Best Moving Average Crossover Settings: 280 USD/JPY Backtests
280 crossover settings backtested on USD/JPY for 2024 and 2025: which timeframe held up, where 10/20 ranked, how the 50/200 golden cross did, what survived both years.
Two moving averages, one faster than the other. Buy when the fast one crosses above, sell when it crosses below. The golden cross and the dead cross are about as old as technical trading gets.
The open question is which two periods, and which kind of average. 5 and 20, 10 and 20, 25 and 75, 50 and 200 all circulate freely, usually without a measurement attached.
So we tested 280 of them. Eight fast periods against eight slow ones gives 56 valid pairs; five averaging methods (simple, exponential, weighted, smoothed and Hull) multiply that to 280. Each one was backtested on USD/JPY for 2025 (1 January to 31 December) on three timeframes — 15-minute, one-hour and four-hour — and the whole grid was run again on 2024 for comparison.
The answer first. The best 2025 result was 12 and 50 weighted moving averages on the 15-minute chart: 751 trades, a 33.95% win rate and +2,978.0 pips at a profit factor of 1.259. But the finding that came through hardest is not that one. Run the same 280 settings on the four-hour chart and 243 of them were profitable in 2024 while only 60 were in 2025. Which timeframe you look at, and whether the year trended, moved the results far more than the choice of periods did.
What a crossover actually measures
A moving average is the last n bars smoothed into one line. Lengthen the period and the line drifts further from price and takes longer to change direction.
Watching two of them cross means watching which smoothing is on top, and taking the moment they change places. They change places when recent price has moved far enough away from the older average, which makes a crossover a proxy for "the trend may have turned".
There are three knobs.
| Knob | What raising or changing it does |
|---|---|
| Fast period | The quick line dulls. Fewer crossings, later reactions |
| Slow period | The slow line dulls. Fewer crossings, longer holds |
| Averaging method | At the same period, the line sits nearer or further from price. Nearer means quicker and more crossings |
The trade counts show it. On the one-hour chart in 2025, settings with a slow period of 20 averaged 398 trades for the year; with a slow period of 200, 108. The method moves it too: averaged across all 280 settings, Hull moving averages produced 455 trades a year and smoothed moving averages 86. Every knob here is partly a trade-count knob, which matters when the costs come up later.
How to put two moving averages on a chart
Moving averages ship with every platform. What differs is the list of methods available.
| Platform | Steps |
|---|---|
| MT4 / MT5 | Drag Moving Average from the indicator list twice and set each period. The method list is simple, exponential, smoothed and linear weighted — Hull is not among them |
| TradingView | Add Moving Average twice from the indicator search and set period and method on each |
| Browser (Formiq) | Add several moving averages and pick from simple, exponential, weighted, smoothed and Hull |
The numbers below cover all five methods. Hull moving averages are not part of the MT4 or MT5 built-ins, so read those rows accordingly if that is where you are working.
How this was tested
Before the numbers, what was measured. Entering this table reproduces these results.
| Item | Value |
|---|---|
| Pair | USD/JPY |
| Period | 2025-01-01 to 2025-12-31 (2024 run identically for comparison) |
| Timeframes | 15-minute / one-hour / four-hour |
| Bars tested | 24,903 on M15, 6,226 on H1, 1,610 on H4 for 2025 |
| Long | Close of the bar where the fast average crosses above the slow one |
| Short | Close of the bar where the fast average crosses below |
| Exit | The opposite crossover, reversing into the new direction at the same close. No stop, target or time limit |
| Spread | 0.3 pips, zero slippage, 0.1 lots. Charged on the exit and the new entry both |
| Grid | 8 fast periods (5, 8, 10, 12, 15, 20, 25, 50) × 8 slow periods (20, 25, 30, 40, 50, 75, 100, 200), keeping the 56 pairs where fast is shorter × 5 averaging methods = 280, per timeframe per year |
| Measurement | Run through Formiq's backtester, with pips re-derived from each trade's fill prices |
Fast and slow use the same method throughout the main grid. Mixing them is measured separately further down.
