RSI Settings for Swing Trading Tested: RSI 14 at 30/70 Lost 0.94 Pips on M5 and Made 4.98 on H1, and H4 Fired Only 10 Times

Almost every RSI test on this blog so far was about scalping on the 5-minute chart, and almost every one lost after costs. The most common reply I get is: “RSI is a swing tool, test it on H1 or H4.” That is a fair point and it is testable, so I tested it on the same data.

There is no separate hourly file here. I built M15, H1 and H4 bars out of the same 16,926 M5 bars I used for the RSI 7 vs RSI 14 scalping test, so the only thing that changes between rows is the timeframe.

What exactly did I test?

• Data: EURUSD, June 3 to August 26, 2026: 16,926 M5 bars, which make 5,642 M15 bars, 1,417 H1 bars and 345 H4 bars. Gold (XAUUSD), June 16 to August 26: 13,736 M5 bars, 1,146 H1 and 292 H4.
• RSI: Wilder’s formula, the same one MT4 uses, periods 7, 9, 14 and 21.
• Fade rule: buy when RSI crosses back up through the lower level, sell when it crosses back down through the upper one. Levels 30/70, 20/80 and 40/60.
• Trend rule: buy when RSI crosses above 50, sell when it crosses below.
• Exit: after 3, 6 or 12 bars. On H1 that is 3, 6 or 12 hours; on H4 it is half a day, a day or two days.
• Cost: 1 pip per EURUSD trade from my spread and swap measurement, $0.30 per gold trade.
• Checks: the data split at July 15 into two halves, and a coin: random entries on the same bars, 200 runs, same hold.

Does RSI work better on H1 than on M5?

On EURUSD, yes, and by a lot. Here is the textbook setting, RSI 14 at 30/70 held 6 bars, on each timeframe:

Pips per trade. Halves split at July 15, 2026.

Look at the “before cost” column first. On M5 the signal was worth 0.06 pips before the spread, which is nothing, so the 1-pip cost decided the result. On H1 the same rule was worth almost 6 pips before cost, and the pip hardly mattered. H4 went the other way, but on 10 trades.

How many RSI settings made money on each timeframe?

“Typical 6-bar move” = average absolute price change over 6 bars. “1 pip is” = the cost as a share of that move. The grid has 48 settings per timeframe; rows with fewer than 30 trades are left out.

Two things are going on. The cost shrinks as the bars get bigger: 1 pip eats 28.2% of a normal 30-minute move on M5 and 7.6% of a 6-hour move on H1. And on H1 the signal itself was better: 32 of 36 settings were positive before any cost, against 5 of 45 on M5. The first is arithmetic and will hold on any data. The second is what this sample showed, and it is the part to be careful with.

The 36 H1 settings also share most of their signals. RSI 14 at 30/70 held 3, 6 and 12 bars is the same 32 entries three times, so 26 of 36 is not 26 separate pieces of evidence.

Is the H1 result better than luck?

Only just, and not for every setting. I ran random entries on the same H1 bars, the same number of trades, the same hold, 1,000 times:

Pips per trade after 1 pip of cost. Coin = random direction on random H1 bars, 1,000 runs per row.

With 32 trades, a coin gets lucky by 4 or 5 pips a trade one run in twenty, so RSI 14’s +4.98 is a hint, not a proof. RSI 7 at 30/70 gave more trades, 113, and was positive in both halves (+1.38, then +4.58). That is the setting I would check first on your own data. The rule vs random entry test explains why I always put a coin next to a result.

What happened on H4?

Not enough. RSI 14 needs a real move to reach 30 or 70, and on H4 that took days. Twelve weeks gave 345 H4 bars and 10 signals at 30/70. The 10 trades lost 11.44 pips each, but a coin with 10 trades on the same bars ranged up to +18 pips in its best 5% of runs, so the loss says almost nothing either way.

The honest summary for H4: to test an H4 RSI swing setup you need years of bars, not months. Every H4 row in my grid with more than 30 trades used the looser 40/60 levels or the 50 cross, and only 6 of 18 were positive.

Does RSI on H1 work on gold too?

No. Gold’s cost was already small on M5, 4.8% of a normal move, and RSI still lost there, so cost was never gold’s problem. On H1, RSI 14 at 30/70 made +$1.85 a trade overall but +$7.57 in the first half and -$2.03 in the second, and a coin on 42 gold H1 trades reached +$7.53 in its best 5%. Gold behaved differently from EURUSD with the same rules and code before, and it did again here.

So what are the best RSI settings for swing trading?

• Use H1, not M5, if you trade RSI levels at all. On EURUSD the same rule went from -0.94 to +4.98 pips per trade. The cost share alone falls from 28.2% to 7.6%.
• Try RSI 7 at 30/70 on H1 before the default 14. 113 trades, +3.02 pips each, positive in both halves. RSI 14 was better per trade on 32 trades and within reach of luck.
• Keep 30/70 on H1. 20/80 fired 9 times in 12 weeks at period 14. 40/60 fired 121 times and made +1.30, less than 30/70.
• Do not trust the 50-line cross for entries. On H1 it made +0.97 over 141 trades, inside what a coin does.
• Treat H4 RSI results with suspicion, mine included. Ten signals in twelve weeks cannot tell a good setup from a bad one.

How to set this up on the platform is in my RSI swing trading setup for MT4, and why RSI with a longer period on M5 is not the same thing as RSI on H1 is in swing RSI vs scalping RSI. For RSI as a turning-point tool on M5, see the RSI divergence backtest.

Frequently asked questions

What is the best RSI setting for swing trading?

On EURUSD H1 in this test, RSI 7 with 30/70 levels, held 6 hours: +3.02 pips per trade over 113 trades after a 1-pip cost, positive in both halves of the data. RSI 14 with 30/70 made +4.98 but on only 32 trades.

Is RSI better on H1 or H4 for swing trading?

H1 had enough signals to measure: 26 of 36 settings made money on EURUSD. H4 did not: RSI 14 at 30/70 fired 10 times in 12 weeks, too few to say whether it works.

Should I use RSI 14 or RSI 21 for swing trading?

RSI 21 at 30/70 fired 19 times on EURUSD H1 and lost 3.14 pips per trade. The longer the period, the rarer the 30/70 signals, and on 12 weeks of H1 bars RSI 21 was too slow.

Does the RSI swing setup work on gold?

Not in this test. 13 of 36 gold H1 settings made money, and the best-known one, RSI 14 at 30/70, went from +$7.57 per trade in the first half to -$2.03 in the second.

Why does RSI lose on M5 but not on H1?

On M5 the signal was worth 0.06 pips before cost, so the 1-pip spread decided everything. On H1 the same signal was worth 5.98 pips before cost and the spread took 1 of them.