What Is RSI (Relative Strength Index)?
Michael Whitman
·
Trading since 2017, full-time since 2020
·
Published 2026-09-05
·
Updated
·
Reviewed
RSI, the relative strength index, compares the size of recent up moves to recent down moves and reports the result on a scale of 0 to 100. Readings above 70 are called overbought and below 30 oversold, yet on SPY since 1994 what followed either crossing was within chance of an ordinary day.
Covered on this page: TradingView.
Relative strength index (RSI) is often misread, and the misreading is specific: people treat 70 as a sell signal and 30 as a buy signal. This page computes RSI exactly as its inventor defined it on more than thirty years of SPY and shows what actually followed each line.
How it works
RSI compares the size of recent up moves to the size of recent down moves, then expresses the result on a scale of 0 to 100.
J. Welles Wilder Jr. introduced it in his 1978 book New Concepts in Technical Trading Systems, with a 14-period default. The calculation, step by step:
- For each bar, take the change in the close. A rise is a gain (the loss is zero); a fall is a loss, written as a positive number (the gain is zero).
- Seed the averages. The first average gain is the simple mean of the first 14 gains, and the same for losses.
- Smooth from then on. Each new average = (previous average × 13 + this bar’s value) ÷ 14. This is Wilder’s smoothing, not a plain 14-bar average, so every past bar keeps a shrinking share of weight.
- RS = average gain ÷ average loss, and RSI = 100 − 100 ÷ (1 + RS).
So a high reading says price has been going up. It is a summary of what already happened, computed from the same bars you are looking at. Nothing in it knows about levels, trend, volume or time of day.
The 70 problem
Above 70 is conventionally called overbought, which sounds like a warning and is not one. A strong advance is made of up bars, and RSI measures up bars, so a high reading is the indicator doing exactly what it was built to do.
On SPY, a cross above 70 did not mark a top. Twenty days after the 87 spaced crossings since 1994, SPY was higher 64.37% of the time, almost exactly its everyday rate. Selling every first overbought reading would have meant selling into a market that went on to behave like any other day.
What the data does not support is the opposite story either. Stays above 70 on SPY were short: a median of 2 days, and the two longest ran 18 sessions each, from 6 November to 2 December 1996 and from 3 to 29 January 2018. “It can stay overbought for weeks” is true of a strong stock now and then, but not the usual case on a broad index.
Why the idea survives anyway
In a sideways market the extremes can genuinely mark the turns, because there is a ceiling and a floor for price to bounce between. The indicator has not changed. The market has.
This is why RSI feels reliable for a while and then stops without warning: it was being used in a range, and the market started trending, or the reverse.
The 50 line
The middle line reads differently from the other two. Above 50, average gains have been bigger than average losses. Below 50, the reverse.
Used that way RSI becomes a filter rather than a signal, a quiet answer to “which direction has recently been winning”, and it has no threshold to argue about.
Length and divergence
A shorter length reacts faster and reaches 70 and 30 far more often. That is not extra information; it is the same information with less smoothing, and more readings at the extreme means more readings that lead nowhere.
Divergence is price making a higher high while RSI makes a lower one. It means the second push was made of smaller up bars than the first. That is a true observation about the last two pushes and the most interesting thing the indicator shows, but a move can slow, hold and resume, so it describes rather than forecasts.
A worked example
SPY on 2 October 2026, by hand. On 1 October SPY closed at $763.99 with an RSI(14) of 49.19: the smoothed average gain stood at about $1.816 and the average loss at about $1.876.
On 2 October it closed at $769.64, a gain of $5.65 and a loss of zero.