Holding to the opposite crossover with no stop keeps this a test of the crossover rather than of the crossover plus an exit. Stops and targets get their own section.
The timeframe decides most of it
Same 280 settings, three timeframes.
| Timeframe | Profitable in 2025 | 2025 median | Profitable in 2024 | 2024 median | Profitable in both |
|---|---|---|---|---|---|
| 15-minute | 231 / 280 | +810 pips | 234 / 280 | +1,420 pips | 205 / 280 |
| One-hour | 188 / 280 | +445 pips | 259 / 280 | +1,671 pips | 172 / 280 |
| Four-hour | 60 / 280 | −1,153 pips | 243 / 280 | +2,151 pips | 32 / 280 |
Judged on 2024 alone the four-hour chart looks like the answer: the highest median of the three and 243 settings in profit. In 2025 that collapses to 60 settings and a median of −1,153 pips. Only 32 of 280 were profitable in both years.
The 15-minute chart is the mirror image. It has the lowest 2024 median of the three, and the most settings that came through both years at 205.
The reason is what the two years did. USD/JPY opened 2024 at 140.873 and closed at 157.197 — up 1,632 pips, with a 2,237-pip range. In 2025 it opened at 157.227 and closed at 156.670, down 56 pips, with a 1,900-pip range. Nearly the same distance travelled; one year arrived somewhere and the other came home. A crossover system is built to catch trends, so that difference lands directly on the results — and the longer the timeframe, the harder it lands.
Where the textbook pairs ranked
The combinations people arrive already having read about, ranked within the 280 for that year and timeframe.
| Combination | M15 2025 | M15 2024 | H1 2025 | H1 2024 |
|---|---|---|---|---|
| 5 / 20 simple | +264.5 (212nd) | +646.1 (207th) | −484.9 (211th) | +3,744.7 (4th) |
| 10 / 20 simple | +1,015.0 (110th) | +1,600.9 (119th) | +377.4 (151st) | +2,888.4 (38th) |
| 12 / 25 exponential | +1,436.6 (69th) | +2,102.1 (73rd) | +970.9 (68th) | +440.6 (239th) |
| 20 / 50 simple | +1,774.9 (46th) | +1,464.8 (132nd) | −220.5 (202nd) | −34.6 (261st) |
| 25 / 75 simple | −1,023.1 (270th) | +3,613.7 (3rd) | −56.8 (192nd) | −253.1 (265th) |
| 50 / 200 simple | +629.3 (171st) | +301.2 (225th) | −1,340.4 (246th) | +2,435.4 (77th) |
(net pips)
25 and 75 simple ranked 3rd of 280 on the 15-minute chart in 2024 and 270th in 2025. The same setting moved from one end of the distribution to the other. 5 and 20 simple did the same thing on the one-hour chart: 4th in 2024, 211th in 2025.
The 10 and 20 simple default landed at 110th and 119th on the 15-minute chart, 151st and 38th on the one-hour. Mid-table in every year. On the four-hour chart, though, it lost in both years: −1,336.2 pips in 2025 (165th) and −1,245.7 pips in 2024 (272nd).
Worth naming the losers too. The worst 2025 result anywhere was 25 and 30 simple on the four-hour chart: 80 trades for −3,582.6 pips — and that same setting made +567.3 pips in 2024. The worst on the 15-minute chart was 8 and 20 Hull, 3,463 trades for −2,098.1 pips, which also lost 2,248.2 pips in 2024.
The 50/200 golden cross, and why trade counts belong next to every ratio
The most famous pair deserves its own look.
| Timeframe | 2025 | 2024 |
|---|---|---|
| 15-minute | 147 trades, 38.10% win, +629.3 pips | 147 trades, 29.93% win, +301.2 pips |
| One-hour | 47 trades, 31.91% win, −1,340.4 pips | 27 trades, 55.56% win, +2,435.4 pips |
| Four-hour | 14 trades, 50.00% win, +173.6 pips | 7 trades, 71.43% win, +1,357.5 pips |
A 71.43% win rate and a profit factor of 2.269 look like a result until you notice that there were 7 trades. Five of seven went the right way; nothing beyond that is supportable. The one-hour version of the same year managed a profit factor of 2.305 on 27 trades.
