How to Read RSI: 70/30 Levels, Settings and Divergence

Read RSI 70/30/50, compare 7/14/21-bar settings, and test trend, pullback and divergence rules with confirmed bars and separate intervals.

Published May 11, 2026 · Content updated October 1, 2026

When RSI reaches 70, first ask whether the trend is still continuing.

RSI, the Relative Strength Index, is a momentum oscillator from 0 to 100 that compares smoothed gains and losses. TradingView defaults to a length of 14 bars and the closing price. On an hourly chart, 14 means fourteen hourly bars, not fourteen days. TradingView RSI guide

Above 70 is commonly called overbought and below 30 oversold. Turning those labels directly into sell and buy orders can miss a continuing trend. In a strong advance, RSI can stay elevated or briefly dip and rise again. Any claim about how long this happened in a particular market needs actual data.

Observe a 40–50 pullback within a trend separately from a 70/30 reaction in a range. The thresholds and setup below are research examples, not trading rules with demonstrated performance.

Conceptual illustration. Its prices and paths are not a verified trading record for a specific asset.

RSI 50 means smoothed gains and losses are equal; it does not determine the presence of a trend on its own. The calculation is RSI = 100 − 100 / (1 + RS), where RS = RMA of gains / RMA of losses. RMA is Wilder's smoothing. A separate moving average added to RSI is distinct from RSI's own calculation. Match inputs using TradingView's RSI documentation.

Length 14 means 14 calculation bars, not necessarily 14 days

Search for the built-in RSI, add it to the TradingView chart, and inspect its inputs. RSI Length counts calculation bars; Source selects the value from each bar. The documented defaults are 14 bars and Close. On a 1-hour chart using the chart's calculation timeframe, it uses 14 hourly bars. Setting the indicator timeframe to daily uses daily data even on an hourly chart. RMA carries its previous smoothed value, so its history does not stop exactly at the latest 14 bars.

Record both chart and indicator timeframes. When Timeframe differs, also check Wait for timeframe closes. Values and crosses can change before a bar closes. See the official timeframe guide.

This compares calculation structure; it is not a return experiment on a market interval. Select trend, range and sharp-change intervals together, record signal frequency and failures, and evaluate unchanged rules on a separate interval. Improved returns after changing a length do not establish an optimal value.

The same synthetic input changes differently with each length

These results use unitless generated values treated as closes, not a market, date, instrument, exchange or return. The 400 warm-up values are 100 + sin(0.2 × i) for i=0–399. The 120 rising values are 100 + 0.2 × j + 1.5 × sin(0.35 × j); the 120 oscillating values are 124 + 3 × sin(0.65 × j), each with j=0–119. Angles are in radians. These two shapes were deliberately constructed to compare the lengths on identical inputs.

RSI starts with mean gains and losses over the first n changes and continues using RMA. After warm-up, the rising window uses overall indices 400–519 and the oscillating window 520–639, without resetting between them. For each 120-value window, we average |current RSI − previous RSI| over its 119 adjacent pairs. Units are RSI points, rounded to two decimal places in the table.

For this input, shorter lengths had larger mean changes in both windows. Early oscillating values still inherit smoothing from the rising window. We do not assume initial-state influence is exactly zero. These ranges establish neither real-market distributions nor signal success rates. This compares calculation response, not a trading-performance ranking of 7, 14 and 21.

Check price direction before interpreting 70

First examine whether price is making higher highs and lows or repeatedly crossing the same range. RSI 70 indicates relatively strong recent upward changes; it does not establish an imminent fall or a confirmed trend. A reading of 70 near a range high and one after an upward breakout can be studied as different hypotheses.

The 200 EMA and ADX can provide additional criteria. For example, group bars whose close is above the 200 EMA and whose ADX is at least 20. ADX does not identify direction, and 20 is not a universal boundary between a trend and a range. Check how a filter changes trade count and results after costs.

Comparison of price environments. A reading of 70 alone cannot confirm a reversal in either one.

Confirm a pullback's recovery above 50 at the bar close

An RSI decline into 40–50 followed by a recovery above 50 is a candidate for studying trend pullbacks. Price can fall again immediately afterward, so define price context and invalidation together.

