RSI explained: overbought, oversold, and does the 30/70 signal work?
9 min read · Updated 2026-09-29
The Relative Strength Index (RSI) compares the size of recent up-moves with recent down-moves and turns the result into a number between 0 and 100. Above 70 is conventionally called overbought, below 30 oversold. The standard length is 14 candles, as J. Welles Wilder defined it in 1978.
The classic RSI signal is the reversal: RSI climbing back above 30 after being oversold (buy) or falling back below 70 after being overbought (sell). We measured it: across our 41 markets, hourly RSI 30/70 reversals worked 51.0% of the time. The chart, scoreboards and streak comparison are below.
See it on a real chart: Gold (XAUUSD) liquidation heatmap — the modeled clusters above and below the current price.
How RSI is calculated
- For each candle, take the change in close. Up-moves go into one bucket, down-moves (as positive numbers) into another.
- Average each bucket over 14 candles using Wilder's smoothing: each new value is 13/14 of the old average plus 1/14 of the new move.
- Relative strength RS = average gain / average loss. RSI = 100 − 100 / (1 + RS).
If there were only gains, RSI is 100; only losses, 0; equal average gains and losses, 50. It measures the balance of recent movement, not the price level.
What overbought and oversold actually mean
Overbought does not mean price is too high or must fall. It means recent candles have been mostly up. In a strong trend RSI can stay above 70, or below 30, for a long time while price keeps going - which is why selling every time RSI crosses 70 can be painful in a trend.
How traders use RSI
- Reversal from the extremes: the 30/70 cross back towards the middle, which is what we tested.
- Trend filter: RSI above 50 read as bullish momentum, below 50 as bearish.
- Divergence: price making a new extreme while RSI does not. Subjective to spot and not part of our test.
- Different bands: some use 20/80 or 40/60 depending on the market and timeframe.
What we measured: RSI 30/70 reversals on real charts
Each time RSI(14) rises from below 30 to 30 or above, that is a buy signal; each time it falls from above 70 to 70 or below, a sell. For each one, a signal counts as a win if price moved one average true range (ATR) in its direction before it moved one ATR against it, within 48 candles, and as a loss if the reverse happened first.
How the signals did in the last month
Now: 1 win in a row · longest 2 wins, 4 losses
Could be luck — chance alone gives 30–70% over 23 signals
A run does not make the next signal likelier to win — we checked 15,192 signals.
Now: 1 win in a row · longest 2 wins, 3 losses
Could be luck — chance alone gives 25–75% over 16 signals
A run does not make the next signal likelier to win — we checked 15,192 signals.
Now: 1 win in a row · longest 3 wins, 6 losses
Could be luck — chance alone gives 28–72% over 20 signals
A run does not make the next signal likelier to win — we checked 15,192 signals.
Now: 4 wins in a row · longest 4 wins, 3 losses
Could be luck — chance alone gives 29–71% over 22 signals
A run does not make the next signal likelier to win — we checked 15,192 signals.
The roulette view: the indicator next to a coin
Below is every decided signal from the four boards above, one after another, and under it the same number of tosses of a fair coin (a fixed simulated sequence, not re-rolled). Look for the runs: both have streaks of several wins and several losses. A streak is what randomness looks like; it is not a sign that the indicator is “hot” or that a win is “due”.
Streaks: the roulette view of RSI
The real RSI record and the simulated coin above both contain streaks. After a few oversold bounces in a row that worked, RSI feels reliable; after a few that failed in a downtrend, it feels useless. In 100 tosses of a fair coin, a run of six or seven of the same side is normal, so neither feeling tells you much.
We checked whether a losing run makes the next signal likelier to win. Across 15,192 hourly signals on 41 markets, after one to five losses in a row the next signal worked 41-56% of the time - within the range chance produces.
What that means in practice
- RSI is a clear measure of how one-sided recent movement has been. On our data, the 30/70 reversal on its own is not a forecast.
- Overbought is a description, not a price target.
- Any claimed RSI edge should come with many signals and a baseline. A few chart examples prove nothing either way.
Every market page shows RSI's scoreboard for that market on four timeframes - for example Bitcoin's signal truth section. Moving averages, MACD and the stochastic are covered in the other articles of this series.
Educational content, not financial or investment advice. Nothing here is a recommendation to buy or sell anything. Figures describe history and models, not what will happen. Trading leveraged products such as CFDs, forex and crypto derivatives carries a high risk of losing money.
See it on a chart
The Gold liquidation heatmap is free on a 24-hour delay — and live for 7 days when you create a free account. No card.