How to Read RSI: 70, 30 and What Data Says About Overbought

What RSI measures: the share of gains

The RSI (Relative Strength Index) compares how much the close rose and how much it fell over recent bars and expresses the result as a number from 0 to 100. The formula is RSI = 100 − 100/(1+RS), where RS is the average gain divided by the average loss. Rearranged, this becomes RSI = 100 × average gain ÷ (average gain + average loss). In other words, an RSI of 70 means that 70% of the average recent movement was to the upside. It uses only closes, not highs or lows. Seen this way, RSI measures not how high price is but the directional makeup of recent moves. Even at an all-time high, RSI can sit near 50 if recent bars have alternated up and down, and even after a steep fall it can climb fairly high if price has risen for several days in a row. If every bar rises, the average loss approaches 0 and RSI pins near 100; if every bar falls, it pins near 0. That is why looking at which zone RSI sits in and for how long, rather than at a single number, keeps you from missing the actual move.

The calculation, step by step, with Wilder's smoothing

J. Welles Wilder introduced the RSI in a 1978 book, and the averages are taken the way he did it (Wilder's smoothing, also called RMA). TradingView's default RSI(14) and this site's RSI tools use the same calculation. Only the first average is a simple average; after that, the previous average gets a weight of 13 and the new value a weight of 1. This is an exponential average with α = 1/14, the same weighting as an EMA with a period of 27. So RSI(14) keeps a faint memory of a longer past than its name suggests, and its value differs slightly depending on where the calculation started. Where this property actually causes trouble is when two tools show different RSI values. With few bars loaded, the influence of the first average lingers and values can differ by a few points; as more bars accumulate, they converge. There are also RSI versions that replace Wilder's smoothing with a simple average or an ordinary EMA, so the same 14 can give different values. Before comparing numbers, matching the calculation method comes first.

• Gain = how much the close rose from the previous bar (0 if it fell)
• Loss = how much the close fell from the previous bar (0 if it rose)
• First average = the simple average of the first 14 values
• Each later average = (previous average × 13 + current value) ÷ 14

70, 30 and the 50 line

Wilder treated readings above 70 as overbought and below 30 as oversold, and most charts still draw these two lines by default. 50 is the point where the average gain equals the average loss, so an RSI above 50 means recent gains were larger than recent losses. That is why some people use the 50 line to gauge the direction of the move, and in markets with large swings some use 80 and 20 as the thresholds instead. Overbought and oversold are just names attached to reference lines; they say nothing about the direction ahead. These lines are widely used less because they are proven boundaries than because they are drawn by default on almost every chart, so many people watch the same numbers. Many people therefore pay attention to the moment RSI crosses 70, but what happens afterward varies by market and period. When changing the thresholds, too, keep in mind that numbers tuned to past results tend to fit only that past.

In a strong trend, RSI stays above 70 for a long time

In an uptrend that keeps pausing briefly and then climbing again, the average gain stays larger than the average loss. The RSI then stays above 70 for long stretches, and even on pullbacks it tends to turn back up from the 60s. In the figure below, price rose about 31% further after the RSI first went above 70. Anyone who read that move above 70 as a sign of an imminent fall would have been out of step with most of the rise that followed. Being above 70 is a summary saying the upward push has been strong so far, not a verdict that it is exhausted. This happens because pullbacks in an uptrend are short and shallow, so the average loss rarely grows. As long as buy orders keep arriving on every dip, RSI stays high. That is why, within a strong trend, some people treat a deep drop of RSI into the 40s or 50s during a pullback, rather than the move above 70, as a clue to a change in the move.

In a range, RSI moves between 30 and 70

When price moves up and down within a set range, the RSI also travels between 30 and 70, rising near the top of the range and falling near the bottom. In stretches like this, reading the area around 70 as a top and the area around 30 as a bottom appears to work well. The problem is that the RSI does not tell you when the range will break. If price breaks out of the range to the upside, the RSI can stay above 70 for a long time, as in the previous figure, and the reading that worked inside the range goes most badly wrong (see the article on sideways markets). Whether price is in a range should be checked on price first, not on RSI. Clues include how many times highs and lows stalled at similar levels and whether volume shrank in between. Also worth remembering: RSI is a ratio, not a size, so a range 1% wide and a range 10% wide can produce similar-looking RSI.

