The PF 7 / 92% Win Rate Trap — 5 Checkpoints to Spot Inflated Backtest EAs

"Profit Factor 7.07, win rate 92.24%, max drawdown 1.85%, 22-year backtest."

Sound familiar? The moment you see numbers like these in an EA ad and think "this is incredible — I have to buy it," stop. This combination of figures is mathematically near-impossible for a conventional single-entry EA.

That doesn't mean the numbers are outright lies. The trick is in how the numbers are constructed. This article explains what's really happening behind flashy backtests — and how to see through them before you buy.

Why "PF 7, 92% Win Rate" Is Abnormal

The Math Behind It

A realistic trend-following EA (e.g., EMA crossover with ATR-based SL/TP) will typically land in this range:

To achieve "92% win rate / PF 7," the EA must rely on at least one of the following:

1. Ultra-tight take-profit with unlimited holding — never cuts losses, books tiny profits on any bounce
2. Multiple simultaneous positions that offset each other — one leg loses, others win, net profit is taken
3. Extremely distant SL or no SL at all — almost every trade closes as a "winner"
4. Martingale, grid, or averaging variants — adds positions in the same direction and exits when average cost is recovered

These designs look beautiful in backtests but blow up a live account on a single sharp move.

Real Example: "Ikoku no Nagrikomi GOLD_V5" Numbers

Here are the published figures for a commercially available EA (quoted from the vendor's page):

The official FAQ explicitly states "No martingale or averaging logic is used." However, the same FAQ includes the fine print: "Because multiple positions may be held simultaneously, the behavior can appear similar to averaging."

Furthermore, among the "12 trading modes" offered are options such as:

• Fixed base lot
• Compound lot (auto-calculated)
• Loss multiplier adjustment ← increase lot size after a loss
• Win multiplier adjustment ← increase lot size after a win
• Position count multiplier adjustment

In other words, "no martingale" simply means it's OFF by default — switching modes enables lot increases based on the number of losses. That is, functionally, a martingale variant.

Combined with "multiple simultaneous positions" and "movement that looks like averaging," this design is exactly what produces a 92% win rate.

None of this is illegal or technically false. It is artful framing.

How Backtest Numbers Are Engineered — 4 Patterns

Here are the most common techniques used to produce spectacular figures. (We could technically produce PF 7+ ourselves using the same methods — that's precisely why we don't.)

Pattern 1: Grid with Unreported Floating Losses

- Place 5 limit orders at a price range (e.g., buy limits every 10 pips)
- Price falls → limits fill → floating loss positions accumulate
- Price bounces → upper positions close with small profit
- The deepest position is held open "until it eventually recovers"
- At backtest end, open positions are treated as "unsettled" — losses not counted

This design produces backtest results of PF 5–15, win rate 85–95%, DD 1–5% — it looks unstoppable. But in live trading, a single large move destroys margin on all those floating positions simultaneously.

Pattern 2: Extremely Distant Stop-Loss

SL = ATR × 10   (typical: ATR × 1.5–2.0)
TP = ATR × 0.3  (typical: ATR × 2.0)

This means "almost every trade closes with a tiny profit." Once or twice a year, a large crash will hit the SL, wiping out all prior gains in a single trade. If the historical 22-year data happened to concentrate those crashes in the "past" portion of the test window, the curve looks pristine.

Pattern 3: Martingale / Averaging

- Initial position goes into the red
- Add 1.5× lot in the same direction (lower average entry)
- Position goes deeper into the red → add 1.5× again
- Eventually a bounce takes the average above break-even → close all at profit

If the 22-year historical data never produced a trend that didn't eventually reverse, this EA approaches 100% win rate. But the mathematical certainty that one catastrophic move will wipe the account never goes away.

Pattern 4: Curve-Fitted Optimization

Running thousands of parameter combinations across 22 years of data and selecting "the best-performing combination for those exact 22 years" will always produce PF 5+. It just won't work in the future. (See our separate article: How to Avoid Over-Optimization (Curve Fitting).)

5 Checkpoints Before You Buy

If you're considering an EA with "PF 7," verify these five things first.

Checkpoint 1: Is the Logic Disclosed?

