Gold EA Failure: Why Our Donchian Strategy Hit Intraday Limits

A beginner-friendly summary of the verification: “Gold EA Failure: Why Our Donchian Strategy Hit Intraday Limits”.

Here’s an interesting case study from our recent research, where a significant daily loss wasn’t what it seemed on the surface.

What’s the idea?

We were testing a gold trading Expert Advisor (EA) based on the Donchian Channel strategy, operating on the M1 (1-minute) timeframe. For those new to EAs, it’s basically an automated system that executes trades for you. The Donchian Channel is a classic trend-following indicator, designed to help us catch and ride big market moves. The core idea is to let profits run as long as the trend continues.

How I tested it

Our test focused on intraday (within a single day) performance, specifically looking for vulnerabilities. We ran simulations with different risk settings: a 1% risk per trade/day and a slightly more conservative 0.7% risk. A crucial rule in our system is a daily loss limit – if the account’s effective equity (which includes both realized and unrealized profits from open positions) drops by more than 5% in a single day, the system stops trading for that day.

What happened?

On January 30, 2026, our EA detected a major intraday loss. At 1% risk, it hit -7.2% for the day, and at 0.7% risk, it was -5.3%. Both figures breached our daily -5% loss limit, marking the day as a “disqualification.” Ouch! But here’s where it gets interesting. When we dug into the root cause (the “真因” or true cause, as we call it), we found something unexpected. This wasn’t a flaw in the EA’s trading logic or a series of bad trades. What actually happened was this: Back on September 1, 2025, the EA had opened a large, highly profitable long position in gold at $3445. By January 30, 2026, this trade was sitting on a substantial unrealized profit (also known as “paper profit”). However, on that specific day, gold experienced a dramatic intraday drop, plummeting from an opening price of $5382 all the way down to $4678. That’s a massive 14.3% range drop in a single day! As gold prices fell, the unrealized profit from our big long position started to rapidly shrink – it was “spat out,” as the notes put it. Even though the trade was still overall profitable, this sudden reduction in paper profit caused the account’s effective equity (remember, that includes unrealized gains) to drop by a whopping 7.7% from its opening value. This, in turn, triggered our daily -5% loss limit, forcing the system to stop. So, in other words, the EA was disqualified not because it made bad trades that day, but because a very profitable, open position saw its paper gains temporarily evaporate during a sharp market correction. This wasn’t a reconstruction bug; it was a genuine, albeit nuanced, risk.

What I learned

This incident highlighted a fundamental tension in automated trading: the conflict between trend-following strategies and strict daily loss rules based on effective equity. Trend-following EAs aim to “let your profits run” – they hold onto winning trades to capture the biggest possible moves. But when you have a large open position with significant unrealized profits, a sudden market reversal can quickly erode those paper gains. If your daily loss limit is calculated using effective equity (which includes these unrealized profits), you can hit your limit and be forced to stop, even if the underlying trade is still well in the black and might recover later. It’s like having a high-performance race car (your trend-following EA) that’s built to go fast and cover long distances, but also having a very sensitive sensor that triggers a “stop” if the potential speed you could achieve (your unrealized profit) drops too quickly, even if you’re still ahead in the race. The “edge” of our EA – its ability to identify and profit from trends – is real. However, this experience showed us that protecting unrealized profits is absolutely crucial. We need mechanisms to safeguard those paper gains from sharp intraday pullbacks, especially when a single, large winning position dominates the account’s equity.

What’s next?

Our next step is to implement solutions that address this specific vulnerability. We’re looking into two main approaches:

1. Trailing Stops: These are dynamic stop-loss orders that automatically adjust as the price moves in your favor, locking in more profit as the trade progresses.
2. Partial Profit-Taking: This involves closing a portion of a profitable trade at certain milestones to realize some of the gains, reducing the overall exposure of the remaining position to market reversals. We plan to re-verify the M1 EA with these modifications to see if they effectively mitigate the intraday unrealized profit drawdown without hindering its trend-following capabilities.

A quick note on the data

It’s also worth mentioning that the gold prices observed in 2026, specifically exceeding $5000 and experiencing a 14% intraday range, seem quite extreme. We’ll be double-checking the integrity and health of our historical data for that period. There’s a possibility that these values might represent anomalies or data errors, which could affect the generalizability of these findings. Always good to verify your data!