RSI does not always confirm before volume. Both RSI and volume can update during an open bar. An intrabar RSI cross can disappear before the close; evaluate a closing-price rule using completed bars. TradingView real-time bar guide

Research example — four-hour long setup, performance untestedUse RSI(14, close), EMA(200, close), and ADX with directional length 14 and smoothing 14.The signal bar must close above the EMA with ADX at least 20; at least one of the preceding five bars must have RSI between 40 and 50 inclusive.If the previous RSI is at most 50 and the signal bar's RSI is above 50, model entry at the next bar's open.Set the stop one tick below the lowest low of the last five bars including the signal bar. Cancel entry if the next open is at or below that stop.After entry, exit at the next open if RSI closes below 50. A fixed stop hit earlier takes priority.Hold only one position and evaluate on a separate period with fees, slippage, and gap fills included.

This tests a different hypothesis from waiting for RSI 30. Do not assume a 40–50 rebound always occurs or performs better; compare both using the same data and costs. A short rule can mirror a rebound into 50–60 followed by a cross below 50, but it needs its own evaluation.

Setup illustration. Waiting for 30 does not necessarily mean missing an opportunity.

Divergence does not confirm a reversal

Bearish divergence occurs when a higher price high corresponds to a lower RSI reading at the two compared highs. Bullish divergence pairs a lower price low with a higher RSI low. This disagreement is a reason to examine the relationship between price and momentum again.

Price can continue in the same direction after divergence. You can add resistance proximity or a close below RSI 50 as conditions, but neither combination guarantees reliability. Abandon the hypothesis when price breaches its predefined invalidation level.

If identifying a high or low requires bars on both sides, the right-hand bars must finish before the pattern is known. Do not treat a signal drawn at a historical peak as information available at that peak. TradingView guide to signals plotted in the past

Conceptual bearish divergence. It is not evidence of a subsequent decline or a profitable trade.

A failure swing needs all four stages

A failure swing uses the RSI path itself, whereas divergence compares RSI with price. A bullish failure swing requires a drop below 30, a recovery above 30, a second decline that holds above 30, and a break above the previous RSI rebound high. The second low alone does not complete it.

The bearish sequence is a rise above 70, a fall below 70, another rise that stays below 70, and a break below the previous RSI reaction low. More conditions do not establish a higher win rate than divergence. TradingView failure swing explanation

In a bullish failure swing, the first low is below 30 and the second is above 30. Wait for the break of the intervening RSI high.

Validate thresholds for each asset and timeframe

RSI distributions can differ by instrument, timeframe, and observation period. Higher price volatility does not necessarily mean more readings above 80, and daily bars do not necessarily have fewer extremes than one-minute bars. The relative composition of upward and downward changes matters.

Suppose you observe five declines after RSI 78. That creates a hypothesis to investigate, not an established overbought threshold. Define the return and number of bars that count as a decline before examining outcomes. Record every eligible case and overlapping signal. Select the threshold on one period and assess it on another, comparing costs, trade count, and maximum drawdown.

Conceptual distribution comparison. The image’s “historical extreme” label has not been verified against actual statistics. The 78/22 levels and BTC/SPY distributions are illustrative, not validated trading thresholds.

Adding conditions does not fix an unsupported hypothesis

• Selling at 70 in a trend: Overbought does not guarantee a selling opportunity. Record continuations as well as declines after 70.
• Overlapping indicators: RSI, Stochastic, and CCI differ in calculation but can share price information. Adding ADX, OBV, or Bollinger Bands does not automatically create independent confirmation. Compare results before and after each addition.
• Shortening the length: RSI(7) can respond more strongly to recent changes. Fourteen is a common default, not the optimal noise filter for every market. Fix the definition of a failed signal and compare lengths across multiple market periods.

Align the confirmation time of the higher timeframe

If you add a four-hour trend filter to an hourly signal, check that only a completed four-hour bar was used at that moment. Treating an unfinished higher-timeframe value as a historical confirmed value can produce signals that cannot be reproduced.

Define support and resistance from information available at the time as well. An RSI recovery above 50 near a level does not establish success. Compare the presence and absence of a filter while holding entry, exit, and cost rules constant; that provides a basis for deciding what to wait for on the next chart.

Illustration of timeframe alignment. Agreement with the higher trend still leaves the possibility of loss.

To connect settings with trade records, inspect the trade-analysis example and then use TradingView trade CSV analysis with your strategy trade list. Compare indicator changes with the same instrument, interval, execution and costs; distinguish calculation diagrams from strategy performance. Continue with the technical-indicator series.

For applying a strategy and matching costs and dates, follow the measured example in the TradingView backtest guide.