Your Grid EA Keeps Winning. Why Is the Account Still Shrinking?

Suppose an Expert Advisor has just finished a busy week. Its trading history is full of green numbers. The win rate looks glorious.

So why does the account's equity look as if it has just been told the restaurant bill is not being split?

This is a hypothetical scene, but the accounting problem is real: a grid EA can close plenty of winning trades while carrying losses large enough to overwhelm them. The trick is to look beyond the percentage of green trades and examine what the system is still holding.

Your trading history and your account are telling two stories

In MetaTrader 5, balance and equity answer different questions.

Balance reflects completed transactions. Equity reflects the account's current value, including floating gains and losses from open positions and other applicable adjustments.

Imagine a fictional account that began with $5,000.

• The EA has banked $75 in closed trading profit.
• Its still-open grid positions are showing $600 in floating loss.
• Ignoring any other adjustments for this simple illustration, the balance is $5,075, but equity is $4,475.

The $75 profit is not fake. It is simply not the whole story.

A screenshot of trading history may show successful closes. It does not necessarily show the price of all the positions still waiting for a reversal.

That is why checking an EA only by the number of winning tickets is a bit like judging a restaurant by how many customers paid, while ignoring the unpaid electricity bill.

Let's try a 95% win rate. Surely that is impressive?

Here is another entirely hypothetical example, separate from any real EA's performance.

A system closes 20 trades:

Nineteen winning trades out of twenty give us a 95% win rate. But the total is a $51 loss, even before commissions or other costs.

The expected average outcome under those assumed win and loss sizes is:

(0.95 × $6) − (0.05 × $165) = −$2.55 per trade.

This is expectancy, not a promise that any particular sequence of twenty trades will lose exactly $51. Actual outcomes and probabilities are unknown until you have credible data, and even a historical estimate may not hold in future conditions.

The important part is the question a win-rate screenshot cannot answer: How large were the winners, how large were the losers, and how long did the account carry risk while waiting?

The quiet danger of an unfinished grid

A grid places or manages orders around a set of price levels or intervals. Different EAs implement that concept differently. Some take small individual profits. Others manage several positions as a combined basket. Some use fixed sizes; others scale volume.

Consider a fictional buy-side grid on EURUSD. It opens one position, price declines, and another level is reached. Then another. And another.

If five same-direction positions are now open, the EA has more combined directional exposure than it had with one. If price keeps falling, each additional adverse move affects those open positions.

A sideways market might give the system opportunities to exit. A persistent downtrend might not.

The market has never signed an agreement to bounce just because your grid spacing says it would be convenient.

This creates two separate risks:

1. Floating loss can grow while closed trading history still looks cheerful.
2. Margin usage can increase as positions accumulate, potentially reducing free margin just as equity is already under pressure.

The exact margin effect depends on the symbol, account mode, broker rules, exposure and the positions involved. You cannot infer a safe account size from an EA's first lot alone.

A grid is not automatically a martingale

These terms tend to get introduced together, like two colleagues everyone assumes are related.

They are not identical.

A grid describes orders or positions arranged around price levels. Martingale-style sizing refers to increasing risk after losses according to a progression.

A grid can use fixed lots. A grid can use increasing lots. A non-grid strategy can also increase lot sizes after losses.

Here's a small hypothetical volume comparison:

• Five entries at 0.01 lot each result in 0.05 lot total open volume.
• Five entries progressing through 0.01, 0.02, 0.04, 0.08 and 0.16 result in 0.31 lot total open volume.

That second sequence has more than six times the combined volume. It does not mean it will lose exactly six times as much under every price path, because entries, direction and contract specifications matter. But it tells you why position-size progression deserves a prominent place in your risk review.

What the Strategy Tester should make you uncomfortable about

A smooth balance curve is pleasant to look at. The more useful question is what the system had to survive to produce it.

When assessing a grid EA in MT5's Strategy Tester, investigate:

Equity drawdown. Did the account experience large floating declines that are not obvious from completed trades?

Maximum simultaneous exposure. How many positions were open together, and what was their combined volume? Were all the positions on the same side of the market?

Margin pressure. How low did available free margin or margin level become under the test assumptions? Does the strategy behave sensibly when it cannot open the next order?

Unfinished positions. Were losing positions still open near the end of the test? Were the final reported statistics affected by how the test handled those positions?

Different market conditions. What happens in a sustained trend instead of a comfortable range? Try periods with adverse movement, changing spreads and less convenient execution assumptions.

Sizing sensitivity. What happens when account size, lot progression or grid spacing changes? Narrow spacing can accumulate entries faster; wider spacing is not automatically safe.

Repeatability. Did you tune the EA repeatedly on the same historical sample until the results looked handsome? Consider separate test periods and forward demo observation instead of treating a fitted backtest as proof.

Be careful comparing tests from different brokers or account types. Symbol specifications, contract size, leverage, margin rules and modeling assumptions can vary. Gold is a particularly good reminder that 0.01 lot is a volume label, not a universal dollar-risk measurement.

None of these checks reveals a guaranteed future maximum drawdown. They make the risks visible enough to evaluate rather than merely admire.

Give the robot rules for when the market refuses to cooperate

A useful risk policy is not just a take-profit target.

Before allowing a grid system to operate, define what it should do when the market is stubborn. Depending on the strategy, that may include:

• A maximum number of grid levels or simultaneously open positions.
• A cap on combined exposure, not merely initial lot size.
• A rule that stops new entries when margin or equity conditions deteriorate.
• A predefined exit or exposure-reduction condition when risk becomes unacceptable.
• A response to rejected orders, connection problems and unexpected platform restarts.
• A plan for monitoring the EA and testing any changes on a demo account first.

These are examples of controls, not a suggested universal configuration. Stops can slip in fast or illiquid conditions, and no rule eliminates trading risk.

Here is the uncomfortable but useful reality: a system that accepts more realized losses may show a lower win rate and still have a healthier risk profile. Whether it is actually better depends on the full distribution of outcomes, costs and the amount of capital at risk.

The question worth asking at the end

The next time somebody shows you an EA with an astonishing win rate, enjoy the green numbers for a second.

Then ask:

What did the account's equity look like during the worst open basket, and what would have happened if the price kept moving against it?

That's a more demanding question than "How often does it win?"

It's also a far more useful one.

Educational discussion only. All numerical examples above are hypothetical, not observed trading results or performance claims. Leveraged trading can result in substantial losses, including loss of capital.