If you have watched an automated system post steady gains for months and then quietly bleed equity for no obvious reason, you are not alone. This is one of the most common frustrations traders raise about algorithmic trading, and it deserves a real answer rather than a shrug. Below is a practitioner-level breakdown of why profitable expert advisors lose their edge, how to tell ordinary drawdown from actual strategy failure, and what a disciplined trader does differently.

Markets Are Not Static — And Neither Is Your Edge

Every rules-based trading strategy, no matter how well designed, is built around assumptions about how price tends to behave: how often it trends versus ranges, how wide typical swings are, how quickly volatility expands and contracts. Those assumptions hold reasonably well most of the time, which is exactly why a system can post a strong run of profitable months. The problem is that markets shift between regimes — trending, ranging, high-volatility, low-volatility — and an EA tuned for one regime will naturally underperform in another.

Gold is a particularly clear example of this. XAUUSD can spend months grinding in a tight range tied to real interest rate expectations, then abruptly become one of the most volatile instruments on the board around a geopolitical shock or a surprise policy move. A strategy that thrives on support and resistance reactions in a range-bound market can struggle badly the moment gold starts trending hard on structural demand shifts or a wave of geopolitical risk pricing. This is a structural feature of trading any single instrument with a fixed rule set, not a flaw unique to one EA. Traders who understand why gold prices move are better equipped to interpret a rough patch correctly instead of assuming the EA is "broken." Before concluding an EA has stopped working, ask whether the broader market regime has changed since the strategy's strongest run.

Overfitting: When Backtest Profits Were Never Real

The single most common reason a "profitable" EA eventually stops working is that it was never actually profitable in a forward-looking sense — it was curve-fitted to history. Curve-fitting happens when a developer adjusts entry filters, exit rules, and parameters repeatedly against the same historical data set until the equity curve looks flawless. The strategy has effectively memorized the noise in that specific data window rather than learning a repeatable, tradable pattern.

An overfit system typically shows textbook-perfect backtest statistics: minimal drawdown, an almost-straight equity curve, and a suspiciously high win rate. Once it goes live, and the market produces price action that differs even slightly from the historical sample it was fitted to, performance degrades — sometimes within weeks. This is why backtesting on MT5 should always include out-of-sample testing, not just a single optimized run. The MQL5 documentation on strategy testing covers the mechanics of proper backtesting and is worth reviewing before trusting any backtest at face value.

A useful gut check: if a strategy's backtest shows fewer than two or three losing months out of thirty, or a maximum drawdown under 5% across several years of gold price history, treat that as a red flag rather than a selling point. Real markets are messier than that, and a strategy that never shows discomfort in testing usually shows plenty of it live.

Worked Example: Overfitting in Numbers

Suppose a developer tests 40 parameter combinations against three years of XAUUSD H4 data and selects the single combination that produced the best equity curve. That combination might show a 68% win rate and a 1.9 average reward-to-risk ratio in the backtest. But because it was chosen specifically because it fit that data best, its true forward-looking edge might be closer to a 52% win rate with a 1.3 reward-to-risk ratio — still potentially profitable, but nowhere near what the backtest implied. Traders who size positions and set expectations based on the inflated backtest numbers are set up for disappointment the moment live results regress toward the strategy's real, smaller edge.

Broker and Execution Conditions Change Over Time

An EA's edge is not just about its entry and exit logic — it's about that logic combined with real-world execution costs. Spreads, commissions, and slippage all eat into the theoretical edge a backtest shows, and those costs are not fixed. Brokers periodically widen spreads during specific sessions, liquidity conditions shift, and the gap between a backtested fill price and an actual live fill can grow over time, especially around high-impact economic news events that move gold sharply in seconds.

This is one of the most underappreciated reasons a strategy that was profitable a year ago produces flat or negative results now, even with identical rules. If your broker's typical XAUUSD spread has crept from 20 cents to 35 cents, and your strategy averages a modest profit target per trade, that widening alone can erase a meaningful share of the edge. Comparing gold spreads across brokers periodically, rather than assuming your original setup is still competitive, is a simple maintenance habit many traders skip.

Worked Example: Spread Creep Eating an Edge

Assume a strategy averages 45 trades per month with an average net profit of $12 per 0.10 lot trade after typical costs at the time of testing, when the average spread was 18 cents. If the broker's average spread widens to 32 cents — a 14-cent increase — and each 0.10 lot XAUUSD trade has roughly $1 of value per cent of spread, that's an added cost of roughly $14 per trade purely from spread. On 45 trades a month, that is $630 in additional cost eating into a strategy that may have only generated $540 in gross monthly profit at the original cost structure. The strategy hasn't changed at all — the execution environment around it has.

Volatility Regime Shifts Hit Gold Strategies Hard

Gold's volatility is not constant, and a strategy calibrated for one volatility regime can behave very differently in another. During periods of elevated futures market volatility, average daily ranges on XAUUSD can expand by 50% or more compared to quieter months. A strategy using fixed stop distances or fixed profit targets designed around a calmer period may get stopped out prematurely in a high-volatility stretch, or leave significant profit on the table if targets are too tight relative to the new range.