The grid holds a sharper example. Third place for profit on the four-hour chart in 2025 went to 20 and 200 smoothed: 2 trades, a 100% win rate, +1,563.3 pips. In 2024 the same setting made 7 trades for +1,251.3 pips and ranked 211th.
Fewer trades, louder ratios, and a rank that moves next year. The same pattern turned up in QQE and Supertrend tested the same way. Read every win rate and profit factor with the trade count beside it.
The longer the slow line, the more the year decides
Moving one knob at a time and averaging over everything else, one relationship came through clearly.
By slow period, on the one-hour chart:
| Slow period | 2025 mean | 2024 mean | Profitable in both |
|---|---|---|---|
| 20 | +736 pips | +1,899 pips | 20 / 25 |
| 25 | +992 pips | +1,798 pips | 30 / 30 |
| 30 | +822 pips | +1,474 pips | 31 / 35 |
| 40 | +709 pips | +1,388 pips | 26 / 35 |
| 50 | +567 pips | +1,468 pips | 27 / 35 |
| 75 | −129 pips | +1,293 pips | 17 / 40 |
| 100 | −158 pips | +1,679 pips | 17 / 40 |
| 200 | −1,179 pips | +1,964 pips | 4 / 40 |
The 2024 column barely moves. In a trending year, long lines and short lines both make money. The 2025 column falls the whole way down and changes sign along the way. All 30 settings with a slow period of 25 were profitable in both years; with a slow period of 200, 4 of 40 were.
The 15-minute chart shows a much weaker version of the same thing. A slow period of 200 there still averaged +343 pips in 2025 and +1,501 pips in 2024, with 24 of 40 profitable in both. Drop the timeframe and even long lines stop being a bet on the year.
The fast period tells a smaller version of the story. On the one-hour chart the 15 settings with a fast period of 50 averaged −693 pips in 2025 and only 2 came through both years. Fast periods from 5 to 15 all landed between +259 and +360 pips in 2025, with 25 to 30 of each 40 profitable in both years — essentially no difference between them.
Does last year's best setting work this year?
This is the part that decides whether any of the above is usable. Take the winner of one year and run it in the other.
| Timeframe | 2024 winner | Its 2024 result | Same setting in 2025 | 2025 rank | 2025 median |
|---|---|---|---|---|---|
| 15-minute | 20 / 75 simple | +3,995.2 pips (387 trades) | −234.6 pips | 237 / 280 | +809.8 pips |
| One-hour | 5 / 25 simple | +3,828.5 pips (309 trades) | +583.3 pips | 115 / 280 | +445.0 pips |
| Four-hour | 8 / 50 weighted | +3,378.8 pips (44 trades) | −1,367.2 pips | 170 / 280 | −1,152.9 pips |
On two of the three timeframes, picking last year's best setting did worse than picking at random. Only the one-hour chart came out slightly above its median.
Going the other way is kinder. The 2025 winner on the 15-minute chart, 12 and 50 weighted, made +2,103.7 pips in 2024 and ranked 70th of 280 — near the top in both years. The one-hour winner (5 and 30 Hull) ranked 226th in 2024, and the four-hour winner (12 and 50 Hull) lost 733.9 pips in 2024 at 263rd.
Rank correlation across all 280 settings puts a number on it.
| Timeframe | 2024 vs 2025 | First half of 2025 vs second |
|---|---|---|
| 15-minute | +0.385 | +0.103 |
| One-hour | −0.178 | −0.171 |
| Four-hour | −0.416 | −0.187 |
The four-hour figure is negative: the better a setting did last year, the worse it tended to do this year. Only the 15-minute chart carried a weakly positive ranking across years. That is new for this series. Neither QQE nor Supertrend, tested the same way, produced a timeframe whose ranking carried forward at all — though the Bollinger breakout does something similar.
It moves inside a single year too. 10 and 20 simple on the one-hour chart lost 548.7 pips over 169 trades in the first half of 2025 and made 885.9 pips over 189 in the second. 50 and 200 simple went the other way: +306.5 pips over 21 trades, then −1,576.0 over 26.
