RSI Settings Tested: RSI 7 vs RSI 14 Across 1,376 Trades, and Why 30/70 Reversion Lost to a Coin Flip (2026)

“Best RSI settings” pages hand out a grid: 7 for scalping, 14 for day trading, 21 for swing. I used to publish one. Then I ran the periods against each other on 16,926 EURUSD 5-minute bars, June 3 to August 26, 2026, with identical rules, and the grid did not survive. Here is what did.

RSI 7 or RSI 14: which period made money?

The argument for a fast RSI is that it catches momentum bursts while they still have room to run. The measurement says the extra trades cost more than they found. Same hit rate within three points, 2.7 times the trades, a quarter of the R and more than twice the drawdown. Every additional signal is a spread paid, which is why the number in spread and swap costs on MT4 matters more to a fast setting than to a slow one. The full test, including the New York session split, is RSI 7 vs RSI 14 for scalping.

Does the 30/70 overbought and oversold rule work?

Not on its own, and not by a small margin. RSI 14 reversion at 30 and 70, held two hours, is one of five rules I ran against 2,000 random portfolios on the same bars.

Buying at 30 and selling at 70 did worse than three quarters of the coin flips. The Stochastic version of the same trade looks the same: 44.7% across 1,075 signals at 30 minutes, and only a 200-period moving average filter turned it positive, per the Stochastic overbought test. Read RSI as a momentum gauge around the 50 line, in the direction of a slow trend gate; the tests in the moving average crossover test say that gate should be an EMA 200, not anything faster.

Should RSI be confirmed with a second oscillator?

No, because the second oscillator is mostly silent when RSI fires. Under identical rules across 1,997 signals, RSI produced 800 signals on a 30-minute hold to the Stochastic’s 1,197, the two finished within half a point on win rate, and they agreed on the same bar only 149 times. With a 200-period moving average filter and a four hour hold the Stochastic reached 51.3% while RSI stayed under 48%. That is Stochastic vs RSI, and Stochastic settings for MT4 shows the same lesson from the other side: five settings differed by 2.3 points, while the hold time moved results four times as much.

What moved RSI results more than the period?

Session first. EURUSD moved 2.75 to 2.95 pips per 5-minute bar between 8 and 11 am New York and 0.96 pips in the early evening, per the best hours to trade EURUSD. Then the filter. Then, a distant third, the period, which should stay at 14. Divergence is a separate use of the same line and is covered in the MACD divergence guide; the timing lesson there, start on H1 and above, matches what the 30/70 numbers show on M5.

So what are the RSI settings?

14, on the default close price, read against the 50 line and only in the direction of an EMA 200, during the New York morning. That is the whole grid. The oscillator’s job is timing inside a trend, and a slow filter did every bit of the work the period was supposed to do. For the trend line itself I use my DeMARK trend line, which is drawn from price pivots and does not move once the candle closes, so RSI has something fixed to be measured against. If you want the four families of indicator laid out before choosing, forex indicators explained is the primer.

FAQ

What is the best RSI setting for scalping?

Fourteen. RSI 7, the usual scalping recommendation, returned +9.0 R against +34.0 R for RSI 14 on the same 5-minute data, with a drawdown more than twice as deep.

Is RSI 14 too slow for forex?

It is slow on purpose. The extra signals from a faster period were the whole reason the fast setting underperformed. The smoothness is the feature.

Should I use 80/20 instead of 70/30 on lower timeframes?

The 30/70 reversion trade lost to random on M5 and wider levels only make it rarer, not better. Use the 50 line and a trend gate instead; the numbers above did not favour any reversion level.

What data are these numbers from?

EURUSD M5, 16,926 bars, June 3 to August 26, 2026, exported from a live MetaTrader terminal. The 7 vs 14 test used identical entry and exit rules for both periods; the 30/70 test used a two hour hold and 2,000 random portfolios as the benchmark.

{“@context”: “https://schema.org”, “@type”: “FAQPage”, “mainEntity”: [{“@type”: “Question”, “name”: “What is the best RSI setting for scalping?”, “acceptedAnswer”: {“@type”: “Answer”, “text”: “Fourteen. RSI 7, the usual scalping recommendation, returned +9.0 R against +34.0 R for RSI 14 on the same 5-minute data, with a drawdown more than twice as deep.”}}, {“@type”: “Question”, “name”: “Is RSI 14 too slow for forex?”, “acceptedAnswer”: {“@type”: “Answer”, “text”: “It is slow on purpose. The extra signals from a faster period were the whole reason the fast setting underperformed. The smoothness is the feature.”}}, {“@type”: “Question”, “name”: “Should I use 80/20 instead of 70/30 on lower timeframes?”, “acceptedAnswer”: {“@type”: “Answer”, “text”: “The 30/70 reversion trade lost to random on M5 and wider levels only make it rarer, not better. Use the 50 line and a trend gate instead; the numbers above did not favour any reversion level.”}}, {“@type”: “Question”, “name”: “What data are these numbers from?”, “acceptedAnswer”: {“@type”: “Answer”, “text”: “EURUSD M5, 16,926 bars, June 3 to August 26, 2026, exported from a live MetaTrader terminal. The 7 vs 14 test used identical entry and exit rules for both periods; the 30/70 test used a two hour hold and 2,000 random portfolios as the benchmark.”}}]}