesting

When we build an EA at Alorny, we don't curve-fit to impress. We build for the real world.

Here's our process:

  1. Multi-pair, multi-timeframe testing. We test across at least 3 pairs and 2+ timeframes to ensure robustness. If the strategy breaks on any pair, we know about it upfront.
  2. Tick-level data, 5+ years minimum. We import real tick data from professional sources and backtest across different market regimes. Boring data is honest data.
  3. Realistic slippage built in from day one. We test with 1.5-3 pips slippage per trade, depending on your broker. The profit report shows gross and net separately so you know exactly what you're getting.
  4. Forward-testing on demo for 30 days. Before you pay a dollar, we run the EA on a live demo account. You watch the results. If the demo doesn't match the backtest, we fix it.
  5. Full disclosure of parameters and methodology. You get the backtest report, the parameter explanation, the data source, everything. We don't hide anything because we don't have to.

We price EA development from $100, and every project includes a full backtest report and 30-day demo proof. If we couldn't back it up, we wouldn't offer it.

That's the bar. If the EA creator you're considering can't match it, they're cutting corners.

Why Do Fraudulent Backtests Exist? (The Uncomfortable Truth)

Most EA creators aren't intentionally defrauding anyone. They're just incompetent. They don't know about look-ahead bias or curve-fitting. They run a backtest in the MT4 default tester, see 90% wins, and think they've built a goldmine.

It's not until they test on a demo account (or worse, a live account) that the illusion shatters. By then, their pride is on the line, so they don't admit the mistake. Instead, they blame "market changes" or "your broker's slippage" rather than admitting their backtest was garbage.

The small percentage of intentional fraudsters are worse. They know exactly how to use look-ahead bias and survivorship bias to pump fake numbers. They sell the EA, take the money, and disappear before the drawdown hits.

Either way, you're the one holding the bag.

Here's the Key Takeaway About Backtest Fraud

A beautiful backtest is a warning sign, not a selling point. Real trading is ugly. Real EAs have drawdowns. Real win rates are 55-65%, not 90%.

If someone is claiming 90%+ win rates and smooth equity curves, they're either lying or delusional. Either way, their EA won't work.

When you evaluate an EA, ask for proof on all five points in the checklist above. If they can't provide it, walk. The best EAs are built by developers who have nothing to hide.

Common Questions About Backtest Verification

Key Takeaways