RSI Explained: What 30, 50 and 70 Really Mean

The Relative Strength Index is probably the first indicator most people switch on, and the first one they misuse. The textbook line is short: above 70 is overbought, below 30 is oversold. The trouble is that "overbought" sounds like "about to fall", and that is not what the number says. This guide works through what RSI actually measures, with real numbers from one of the most-watched charts of 2024.

What RSI measures

J. Welles Wilder introduced RSI in 1978. For each bar you record the close-to-close change. Up days go into a "gain" bucket, down days into a "loss" bucket. RSI compares the average gain with the average loss over the look-back window, which is 14 bars by default:

1. Average gain and average loss: the first value is a simple mean of the last 14 bars. After that, Wilder smooths it: new average = (previous average × 13 + today's value) ÷ 14.
2. RS = average gain ÷ average loss.
3. RSI = 100 − 100 ÷ (1 + RS).

If gains and losses have been the same size, RS is 1 and RSI is 50. If there have been no losses at all, RSI is 100. That is the whole idea: RSI is a ratio of recent up-move size to down-move size, squeezed onto a 0–100 scale. Nothing in the formula knows about valuation, earnings, or how far price has travelled from any "fair" level.

One day worked through

On June 17, 2024 NVDA closed at 130.98. Its smoothed average gain was 1.9612 and its average loss 0.5301, so RS was 3.700 and RSI 78.72. The next day it closed at 135.58, up 4.60:

Notice what moved the number. A single strong day lifted the average gain, and the absence of a down day let the average loss decay. RSI climbed because the last two weeks had been lopsided, not because the price crossed some ceiling.

NVDA in 2024: overbought for a month

NVDA reported earnings after the close on May 22, 2024 and gapped from 94.95 to an open of 102.03 the next morning. RSI closed at 71.6 that day. Anyone treating 70 as a sell signal would have been out at 103.80.

From May 23 to June 20 the stock closed above 70 on 18 of 19 sessions. The one exception was May 31, at 68.8. Over that stretch the close went from 103.80 to a peak of 135.58 on June 18, roughly 31% higher, while RSI said "overbought" almost every day. In a strong trend, a high RSI is a description of the trend, not a warning.

The turn did come, but RSI did not time it. From the June 18 close the stock fell to 118.11 by June 24, a 12.9% drop in four sessions. On June 20, the first down day, RSI was still 71.4. You only knew the run had paused because price failed: an outside day that opened at 139.80 and closed at 130.78.

Oversold is not a floor either

The same stock gives the mirror image. In early August 2024, during a broad market sell-off, NVDA fell to 98.91 on August 7, and RSI only reached 36.5. The textbook "below 30" buy signal never fired. Across 2023 to 2026 NVDA closed below 30 exactly once: April 4, 2025, at 94.31 with RSI 28.5. The next morning it opened at 87.46 and traded as low as 86.62, more than 8% below the "oversold" close, before bouncing to 114.33 on April 9.

So a reading of 28 told you the prior two weeks had been one-sided. It did not tell you the selling was finished. In that case, a buyer at the oversold close had to sit through another 8% drop first.

Three readings that are more useful than 30/70

The 50 line as a trend filter

In a healthy uptrend, pullbacks tend to hold RSI in the 40–50 area. In a downtrend, bounces tend to stall at 50–60. From July 17, 2024 onward NVDA's bounces kept failing near 50 (50.7 on July 22, 49.6 on July 23) and the stock kept falling. Watching where RSI stalls often tells you more than watching whether it reached an extreme.

Divergence, with a caveat

A bearish divergence is when price makes a similar or higher high but RSI makes a lower one. On July 10, 2024 NVDA closed at 134.91, within 0.5% of the June high, yet RSI was 65.8 against 81.4 in June. The advance was being carried by smaller up days. The stock then fell to 98.91 within a month. That is a textbook case, and it is also a warning against hindsight: plenty of divergences resolve by going sideways. Treat divergence as a reason to tighten your plan, not as an entry on its own.

Range extremes, but only in a range

The 30/70 rule works best where it was born: a market moving sideways between clear support and resistance. If the chart shows a flat box and RSI reaches 70 at the top of it, the two pieces of evidence agree. If the chart shows higher highs and higher lows, 70 is just the trend doing its job.

Settings: 14, 7 or 21?

Fourteen is Wilder's default and what most charting tools show, which matters because other traders are watching the same line. A 7-period RSI reaches the extremes far more often and is noisy on daily charts. A 21-period RSI rarely reaches them at all. Changing the period does not add information; it only changes how often the line crosses your thresholds. Pick one and learn how it behaves on the markets you watch.

A short routine for reading RSI

1. Look at price first. Is it trending, ranging, or breaking out? Mark the last swing high and low.
2. Note which side of 50 RSI has been on for the last month.
3. If RSI is at an extreme, ask whether price structure agrees: a rejection wick at resistance, a failed breakout, a lower high.
4. Check for divergence only at meaningful highs or lows, not on every wiggle.
5. Decide what would prove you wrong before you act, and write it down.

RSI pairs naturally with a trend tool. Moving averages tell you which way the trend runs, MACD shows whether it is speeding up, and RSI tells you how one-sided the recent closes were. Each answers a different question.

The fastest way to build a feel for this is to watch RSI while the next candle is still hidden. In the One Candle Ahead simulator, turn on RSI(14), guess the next bar, then reveal it. After a few dozen reps you start to see the difference between "overbought and tiring" and "overbought and just getting going".