- New average gain = ($1.816 × 13 + $5.65) ÷ 14 ≈ $2.090
- New average loss = ($1.876 × 13 + $0) ÷ 14 ≈ $1.742
- RS = 2.090 ÷ 1.742 ≈ 1.20
- RSI = 100 − 100 ÷ (1 + 1.20) ≈ 54.5; unrounded, the script gives 54.54
One good day moved RSI from just under 50 to about 54.5, a shift from “losses slightly bigger” to “gains slightly bigger”. Nothing about that says where to act. The entry comes from the chart, not the panel: RSI said which direction was winning, the price level said where to act, and market structure said the price at which the idea was finished.
The original data
RSI(14) on SPY, computed with Wilder’s formula above on daily closes from Yahoo Finance (29 January 1993 to 2 October 2026, downloaded 3 October 2026; prices not adjusted for dividends, as a chart shows them). The first reading is on 19 February 1993; statistics start on 3 January 1994 so the seed has worn off, giving 8,243 trading days. The daily RSI file is published with every value.
How often the lines were reached. RSI closed above 70 on 587 days (7.12%) and below 30 on 125 (1.52%). The highest reading was 87.03 on 26 January 2018 and the lowest 16.70 on 8 August 2011. In the year from 1 October 2025 there were 22 days above 70 and 3 below 30, with a high of 78.54 on 14 May 2026 and a low of 27.30 on 30 March 2026.
What followed a crossing, against the base rate. The base rate is every day with a close 20 sessions later: 5,257 of 8,223 were higher, 63.93%. A crossing is the first close beyond the line after a close on the other side. Crossings bunch together, so the test counts a crossing only if it comes at least 20 sessions after the last one counted; that keeps the 20-day windows from overlapping. The p values use the normal approximation to the binomial.
| Event | Count | Higher 20 days later | Median change | p vs 63.93% |
|---|---|---|---|---|
| Cross above 70, spaced | 87 | 64.37% | +1.32% | 0.932 |
| Cross below 30, spaced | 40 | 67.50% | +2.31% | 0.638 |
| All days (base) | 8,223 | 63.93% | +1.31% |
Both results are within chance. Neither crossing separates from an ordinary day on these counts, and 40 oversold crossings in 32 years is a small sample. Counting every crossing (179 above 70, 67 below 30) gives 68.72% and 73.13%, still within chance (p = 0.183 and 0.117). The crossing list has every date and result.
How long the extremes lasted. There were 179 separate stretches of closes above 70. Their median length was 2 days, 73 of them (40.78%) lasted a single day, and 45 ran five days or more.
In the search study behind this site, 594 of 24,971 videos have RSI or relative strength index in the title, with a median of 5,377 views; 77% never pass 50,000.
When it fails
Reading 70 as a sell
On SPY the overbought line was followed by an ordinary 20 days, not a drop. A strong market is made of up bars, and a rule that sells the first high reading sells into exactly the bars RSI is counting.
Reading 30 as a floor
The oversold side looked better and was still within chance. Forty spaced crossings is too few to separate a real tendency from luck. In June and July 2008, RSI crossed below 30 four times in 11 sessions while SPY slid from $128.23 to $123.84.
The divergence does not resolve
Momentum slowing is not momentum stopping. A move can slow, hold and resume, and the same divergence appears on a chart that keeps going.
You are trading the panel instead of the chart
The indicator has no idea where price is. Two identical readings at a major level and in the middle of nowhere mean different things, and the panel cannot tell them apart.
Related
Moving averages have the same character: a smoothed summary of recent price that describes rather than predicts. The lag is easier to see there.
Market structure is what RSI cannot give you: a specific price at which the read is wrong.
And risk management is the difference between an indicator reading and a position, which is a gap no oscillator closes.
Treat this the same way as a moving average, as context rather than a trigger. A high reading tells you the recent bars have been mostly up, which you can see by looking at them. What it never tells you is where to get in or where you are wrong, and those are the only two decisions that cost money.
— Michael Whitman
This page is educational, not financial advice. Test every idea on your own charts before risking money. Some links on this page earn a commission if you buy through them. It costs you nothing and it does not decide what appears here or in what order — how these pages are made is set out in our methodology.