Measured: what actually happened after overbought readings

Measured on the daily bars of 10 coins on Binance (each from its Binance listing date to September 2026, 29,946 bars in all), the share of bars with RSI(14) above 70 whose close was higher 20 bars later was 58.0%, above the 50.4% baseline measured over all bars. The median, at +3.52%, was also larger than the baseline's +0.15%, and after bars below 30 the share was 56.3%, also above the baseline. For BTC daily bars alone, the share after the 369 bars above 70 was 71.8% (baseline 53.7%). In this period, overbought readings marked stretches where a strong move carried on rather than foreshadowing a decline. The baseline is placed alongside because coins generally rose in this period, so more than half of all bars were followed by higher prices no matter which bar you picked. That is why 58.0% means something only when compared with 50.4%, not with 50%. That both the bars below 30 and the bars above 70 came out above the baseline can also be read as saying that RSI being at one extreme was not enough, on its own, to tell the direction that followed.

• Baseline (all bars, 20 bars later): up 50.4%, median +0.15%
• 2,294 bars with RSI above 70: 58.0%, +3.52%
• 577 bars where RSI dropped from above 70 to below it: 56.3%, +2.75%
• 1,087 bars with RSI below 30: 56.3%, +2.24%

Bar length, divergence and tools

The 14 in RSI(14) is a number of bars, not days. On 1-minute bars it covers 14 minutes and on daily bars two weeks, so the same reading of 70 contains completely different spans of time (see the article on timeframes). RSI Radar shows the 1-minute RSI(14) of 20 coins on Binance, so its values can swing between 30 and 70 within minutes. When price and RSI disagree, for example when price makes higher highs while the RSI makes lower highs, it is called a divergence; this is covered separately in the divergence article in Part 6. The RSI Divergence Finder detects these and draws them on the chart as lines. At the same moment, it is common for 1-minute RSI to be overbought while daily RSI sits near 50. Rather than arguing about which one is right, it causes less confusion to decide first which bar length matches the period you care about. The shorter the bar, the more a single burst of trades can swing RSI.

What these numbers do not say

The measurements above are a record of one period, not a rule. Bars with RSI above 70 come in runs during uptrends, so the 2,294 are not independent cases, and because coins rose sharply over this period, even the average 20-bar return across all bars (+4.62%) is positive. Changing the bar length and the horizon also changes the gap. On BTC 4-hour bars, the share higher after bars above 70 was 53.1% 6 bars (one day) later, almost the same as the 52.0% baseline, and 58.2% (baseline 52.8%) 30 bars (5 days) later. All that can be said here is that, in this data, 'overbought means a fall is coming' was not confirmed. How to use the moments RSI crosses 70 or 30 in trading is not something this article can decide. For the same reason, a day when a tool showing many coins at once flags several of them as overbought does not amount to as many independent pieces of evidence as there are coins.

What looks different in crypto and stocks

RSI is calculated from differences between closes, so whether a market takes breaks feeds directly into its value. Korean stocks stop trading after the regular session until the next open, and news in between is reflected all at once in the next day's first move. When Samsung Electronics or SK hynix gaps sharply higher, that whole day goes in as a single gain, RSI jumps at once, and because of Wilder's smoothing the trace lingers for a while. The same happens with US indexes or large tech stocks when a gap follows an earnings report. Crypto trades around the clock, so a change of the same size tends to be split across several bars. Korean stocks have a daily price limit of ±30%, which caps the gain in a single bar, while crypto has none. Volatility differs too. Coins tend to move more in a day, so RSI crosses 70 and 30 more often, which is why some people use 80 and 20 instead. Volume is not part of the RSI calculation, so in both markets a small move during a nearly empty trading hour moves RSI with the same weight.

A practical checklist

On a live chart, RSI is not final until the last bar closes. The current price takes the place of the close in the bar still in progress, so it is common for an RSI that went above 70 mid-bar to end below 70 when the bar closes. In a tool that watches 1-minute bars, like the RSI Radar, this kind of change happens within seconds. All the measurements above were taken on closed-bar values, so they should not be compared directly with the value of a bar still in progress. Below is the order of things to check before reading an RSI number.

• Bar length: how many minutes, hours or days the 14 covers
• Calculation: whether it is RSI(14) with Wilder's smoothing
• Whether the last bar is closed or still in progress
• Whether price is trending or in a range (price before RSI)
• How long RSI has stayed beyond 70 or 30, and which side of the 50 line it is on
• Whether a gap or a single sharp bar pushed the value up