Which averaging method — it depends on the timeframe
Five methods, counted as "settings profitable in both 2024 and 2025 / settings tested".
| Method | 15-minute | One-hour | Four-hour |
|---|---|---|---|
| Simple (SMA) | 48 / 56 | 28 / 56 | 2 / 56 |
| Exponential (EMA) | 52 / 56 | 34 / 56 | 2 / 56 |
| Weighted (WMA) | 49 / 56 | 39 / 56 | 8 / 56 |
| Smoothed (SMMA) | 43 / 56 | 22 / 56 | 13 / 56 |
| Hull (HMA) | 13 / 56 | 49 / 56 | 7 / 56 |
Hull moving averages came last on the 15-minute chart at 13 of 56 and first on the one-hour at 49 of 56. One method, opposite verdicts, decided by the timeframe. Smoothed averages do something similar in reverse: 43 of 56 on the 15-minute chart, 22 on the one-hour, and the best of the five on the four-hour.
The means agree. Hull on the 15-minute chart averaged +243 pips in 2025 and −449 pips in 2024, where only 16 of 56 settings were profitable. Hull on the one-hour chart averaged +1,155 pips and +1,565 pips.
Trade counts explain the 15-minute case. Hull settings there averaged 1,852 trades a year against 324 to 875 for the other four — twice the nearest and more than five times the smoothed averages. Costs scale with trade count, as the last section shows.
There is no answer to "which average is best" here that does not name a timeframe alongside it.
What happens when you mix methods
Different methods on the fast and slow legs come up often enough to test, so the full 5 × 5 grid was run on the one-hour chart.
First, why this is not a matter of taste. The figure below is how far each method sits from the close, on average.
A 20-period Hull average sits 18.7 pips from the close; a 10-period simple average sits 22.1 pips away. The one with double the period is the closer of the two. Put a Hull on the slow leg of a 10/20 pair and the "slow" line is the quicker one.
The results follow. On 10 and 20, the four combinations with Hull on the slow leg only had 1 of 4 profitable in 2025 (mean −420 pips) and 0 of 4 in 2024 (mean −2,953 pips). The four with Hull on the fast leg only were profitable 4 of 4 in both years, averaging +820 and +3,089 pips. The five matched pairs were profitable 5 of 5 in both years at +978 and +2,106 pips.
At the extreme: 10 and 20 with a simple fast line and a Hull slow line lost 3,393.4 pips over 385 trades in 2024. Swap them — Hull fast, simple slow — and the same year returns +3,739.2 pips over 397 trades. Same periods, opposite sign.
This is not a universal rule. It did not repeat on 25 and 75, where the four Hull-on-the-slow-leg combinations were all profitable in 2024 at a mean of +845 pips. The figure explains that as well: a 75-period Hull sits 39.7 pips from the close and a 25-period simple 38.1 pips, so the two never change places.
The practical version: if you mix methods, check which line is actually faster instead of reading it off the periods.
Does a higher win rate mean more profit?
Not consistently, and the timeframe changes the answer.
| Timeframe | Rank correlation of win rate and profit, 2025 | Win rate range |
|---|---|---|
| 15-minute | −0.110 | 19.30% – 44.22% |
| One-hour | +0.714 | 12.24% – 48.75% |
| Four-hour | +0.680 | 10.94% – 100% |
On the 15-minute chart the relationship is slightly negative. The highest win rate of 2025 — 15 and 20 Hull at 44.22% over 2,680 trades — ranked 263rd of 280 for profit at −755.1 pips. The profit leader, 12 and 50 weighted, won 33.95% of its trades: +56.74 pips per winner against −23.17 pips per loser, a 2.4-to-1 ratio that carries a one-in-three hit rate.
The one-hour and four-hour correlations are positive, but the four-hour range topping out at 100% is the two-trade setting from earlier, so that figure rests on a handful of rows.
A trend-following rule landing in the 30–40% band and still making money matched the rest of this series. Mean-reversion indicators run the other way — higher win rates, negative results (Williams %R). Ranking settings or indicators by win rate can invert the ranking by profit.
