Does RSI actually work for trading?
It depends entirely on the setting, and the difference is dramatic. We walk-forward tested 6 RSI strategies across 28 tickers and ran each through a 100-run random-entry gauntlet. The fast RSI(2) dip buys held a real edge on up to 10 of 28 tickers. The two most preached RSI techniques, buying when RSI(14) drops below 30 and trading bullish divergence, both busted. Same indicator, opposite outcomes.
Based on the same process behind every Library report. How we test.
The three ways traders use RSI
RSI shows up in three roles. As a dip trigger: buy when a fast RSI like RSI(2) collapses, betting on a snap back. As an oversold alarm: buy when the classic RSI(14) drops below 30, the version every textbook teaches. And as a trend gauge: treat the 50 line as the border between bullish and bearish momentum. We tested all three, always with price above its 200-day average so every entry happens inside an uptrend.
The RSI(2) dip buys that held an edge
| Setup | Ticker | Return | Sharpe | Win rate | Edge held on |
|---|---|---|---|---|---|
| RSI(2) below 5 dip buy | QQQ | +35.4% | 0.56 | 82.1% | 10 / 28 |
| RSI(2) dip, MA exit | SPY | +44.4% | 0.76 | 74.7% | 9 / 28 |
| RSI(2) dip buy | QQQ | +74.0% | 0.81 | 77.0% | 7 / 28 |
The classic Larry Connors family: a 2-period RSI collapses within an uptrend and you buy the snap-back. High win rates with modest per-trade returns, and real breadth: the strictest version held an edge on 10 of 28 tickers. Return shown is a representative index, not the best single ticker.
RSI as a trend gauge: a narrow edge
| Setup | Ticker | Return | Sharpe | Win rate | Edge held on |
|---|---|---|---|---|---|
| RSI(14) 50-line cross | GOOGL | +165.5% | 0.80 | 37.5% | 4 / 28 |
Crossing the 50 line is a momentum signal, not a dip buy, and it behaves like one: fewer wins, bigger winners, and an edge on only 4 of 28 tickers. It works where trends run long; it is not a broad, everywhere edge.
The textbook RSI signals that failed
Here is the uncomfortable part. The two RSI techniques traders actually learn first both busted. Buying when RSI(14) turns up through 30 beat 98 of 100 random entries on its best ticker, which sounds unbeatable, until you count the trades: three in a decade, positive in only two of four test periods. Bullish RSI divergence was the same story, three trades in ten years on its best name. The reason is the period setting. With a 14-day lookback and a 200-day uptrend filter, RSI almost never reaches 30, because stocks in uptrends rarely get that oversold. RSI(2) reaches its trigger constantly. One setting produces a sample you can trust; the other produces a coin flip with a good story.
The pattern: the period setting decides everything
Every RSI result in this guide is explained by one variable: how often the signal fires. RSI(2) collapses below 10 dozens of times a decade, so its edge is measured on a real sample. RSI(14) below 30 fires a handful of times, so its great-looking score is statistical noise. This is not an RSI quirk, it is the same lesson our moving-average crossover test taught: before you trust any indicator backtest, check the trade count, then the win rate.
How to use this
- If you use RSI to buy dips, use a fast setting like RSI(2) below 10, not the textbook RSI(14) below 30.
- Keep the trend filter. Every surviving setup required price above its 200-day average; an oversold reading in a downtrend was outside our test.
- Expect the dip-buyer profile from RSI(2): high win rate, small average win, regular signals.
- Treat RSI divergence claims with suspicion. It fired 3 times in a decade on its best ticker; nobody has a trustworthy sample of it.
- Count the trades before you trust any RSI backtest. Under about 30 trades, the score is a hint, not evidence.
Common questions
Does RSI actually work for trading?
The indicator works when the setting gives it enough signals. Fast RSI(2) dip buys held a walk-forward-tested edge on up to 10 of 28 tickers, while the classic RSI(14) oversold signal and bullish divergence both failed our test.
What is the best RSI setting for swing trading?
In our test, RSI(2) was the clear winner for dip buying in an uptrend, with the strictest version (buy below 5) holding an edge on 10 of 28 tickers. The textbook 14-period setting fired too rarely to trust.
Does buying when RSI is below 30 work?
Not with the standard RSI(14) in an uptrend. It triggered only three times in a decade on its best ticker, far too few trades to prove an edge, and it cleared our bar on zero of 28 tickers.
Does RSI divergence work?
It busted in our test. Confirmed bullish divergences are so rare, about three trades in a decade on its best ticker, that no one can honestly measure an edge from them.
Is RSI(2) better than RSI(14)?
For dip buying, yes, and by a wide margin in our test. RSI(2) fires often enough to build a trustworthy sample and held an edge across multiple tickers; RSI(14) oversold almost never fires inside an uptrend.
We take each setup's rules as commonly stated. Your parameters may differ, so run your own on your own account and watch the walk-forward result before you trust it.