This is precisely why systems built around adaptive, risk-based position sizing tend to hold up better across changing conditions than those with rigid, fixed-lot rules. Understanding how drawdown behaves under different volatility conditions, and reviewing how EA settings interact with volatility, helps traders separate a temporary rough patch caused by a volatility spike from a genuine breakdown in the underlying logic.

Trader-Side Changes Are an Overlooked Culprit

Not every "the EA stopped working" story is actually about the EA. A significant share of cases trace back to changes the trader made, often without realizing the downstream effect: switching from a conservative risk setting to an aggressive one after a winning streak, running the same EA on multiple accounts with overlapping trades, misconfiguring a magic number during an update so two versions of a system trade the same account, or changing a VPS or broker without re-verifying that execution conditions still match what the strategy was designed for.

Reviewing how EA magic numbers work is worth doing any time you run more than one automated strategy, since a misconfiguration there can cause one system to interfere with another's trade management. Likewise, if you've recently changed your VPS provider, latency and connection stability differences can alter fill quality enough to visibly affect results, even though the strategy logic never changed.

Worked Example: Risk Setting Escalation

Consider a trader running an EA on a Conservative risk setting for six months with a $5,000 account, averaging 2% monthly growth with a maximum observed drawdown of 6%. After a strong quarter, the trader switches to Aggressive, roughly tripling position size per signal. The strategy's win rate and trade selection do not change at all — but a losing streak that would have cost 6% under Conservative sizing now costs closer to 18%, which feels like the system "broke" even though the only variable that changed was the trader's own risk dial.

Normal Drawdown vs. Genuine Strategy Breakdown

One of the most valuable skills an EA trader can develop is telling the difference between a normal, expected losing stretch and an actual breakdown in the strategy's edge. Every legitimate strategy has losing trades and losing months built into its statistical profile — that is simply the nature of probability-based trading, and sound risk management exists precisely because no system wins every time. The table below outlines the practical differences.

That last row matters more than most traders realize. If a strategy is backed by a verified live track record on Myfxbook, you can directly compare your own account's performance over the same window to the published results. A wide gap between your results and the verified account over the identical period points strongly toward a local, trader-side or broker-side issue rather than a flaw in the strategy itself. Myfxbook's own guidance on account verification explains what a genuinely audited track record looks like, which is a useful reference point when evaluating any EA's claims, including your own experience running one.

Ranking the Real Causes of EA Decay

Not all causes of "an EA stopped working" are equally common. Based on how these issues typically surface among gold EA users, here is a practical ranking with the usual fix for each.

Scams and Fabricated Track Records Masquerading as "Decay"

It's worth being direct about a darker version of this question: sometimes an EA never actually worked, and its "profitable" reputation was built on a fabricated backtest, a cherry-picked screenshot, or an account history that was never independently verified. When that kind of product eventually stops posting fake gains and starts losing real money, it isn't decay — it was never a real edge to begin with.

The CFTC's advisory on trading system fraud and its broader forex fraud warnings outline the classic pattern: guaranteed-return claims, pressure to buy quickly, and an absence of independently verifiable results. The FTC's guidance on investment scams adds that any product promising consistent gains with no real risk is a warning sign, not a selling point — legitimate automated trading always carries the possibility of loss. Any EA or signal service that cannot show a public, independently verified track record — for example through a Myfxbook-verified account or an MQL5 verified signal — should be treated with skepticism regardless of how convincing the marketing looks.

This is also why it's worth reading reviews and comparisons critically. Resources profiling proven trading systems or ranking the best performing MT4 EAs are only useful if they emphasize verified, audited results over marketing claims.

Technical and Platform-Level Failures

Sometimes the EA hasn't lost its edge at all — it has simply stopped executing correctly. Common culprits include a VPS outage that leaves the terminal disconnected during a key session, a platform update that changes default behavior, insufficient margin blocking new trades, or "Algo Trading" being toggled off after a terminal restart. These produce results that look like poor performance but are actually operational, not strategic.

Checking the MetaTrader 4 platform documentation or the equivalent MetaTrader 5 automated trading guide after any update catches a surprising share of "why isn't my EA trading" cases. A broader troubleshooting reference like a guide to common EA problems and fixes is worth bookmarking, since many issues that look like strategy failure trace back to a terminal setting or a missed permission.

What To Do When Performance Genuinely Slips

Once you've ruled out execution and technical issues, and confirmed the drop isn't simply normal variance within the strategy's historical range, there is a sensible sequence to follow rather than reacting emotionally.

Start by comparing your account's recent results against the strategy's own published, verified track record over the identical date range — a real divergence there tells you the issue is local to your account, broker, or settings, not the strategy itself. Next, review whether you've changed any settings, including risk mode, lot sizing, or running additional EAs on the same account. Then assess whether the broader gold market has shifted regime — a sudden volatility spike tied to central bank policy, for instance, explains a lot. Finally, consider your overall capital allocation: capital preservation principles suggest that no single automated strategy, however strong its track record, should represent all of a trader's risk capital, and diversifying across multiple EAs or strategies can reduce the impact when any one system goes through a rough stretch.