Filters, stops and targets
Two 15-minute settings, with the filters that usually get suggested.
| Condition | 2025 | 2024 |
|---|---|---|
| 10/20 simple, crossover only | 1,402 trades, 37.52%, +1,015.0 pips | 1,397 trades, 38.08%, +1,600.9 pips |
| + ADX ≥ 20 | 674 trades, 35.31%, −495.4 pips | 705 trades, 37.02%, +923.1 pips |
| + ADX ≥ 25 | 354 trades, 34.75%, −271.0 pips | 400 trades, 40.25%, +1,847.9 pips |
| + ADX ≥ 30 | 170 trades, 34.71%, +195.7 pips | 217 trades, 37.33%, +703.8 pips |
| + London and New York hours (UTC 7–21) | 802 trades, 37.16%, +846.9 pips | 797 trades, 37.77%, +41.1 pips |
| + Tokyo hours (UTC 0–8) | 460 trades, 38.91%, +309.7 pips | 443 trades, 39.73%, +2,194.9 pips |
The ADX filter helped in 2024 and hurt in 2025. ADX ≥ 25 beat the unfiltered version in 2024 at +1,847.9 pips and erased the profit in 2025 at −271.0.
The session filter reads differently per trade. Tokyo hours in 2024 made +4.95 pips a trade against +1.15 unfiltered. In 2025 it made +0.67 against +0.72 — marginally worse. On the profit leader, 12 and 50 weighted, Tokyo hours made +5.04 pips a trade in 2025 against +3.97 unfiltered, and +7.22 against +2.57 in 2024. Three of the four setting-and-year combinations improved per trade; not all four.
Stops and targets, on the same two settings:
| Exit rule | 10/20 simple 2025 | 10/20 simple 2024 | 12/50 weighted 2025 | 12/50 weighted 2024 |
|---|---|---|---|---|
| Opposite crossover only | +1,015.0 pips | +1,600.9 pips | +2,978.0 pips | +2,103.7 pips |
| SL 30 / TP 60 | +189.7 pips | +124.3 pips | +1,809.6 pips | +547.7 pips |
| SL 50 / TP 100 | +510.8 pips | +932.1 pips | +2,475.1 pips | +749.2 pips |
| SL 30 / TP 90 | +558.9 pips | +1,113.2 pips | +2,268.9 pips | +1,361.3 pips |
| Time exit, 24 bars | +621.7 pips | +273.0 pips | +641.2 pips | −108.5 pips |
All sixteen were worse than simply holding to the opposite crossover. Two settings, two years, four exit rules, no exceptions.
The one-hour chart does not agree. There, 10 and 20 simple with a 30-pip stop and a 90-pip target made +1,224.0 pips in 2025 against +377.4 for the plain version — then made +980.8 pips in 2024 against +2,888.4. On the 15-minute chart, leaving it alone won both years; on the one-hour chart the answer swapped.
What the spread costs is the trade count
Re-running 2025 with nothing but the spread changed:
| Spread | M15 10/20 simple (1,402 trades) | M15 12/50 weighted (751) | H1 10/20 simple (358) | H4 50/200 simple (14) |
|---|---|---|---|---|
| 0.0 pips | +1,435.6 | +3,203.3 | +484.8 | +177.7 |
| 0.3 pips | +1,015.0 | +2,978.0 | +377.4 | +173.6 |
| 0.6 pips | +594.4 | +2,752.7 | +270.0 | +169.4 |
| 1.0 pips | +33.6 | +2,452.3 | +126.8 | +163.8 |
| 1.5 pips | −667.4 | +2,076.8 | −52.2 | +156.8 |
| 2.0 pips | −1,368.4 | +1,701.3 | −231.2 | +149.8 |
Going from 0 to 2.0 pips costs 2,804.0 pips on the 15-minute 10/20 — exactly 1,402 trades × 2.0. The others: 1,502.0 pips for 12/50 (751 × 2.0), 716.0 pips for the one-hour 10/20 (358 × 2.0), and 27.9 pips for the four-hour 50/200 against the 28.0 that 14 × 2.0 predicts, a rounding difference.