Does the sales page describe a concrete algorithm, like "EMA crossover + ATR-based SL"?

• Clearly stated: EMA 37/80 cross, Bollinger Band breakout, etc. ✅
• "Dozens of indicators used adaptively," "proprietary secret logic," "AI-driven decisions" ⚠ — this is a black box. When something goes wrong, you can't diagnose why.

Never buy a black-box EA. That's the iron rule.

Checkpoint 2: Are Both the Balance Curve and the Equity Curve Shown?

A backtest report can display two types of graphs:

EAs with large floating losses will show a smooth upward Balance curve but periodic deep Equity dips. If only the Balance curve is shown, the vendor may be hiding floating losses.

Ask the vendor for the Equity curve. If they can't provide it, walk away.

Checkpoint 3: Maximum Lot Usage and Open Positions at Peak Losing Streak

The simplest questions to expose a martingale/grid EA:

• "What is the total lot size at the worst losing streak?"
• "What is the maximum number of consecutive averaging entries?"
• "Is there a MaxNanpinCount (max averaging count) parameter?"

If the answer is vague, or uses phrases like "multiple positions may be held" or "depends on market conditions," it's a grid or averaging system.

Checkpoint 4: Is Walk-Forward / Out-of-Sample Validation Published?

Legitimate developers publish separate metrics for Train (first 10 years) and Test (last 10 years):

An EA that collapses in the Test period is optimized only on historical data. Live trading is the future, so Test performance is what actually matters.

Checkpoint 5: Lot Size Calculation Logic

Request a screenshot of the input parameter screen. Be cautious of any EA with parameters named:

• LotMultiplier
• MartingaleStep
• NanpinDistance
• GridStep
• RecoveryMode
• MaxPositions set to 10 or more (assumes multiple positions)
• UseAveraging

If these exist yet the vendor claims "no martingale / no averaging," there is a high probability of deliberate wordplay to avoid liability.

If You Still Want to Use a "Flashy Number" EA

Honestly, martingale and grid EAs can produce real profits when used correctly. FXEA also offers averaging EAs such as "KAMIKAZE," "TETSUBEKI," "HEDGE_RECOVERY," and "ATR_DYNAMIC."

However, you must follow these rules without exception:

1. Only invest money you can afford to lose entirely — no more than 5% of total capital
2. Withdraw profits daily — letting profits accumulate in the account means they all disappear when the crash comes
3. Accept bankruptcy as a built-in outcome — understand mathematically that this type of strategy will eventually fail
4. Choose EAs that explicitly state "bankruptcy-risk / daily withdrawal required" — vendors who are upfront about this are more trustworthy than those who hide it

If you take the flashy numbers at face value and deposit serious capital, you are heading for the classic martingale/grid outcome: months of profit erased in a single volatile session.

See Martingale EA Operating Rules for full guidance.

Summary — Honest Numbers, Honest EAs

At FXEA, we fully disclose our strategy logic, publish MQL5 source code, and share backtest results produced from real MT5 data.

As a result, our EAs produce numbers like these — unspectacular, but real:

Compared to "PF 7," these numbers look modest. But they represent the realistic tradeoff for near-zero risk of a single move wiping out the account.

The choice is simple:

• Want to grow capital fast → Use martingale/grid EAs with a "bankruptcy-accepted" mindset (daily withdrawal is non-negotiable)
• Want to trade long-term → Accept single-entry EAs at realistic annual returns of 5–15%

No EA can deliver both simultaneously. Any vendor claiming otherwise is hiding something.

The perfect EA doesn't exist. But an EA with disclosed logic, verifiable results, and a strategy that fits your trading style does exist. That's what FXEA is built to provide.

Related Articles

• How to Avoid Over-Optimization (Curve Fitting)
• How to Read a Backtest Report Correctly
• Preventing Over-Optimization with Walk-Forward Analysis
• How to Spot EA Scams
• Martingale EA Operating Rules

Related Pages

• How to Identify Fraudulent and Scam EAs
• How to Read Backtest Metrics Correctly
• Verifying Over-Optimization with Walk-Forward Testing
• Our Free EAs (All Backtest Data Published)