Moving Average Crossover: 8 Pairs Tested on EURUSD (2026)

Moving average crossovers on EURUSD 5-minute charts lost money in seven of eight configurations I tested across 16,926 real bars from June 3 to August 26, 2026. The single exception was the EMA 50/200 golden cross, which won 51.1% of 88 signals and averaged +2.31 pips. The pattern is blunt: the slower the pair, the fewer the trades, and the better the result.

How were the crossovers tested?

Same dataset and same rules as my Stochastic test and my Bollinger squeeze test, so the three are directly comparable.

• Signal: the fast average closing above the slow one is a long, closing below is a short. No confirmation candle, no filter, no discretion.
• Pairs: 9/21, 12/26, 20/50 and 50/200, each run twice, once as EMA and once as SMA.
• Measurement: the move in pips two hours after the cross, in the signal direction. No stop, no target, so the number describes the signal and nothing else.
• Overlap rule: after a signal the clock skips forward by the hold length, so one move is never counted twice.

Which moving average pair actually made money?

Why do the fast pairs lose?

Because a 5-minute chart spends most of its life going sideways, and a fast crossover fires every time the range wobbles. Across the dataset the 12/26 pair produced 676 signals, and I measured clusters of up to five consecutive crosses with no move between them. Each one pays the spread.

Does holding longer fix it?

Partly. The win rate climbs steadily from 36.4% to 48.2% as the hold stretches, but it never crosses 50% and the average trade only reaches break even. Time alone does not rescue a fast crossover on this pair. Changing the pair does.

Why did the trend filter make it worse?

This is the most interesting result in the test, because it contradicts what I found with the Stochastic, where an EMA 200 filter turned a losing setup positive. Here the same filter dropped the win rate from 44.6% to 41.2%.

The reason is that a crossover is already a trend signal. Filtering it by a longer moving average does not add information, it just removes the trades taken early in a turn, which is exactly where a crossover earns whatever it earns. An oscillator has no view on direction, so a direction filter tells it something new. A moving average pair does have a view, so the filter only censors it. The same logic applies when you stack a crossover on top of Supertrend or a 2-line MACD, both of which are crossovers wearing different clothes.

What should you take from this?

• Use 50/200, or do not use crossovers on this timeframe. It was the only pair above 50%, and by a clear margin on average pips.
• Prefer EMA at the slow end. EMA 50/200 beat SMA 50/200 by nine percentage points on the same bars.
• Do not filter a crossover with another moving average. You are filtering a trend signal with a trend signal and paying for it in missed entries.
• Count your signals before your win rate. The 12/26 pair traded four times more often than 50/200 and turned a small edge into a steady leak.
• 88 signals is a small sample. The 50/200 result is a lead worth following, not a settled fact. I will re-run it as the dataset grows.

Frequently Asked Questions

What is the best moving average crossover for forex?

On this EURUSD 5-minute data, EMA 50/200 was the only profitable pair of the eight tested: 51.1% win rate, +2.31 pips per trade, 88 signals. Every faster pair lost money.

Is EMA better than SMA?

Only at the slow end here. EMA 50/200 beat SMA 50/200 by 9 percentage points, but SMA 20/50 beat EMA 20/50, and the two 9/21 versions finished within one point of each other. The pair length mattered far more than the averaging method. The full comparison is in my EMA against SMA article.

Why do so many courses teach 12/26?

Because those are the MACD defaults, and MACD is the most widely taught indicator there is. That is a historical accident, not a measured result. On this data 12/26 was the second worst pair of the eight.

Does the golden cross work on other pairs?

I have only measured EURUSD 5-minute bars over this window, so treat the number as specific to that. What generalises is the shape: slower pairs trade less and lose less.

Should I use a crossover as my only entry signal?

No. Even the winning pair produced 88 signals in three months, roughly one per day, with a small edge. It works better as a context filter for other setups than as a trigger on its own. Settings by timeframe are in my moving average settings guide.