The practical reading: the 15-minute 10/20 default is roughly break-even at a 1.0-pip spread and loses at 1.5. The 12/50 setting on the same chart, at half the trade count, still returns +1,701.3 pips at 2.0 pips. Checking your own spread can matter more than moving a period by two.
What this test supports
Only the things that pointed the same way in both years.
- Choose the timeframe first. 205 of 280 settings were profitable in both years on the 15-minute chart, 172 on the one-hour, 32 on the four-hour — and the four-hour chart had 243 winners in 2024 alone. That is too wide a swing to plan around
- The longer the slow line, the more the result is just whether the year trended. On the one-hour chart, all 30 settings with a slow period of 25 survived both years; 4 of 40 did with a slow period of 200
- The averaging method only means something paired with a timeframe. Hull was 13 of 56 on the 15-minute chart and 49 of 56 on the one-hour
- If you mix methods, verify which line is faster. A 20-period Hull sits closer to price than a 10-period simple, and putting it on the slow leg swapped the two lines' roles
- Do not pick last year's best setting. It ranked 237th of 280 on the 15-minute chart and 170th on the four-hour, below the median both times
Narrowing the periods to one pair produced nothing usable within the scope of this test. What did matter was the timeframe, avoiding very long pairs, and cost. If you want to run the same tables on your own pair and your own dates, building the conditions without writing code covers how.
Limits of this test
- One pair, USD/JPY, and two years. Nothing here guarantees the same pattern on other pairs or other years
- The two years are very different. 2024 rose 1,632 pips through a 2,237-pip range; 2025 finished 56 pips lower through a 1,900-pip range. A crossover system is built for trends, so that difference produces most of what is above. A third year could well change the conclusions
- Entries and exits happen at bar closes. Real fills differ
- The reversal happens at the same close that ends the previous trade. In practice there is a gap between the two
- The spread is a flat 0.3 pips throughout. Real spreads move with the session and around releases
- Settings with a slow period of 200 build their line from that many bars of history before the test window opens. Opening a fresh chart does not necessarily give you the same line
- Price moves by different amounts at different hours. The session-filter results carry that effect
Questions people ask
- What are the best moving average crossover settings?
- On USD/JPY in 2025 the best result came from 12 and 50 weighted moving averages on the 15-minute chart: 751 trades, a 33.95% win rate and +2,978.0 pips. That is not a recommendation. The setting that won 2024 on the same chart — 20 and 75 simple — returned −234.6 pips in 2025 and ranked 237th of 280, below the year's median of +809.8 pips.
- Which moving average periods work best?
- The slow period mattered far more than the fast one. On the one-hour chart every one of the 30 settings using a slow period of 25 was profitable in both 2024 and 2025; with a slow period of 200 only 4 of 40 were. A slow period of 200 averaged +1,964 pips in 2024 and −1,179 pips in 2025 — the sign flips with the year.
- What is the win rate of a golden cross?
- Across 280 settings on the one-hour chart in 2025 win rates ran from 12.24% to 48.75%. Held to the opposite crossover with no stop, a crossover system wins less than half its trades and makes money on the size of the winners. The best 15-minute setting won 33.95% of its trades, averaging +56.74 pips per win against −23.17 pips per loss.
- Does the 50/200 golden cross work on forex?
- On the four-hour chart 50 and 200 simple made 7 trades in 2024 at a 71.43% win rate for +1,357.5 pips, and 14 trades in 2025 at 50% for +173.6 pips. Seven trades is not a sample. The same pair on the one-hour chart lost 1,340.4 pips over 47 trades in 2025, ranking 246th of 280.
- Which timeframe suits a moving average crossover?
- Running the same 280 settings on three timeframes, the number profitable in both 2024 and 2025 was 205 of 280 on the 15-minute chart, 172 on the one-hour, and 32 on the four-hour. The four-hour chart had 243 profitable settings in 2024 and 60 in 2025 — the widest year-to-year swing of the three.
Formiq is a free browser-based FX terminal with replay practice and no-code backtesting. Open the chart or see what the free plan includes.