Is RSI 30 a buy signal?

Our last two Long Trend pieces tested moving averages. Today, we will test one of the most famous technical indicators, the Relative Strength Index (RSI).

J. Welles Wilder published RSI in 1978. It compares the size of a stock's recent up days with its recent down days, usually over 14 days. The result runs from 0 to 100. Below 30, the stock is called “oversold”. Above 70, it is called “overbought”.

The common wisdom among the market technicians is simple.

I have leaned on that wisdom myself, more than once. Ok, maybe even more than a couple of times.

Here is SPY (the S&P 500 ETF) over the last two years, with its RSI in the lower pane.

Take the last signal. On 27 March 2026, SPY's RSI closed below 30. Bought at the next open and held five days, SPY gained 2.6%.

How tempting is it to conclude now that this is the gold-dust indicator?

One good trade proves nothing, though. Was the signal useful, or did I just pick the winner after the race? So I tested it on all 193 names I cover.

1. A drop below 30 paid. Across 4,840 signals, buying at the next open beat an ordinary five days by 0.86%. It beat SPY by 0.67%.

2. The edge came in the first week. Holding longer added nothing on top.

3. A bigger fall paid more. A drop below 25, or RSI(21) below 30, earned over 1%. 30 is just a line.

4. Waiting for the climb back above 30 gave the edge away.

5. On its own, it is too rare. One stock at a time, the money sat in cash about 95% of the time.

6. As a full-time strategy, it is too crowded. The signals bunch in sell-offs.

7. As a swing trade on the side of SPY, it added about 2% a year after costs.

A signal is the first close with RSI (14) below 30, after a close at 30 or above. RSI needs the closing price. The signal is only known after the close. I buy at the next day's open and sell at the open five trading days later.

I count at most one signal every 10 trading days. A stock hovering around 30 does not count again and again. On the SPY chart, the hollow circle marks the drop I didn’t count.

For each signal, I take the stock's return from that open to the open five days later. Then I subtract the same stock's return over an ordinary five-day period, measured the same way. What is left is the excess return vs an average day: what the signal adds.

Since 2011, SPY has given 22 signals. Its average excess return was 1.43%. But 22 signals are far too few to trust on their own.

Therefore, I tested the entire universe of tickers that I cover. Each result below is an average across the 193 names. I average each stock's signals first, then average the 193 stocks. A stock that sells off often does not count more. The 95% range is in brackets. A range that stays above zero is a result I trust.

Yes. Across 193 names and 4,840 signals, buying at the next open and holding five days beat an ordinary five days by 0.86% (0.61% to 1.12%).

The trades also beat the market. Over the same five days, the stock beat SPY by 0.67% per trade (0.46% to 0.91%).

A stock with RSI below 30 has just fallen hard. Maybe any stock bounces after a fall like that, RSI or not. I also compared each signal with other days on which the same stock had moved by a similar percentage over the previous 20 trading days. The signal still added 0.54% (0.30% to 0.79%).

I also held each signal for ten and 20 days, each against an ordinary stretch of the same length.

Held ten days, the lead was 0.43% (0.06% to 0.75%). Held 20 days, it was 0.64% (0.12% to 1.09%). Both are below the five-day figure. The extra return arrives in the first week. Holding longer adds nothing on top.

Wilder chose 30. I tested the levels around it.

The lower the RSI, the bigger the bounce, down to about 25. A drop below 35 RSI earned 0.32%. Below 30, it earned 0.86%. Below 25, it earned 1.06% (0.58% to 1.58%). Below 20, the signals get rare. The range runs through zero.

So the 30 threshold isn't magical.

Wilder used 14 days to calculate the RSI. I also tested a faster RSI(7) and a slower RSI(21).

RSI(7) earned just 0.19%. RSI(21) earned 1.19% (0.62% to 1.85%), the most per signal. It fires about a third as often as RSI(14).

RSI(21) moves more slowly. It only reaches 30 after a deeper fall. RSI(14) below 25 fires about as often. RSI(21) below 30 beat RSI(14) below 25 by just 0.12% per signal (minus 0.28% to plus 0.52%). A slower RSI and a lower line do much the same job. Both wait for a bigger fall.

One version of the rule waits for confirmation. Let RSI drop below 30, then buy when it climbs back above 30.

Waiting gives the edge away. Buying the climb back above 30 earned minus 0.09% over five days and minus 0.32% over ten. Both ranges run through zero. Buying the drop itself earned 0.86%.

0.86% in five days sounds like a lot. Earned every week, it would compound to about 54% a year. I built the simplest version on each stock. Buy at the open after every signal, sell five days later.

While it holds the stock, the rule does well. Over just those days, its Sharpe ratio, which measures return per unit of risk, was 1.27. Owning the stock all the time scores 0.60.

The catch is time. RSI(14) sits below 30 on about 3% of trading days. A stock gives about 1.9 signals a year. The rule holds it on just 4.5% of days. Across the full year, it made 2.2% a year, against 17.1% for owning the stock.

The timing itself is real. Buying the same stock on random days, the same number of times, scored a Sharpe ratio of 0.11, against 0.26 for the RSI rule.

The textbook pair does not fix it. Buying below 30 and selling above 70 holds each trade for about five months. That made 7.1% a year. It beat owning the stock on just 18% of names.

No. I put all the money, split equally, into every stock with a fresh signal. With no signal, the money sat in cash. Before costs, this made 17.2% a year. After costs, it made 10.1%. Owning the same 188 stocks equally made 22.3%. Its worst fall was 52%.

The signals arrive in crowds during sell-offs. On 8 April 2025, 116 of the 188 stocks were held at once. Buying on random days instead did better in 19 of 20 tries.

Picture a $100,000 portfolio held in SPY. When a stock's RSI closes below 30, $2,000 moves out of SPY into that stock at the next open. Five trading days later, at the open, the stock is sold. The $2,000 goes back into SPY.

Most weeks, only a few trades are open. In a sell-off, many stocks signal at once. If ten signal in the same week, $20,000 is out on swing trades. The cap is half the portfolio.

This version works. Before costs, it made 18.3% a year, against 15.2% for SPY alone. The Sharpe ratio rose from 0.94 to 1.06. The worst fall was about the same, 33% against 34%.

The gain over SPY was 2.72% a year (1.18% to 4.32%). After costs, it was 2.14% a year (0.63% to 3.75%).

Part of that gain comes from holding stocks instead of SPY. The same swing trades on random days added 0.67% a year. The RSI version beat the random version in all 20 tries.

The answer held at other sizes. At $1,000 a trade, the gain was 1.63% a year. At $5,000, it was 4.15%. With the 188 stocks held equally as the base, instead of SPY, the $2,000 version added 1.60% a year (0.42% to 2.86%).

Here RSI(14) beat RSI(21). The slower setting fires a third as often. It adds fewer trades. Its swing trades added 0.94% a year. RSI(14) added 1.78% a year more (0.51% to 3.06%).

RSI 30 marks a better week to own a stock, about twice a year per stock. It works as a swing trade on the side, not as a strategy on its own.

The daily Systematic Portfolio and Multi-model Signals, including the deep dives on every holding, sit alongside this letter. You can see how they work at yxinsights.com/systematic.

DISCLAIMER: This newsletter is strictly educational. Any information or analysis in this note is not an offer to sell or the solicitation of an offer to buy any securities. Nothing in this note is intended to be investment advice and nor should it be relied upon to make investment decisions. Any opinions, analyses, or probabilities expressed in this note are those of the author as of the note's date of publication and are subject to change without notice.