What the 30 January 2026 gold drop did to grid, martingale and stop-loss EAs
On Friday 30 January 2026, gold had its sharpest one-day fall since 1983. For anyone running an EA on XAUUSD, that day is now the most useful stress test in the data. Here is what was reported, and what it means for each type of EA.
The day in numbers (as reported)
- Record first. Spot gold set an all-time high of about $5,595 an ounce on Thursday 29 January.
- The fall. On 30 January spot gold fell about 9.5% to 9.8%. Reuters described it as the biggest one-day percentage drop since 1983.
- Silver was worse. Spot silver fell roughly 26% to 28%, reported as its worst day on record.
- Reasons given. The nomination of Kevin Warsh as Fed chair (read as hawkish), a stronger dollar, and profit-taking after a crowded rally.
- Margin after the fall. CME announced higher gold and silver futures margins on 30 January, effective after the close on Monday 2 February. Selling continued on the Monday.
Exact intraday lows and broker-level spread or margin changes were not published, so we don't estimate them.
How far can a grid survive? A worked example
Take a $10,000 account, a grid that buys 0.01 lot (1 oz) at the first level and adds one position every fixed step as price falls, with no take-profit hit on the way down. We ignore spread, swap and the broker's stop-out level (a real stop-out comes earlier, so real results are worse). The question is simple: how far does gold have to fall before floating loss equals the whole balance?
| Grid step | Lot multiplier | Move that wipes $10k | Positions | Total lots |
|---|---|---|---|---|
| $10 | 1.0 (flat) | ~$442 (≈8.0%) | 45 | 0.45 |
| $10 | 1.2 | ~$199 (≈3.6%) | 20 | 1.87 |
| $10 | 1.5 | ~$126 (≈2.3%) | 13 | 3.87 |
| $10 | 2.0 | ~$90 (≈1.6%) | 10 | 10.23 |
| $20 | 1.0 (flat) | ~$622 (≈11.3%) | 32 | 0.32 |
| $20 | 1.2 | ~$330 (≈6.0%) | 17 | 1.06 |
| $20 | 1.5 | ~$221 (≈4.0%) | 12 | 2.57 |
| $20 | 2.0 | ~$160 (≈2.9%) | 9 | 5.11 |
Percentages use a $5,500 reference price.
A fall of roughly 9.5–9.8% from that level is about $520–540. Read against the table:
- A flat grid with a $10 step is gone before the day's move is complete.
- A flat grid with a $20 step survives on paper, but with very little left once a real stop-out level is applied.
- Any lot multiplier wipes the account after a 2–6% move. That is a fraction of what gold did in one session.
The lesson is not "grids are bad". It is that a grid's survival is set by three numbers you choose before the day starts: step size, lot growth and free margin. Its average past result tells you nothing about that day.
Stop-loss EAs
A stop limits the loss only if it fills near its price. In a fast market, fills can be worse than the stop and spreads widen. A model tested with a fixed spread and perfect fills will look safer than it was.
Trend-following EAs
A short-side trend model could gain on the day itself. The risk moves to what follows: after a record move, wide ranges and sharp reversals can produce a run of false signals.
Margin
Exchange margin rose after the fall. Brokers set their own CFD margin. Any EA that runs close to its margin limit should be tested against a higher requirement, not only against price.
A five-point test for your own EA
- Make sure your test data covers 29 January to 2 February 2026, on real ticks if available.
- Run it at today's lot size and balance, not those of the original optimization.
- Widen the spread and add slippage for the session of the fall.
- Re-run with a higher margin requirement and check free margin at the worst point.
- Test silver separately. Its move was close to three times larger in percentage terms.
Limits
Event figures come from public reporting and differ slightly between sources, so we give ranges. The grid table is arithmetic on a straight-line fall, not a test of any real EA. Real days have bounces, which can help a grid or trap it deeper.
What did this day do to the EAs you run?