It's also worth revisiting your original expectations. If you started with modest starting capital, recalibrating what realistic monthly and annual returns look like can reveal whether what you're seeing is actually underperformance or simply the normal variance of a smaller sample size. The checklist below is a practical monthly review routine.

Building Resilience Instead of Chasing the Next EA

The traders who handle this best don't treat every drawdown as a signal to abandon ship and chase the next "hot" EA. Jumping from system to system after every rough patch almost guarantees underperformance, because you exit right when a system is statistically due for reversion and enter a new, unproven one at what is often its own temporary peak. Resilient traders instead run only strategies with a long, independently verified history, size positions conservatively relative to account equity, and periodically re-verify that broker conditions and platform settings still match what the strategy was designed around.

This is also where selectivity matters. A system that only takes roughly one high-conviction setup per day on a single instrument, using trend confirmation techniques rather than firing on every minor fluctuation, tends to be less exposed to the kind of noise-driven whipsaws that erode over-traded, high-frequency systems during choppy periods. Fewer, higher-quality trades don't eliminate drawdown, but they reduce the number of opportunities for execution costs and short-term noise to compound against you.

Risk Disclosure

Trading gold and other financial instruments carries real risk, including the possibility of losing some or all of your capital. Past performance, including any verified historical track record, does not guarantee future results. No automated trading system, regardless of its track record or design, can eliminate the risk of loss. Only trade with capital you can genuinely afford to lose, and treat any strategy — automated or manual — as one part of a broader, diversified approach to risk.

Frequently Asked Questions

Why did my EA stop making money after months of consistent profit?

The most common reasons are a shift in market conditions (volatility or trend character changing), execution costs like spread widening over time, or a change you made to risk settings or lot sizing. Less commonly, the original backtest was curve-fitted and never reflected a real forward-looking edge. Comparing your results against the strategy's verified live track record over the same period is the fastest way to identify which of these applies.

Is drawdown a sign that an EA is broken?

Not necessarily. Every legitimate rules-based strategy has losing trades and losing stretches built into its statistical profile. A drawdown that stays within the historical range shown in a strategy's verified track record is normal variance, not a sign of failure. A drawdown that far exceeds anything seen historically, with no external explanation, is more concerning.

How long should I wait before deciding an EA has genuinely stopped working?

There's no universal number, but comparing the current losing stretch to the longest historical drawdown in the strategy's verified track record is a reasonable benchmark. If you're within that historical range and the trade logic still matches the documented rules, it's premature to conclude the edge is gone.

Can changing my risk setting cause an EA to appear to stop working?

Yes, and this is one of the most overlooked causes. Increasing lot size or switching from a Conservative to an Aggressive risk mode amplifies both gains and losses proportionally. A losing streak that was mild under conservative sizing can look like a system breakdown under aggressive sizing, even though the underlying strategy logic never changed.

Does gold's volatility make EAs stop working more often than other instruments?

Gold can shift between calm, range-bound periods and sharp volatility expansions tied to interest rate expectations or geopolitical events. A strategy calibrated around one volatility regime may need to adjust to a materially different one, which can look like underperformance even though the underlying logic hasn't changed.

How can I tell if an EA's original backtest was overfitted?

Warning signs include backtest results with an almost-straight equity curve, very few losing months, and drawdown well under what similar strategies typically show over multi-year periods. A strategy validated only against the exact data it was optimized on, without out-of-sample or forward testing, is at high risk of being curve-fitted.

What role does my broker play in an EA "stopping working"?

A significant one. Spreads, slippage, and execution quality directly affect a strategy's realized profitability, and these conditions can drift over time even without any change to the strategy itself. Comparing your broker's current spreads against what was typical when the strategy was originally tested can reveal whether execution drift is eating into your results.

Should I switch EAs every time I see a losing month?

Generally no. Switching strategies after every losing stretch tends to produce worse long-term results, because you often exit a system right as it's statistically due for a recovery and enter an unproven one. A more disciplined approach is to compare your results against the strategy's verified track record and only reconsider if the divergence is significant and sustained.

How do I know if an EA's track record was ever real to begin with?

Look for independent verification — a publicly auditable Myfxbook account or an MQL5 signal history you can review directly — rather than screenshots or marketing claims alone. The CFTC and FTC both warn that guaranteed-return promises and an absence of independently verifiable results are classic signs of a fraudulent trading system rather than a genuine, if imperfect, strategy.

What's the single best habit for catching EA problems early?

Regularly compare your live account's performance against the strategy's own verified, published track record over the same time window. If the two are broadly aligned, any drawdown you're seeing is likely shared market-wide variance. If your results diverge sharply from the verified record, the issue is more likely local — settings, broker, or technical setup — and worth investigating directly.