With gold in 2026, why the martingale was burned

Oct 29, 2026 ·

Self-made gold martingaleI ran an EA on data from July 2024 to October 2026. Until February 2026, it was almost always a clean upward trend with monthly profit. In the seven months from March, it spit out all the previous profits.

Then I changed the parameters, added features, and ran about a dozen backtests. Still,it did not become a winning martingale for gold in 2026.

This article is a record of that failure. All numbers are from our backtests (GOLD, all ticks, modeling quality 99.90%).(GOLD, all ticks, modeling quality 99.90%).

What happened in 2026

The profits accumulated over 19 months were lost in just four withdrawals.

 

Our martingaleEA (Zanshin v3.30) backtest · GOLD 2024 Aug–2026 Oct · initial capital $30,000 · 0.01 lot

LossThe four losses all followed the same pattern. Enter the trend, and after several weeks there would be a 500–1,000 USD correction, add a martingale step daily, and terminate when all steps were filled.

Withdrawal date	Direction	Loss	What happened
March 23, 2026	Buy5 levels	−$3,037	Crash from $5,200
June 24, 2026	Buy6 levels	−$2,711	Drop from $4,650 to $4,015
August 21, 2026	Sell6 levels	−$2,218	Rebound from $4,050 to $4,590
October 1, 2026	Buy5 levels	−$1,529	Decline from $4,640 to $4,180 (unrealized loss at test end)

If you buy, it goes down; if you sell, it goes up. A back-and-forth market. Moreoverdaily moves expanded several times compared to 2024. The gaps between martingale levels were jumped over in a single day.

Records of ongoing improvements

What I did can be summarized into four items. Each could change the way you lose, but the act of losing itself couldn’t be changed.

What I did	Aim	Result
① Widen the martingale spacing to match daily price movement (ATR)	Prevent full averaging out in a matter of hours	OriginallyEven 12 levels only had about $10 width. Widening increased the buffer, but the unrealized loss when all levels filled also grew (on a $10k account, −$3,543, max DD 53%).
② Create a withdrawal line	After all levels fill, if it moves against you further, close all	No single blow to the account, but each withdrawal loses about 10–40% of the balance
③ Cap a single loss (10% of balance) and skip entries if close	Pause during turbulent markets
④ If price moves beyond a certain % of the range, pause regardless of account size	Fill the gap left by #3

The result of #3 was the harshest. The $10k account was able to pause through 2026 largely because of lack of funds. Those with more capital can enter volatile markets and still lose, which is fatal for an EA.

Also, with any line, there was either “keep the loss” or “erase profits.” The entry that loses and the entry that wins could not be distinguished by market intensity.

Tried doubling down again

“If it goes the other way, switch at the withdrawal point.” I tested price action after withdrawals.All three times, it failed.

Withdrawal	Withdrawal price	If doubled down	Subsequent price movement	Result
March 23	$4,358 (withdraw buy)		Rose to $4,650 in May
June 24	$4,015 (withdraw buy)		Rose to $4,590 in August
August 21			Declined to $4,180 in October

Withdrawing at the point of maximum counter-move means entering opposite positions at the worst possible timing.

Why it didn’t fix itself

Martingale is built on “add when price moves against you.” It is strong in ranges that bounce up and down, but structurally loses in trends. This cannot be fixed by parameters.

Budget and lot size determine how much you can endure. Increasing levels only divides the same budget into finer steps.With $30k and 0.01 lots, endurance was about $700. The 2026 adjustments were $650–$900.

Even with the same lot, risk grows with price movement. As gold price rose and daily moves widened, the worst-case single loss with 0.01 lots multiplied.

You exhaust the budget while prices are falling. Each day of adding one level during a crash drains the ammo before hitting bottom.

From here, the story continues(Free registration)

So what did I do? I stopped martingale.

I flipped “add when counter-trending” to “add when trending.” Ride the trend, and add the same lot as it progresses in your favor(pyramiding). If it goes the other way, you decide and exit immediately when you enter. That’s all.

The same data, same period produced this result.

 

Zanshin v4.03 backtest · GOLD 2024 July–2026 Oct · All ticks · Initial capital $10,000 · 0.01 lot

MartingaleEven in March–October 2026, when there were four withdrawals and big declines, it did not collapse. The March crash occurred from martingale buying the dips, but trend-following was riding the sell side.

However, the way of winning is completely different.

 

Zanshin v4.03 backtest · 53 entries divided by holding time

Martingale is “small wins many times, occasional large loss.” The trend-type is the opposite, “small losses many times, occasional large wins.” It can lose up to a maximum of9 consecutive losses. That said, a single loss maximum is $194 (0.01 lot), so you won’t blow the account.

ThisEA is under review as “Zanshin EA” by Chicken Farm on GogoJungle.

Summary

•Gold in 2026 moved up and down by $500–$1,000 in a matter of weeks. For martingale, the hardest form.

•Spacing, levels, withdrawal lines, waiting, doublings. All could change how you lose, but you couldn’t erase the loss.

•If you use martingale EA, remember that more capital does not guarantee safety. Those with more funds can get into rough markets and lose.With current gold, calculate what your one-loss exposure would be at your chosen lot size.

•Even with the same lot, increased price movement multiplies risk. Please compute in advance what your “worst one loss” would be under current gold conditions.Note: All published numbers are from past data (GOLD, July 2024–October 2026) backtests. They do not guarantee future results. This period featured a gold price uptrend favorable to trend-following.— CHICKEN FARM (farm of the cowards)

Note: All published numbers are from past data (GOLD, July 2024–October 2026) backtests. They do not guarantee future results. This period featured a gold price uptrend favorable to trend-following.

— CHICKEN FARM (farm of the cowards)