RSI — the Relative Strength Index — compares the average size of recent gains to the average size of recent losses and scales the result from 0 to 100. It measures how one-sided recent price movement has been. It does not measure value, and a high reading means a market is strong, not that it is finished.
That distinction is where almost all RSI losses come from. The indicator is a momentum thermometer, and traders keep reading it as a price tag.
How the number is built
Over a lookback window — 14 periods by default — RSI takes the average gain on up candles and the average loss on down candles, forms the ratio between them, and converts it to a 0–100 scale. The mechanics matter less than what they imply:
- RSI 50 means average gains and average losses have been roughly equal over the window. It is the balance point.
- RSI 70 means up-moves have been substantially larger than down-moves recently. Buyers have been in control.
- RSI 30 means the reverse. Sellers have been in control.
- RSI 100 would require every candle in the window to close up. It is an arithmetic limit, not a prophecy.
The 14-period default comes from Welles Wilder, who introduced the indicator in New Concepts in Technical Trading Systems in 1978. Nothing about 14 is optimal; it was a sensible starting point for daily charts in an era of hand calculation, and it stuck because enough people use it that it has become a shared reference.
Why "overbought" gets people short in uptrends
The word overbought sounds like a verdict. It is a description of the recent past with no forward content whatsoever. A market in a genuine trend will spend long stretches above 70, and each of those candles is one where the RSI-based short seller is losing money.
The correct reading is the opposite of the intuitive one. Sustained readings above 70 are evidence of trend strength, not evidence of exhaustion. If an instrument reaches 70 and stays there for two weeks while grinding higher, the indicator is telling you the trend is real. Fading it is a decision to be systematically wrong during the part of the move that pays most.
Read the range, not the line
The most useful RSI technique gets the least attention: watch which part of the 0–100 scale the indicator is living in, because that band shifts with the market's regime.
| Regime | Where RSI tends to travel | What that tells you |
|---|---|---|
| Uptrend | Roughly 40 to 80; pullbacks stall near 40–50 | Dips are being bought; 40 acts as a floor |
| Downtrend | Roughly 20 to 60; rallies stall near 50–60 | Rallies are being sold; 60 acts as a ceiling |
| Range | Full swings between 30 and 70 | Neither side in control; extremes mean more here |
Two things follow. First, the classic buy-30 / sell-70 approach only makes sense in the third row — a genuine range — and applying it in the first two rows is what produces the endless chop traders complain about. Second, the moment a market breaks its established RSI band is genuinely informative. When an uptrend that had been holding 40 on every pullback suddenly closes at 28, the character of the market has changed before the price structure has confirmed it. That is the one place RSI is close to leading rather than lagging.
Divergence, honestly
Divergence is when price makes a higher high but RSI makes a lower high — or price makes a lower low while RSI makes a higher low. The observation is real: the second push was weaker in momentum terms than the first.
The problem is timing. Strong trends routinely print three or four divergences before anything reverses, and there is no way to know from the divergence itself which one is the last. Every failed divergence traded as a reversal is a loss taken against the prevailing trend, which is the most expensive way to be wrong. Treat it as a state change worth respecting — a reason to trail a stop closer or take partial profits on a position you already hold in the direction of the trend. Not a reason to open a new position in the opposite one.
Settings, and the trap in choosing them
A shorter lookback (7) reaches the extremes far more often and suits fast intraday work. A longer one (21) is steadier and better for swing time frames. Neither is more correct. The trap is not the number itself — it is the process of hunting for it.
The published evidence makes this concrete. Chong, Ng and Liew revisited two widely-used oscillator rules across several developed markets and found the result depends heavily on both the market and the parameter set: MACD(12,26,0) and RSI(21,50) consistently generated significant abnormal returns in the Milan Comit General and the S&P/TSX Composite Index, while RSI(14,30/70) was the profitable configuration in the Dow Jones Industrials (Chong, Ng & Liew, "Revisiting the Performance of MACD and RSI Oscillators", Journal of Risk and Financial Management 7(1), 2014, pp. 1–12).
Notice what that actually says. Different settings worked in different markets — which is exactly the pattern you would expect from a search across many parameter combinations, and exactly why a result that holds in one index is not a promise about the next. Pick a setting for a reason you can state, then leave it alone. If you find yourself testing your fifth RSI configuration this month, you are not improving the tool; you are fitting it to history, the same failure documented for moving average rules.
Frequently Asked Questions
What does RSI actually measure?
RSI compares the average size of gains to the average size of losses over a lookback window, usually 14 periods, and scales the result from 0 to 100. A reading of 70 means recent up-moves have been substantially larger than recent down-moves. It is a measure of the character of recent price movement, not of value, and it says nothing about whether an instrument is expensive.
Is an RSI above 70 a sell signal?
No. Above 70 means momentum is strong, and strong markets stay above 70 for long stretches. Selling every time RSI crosses 70 in an uptrend means shorting the strongest part of the move repeatedly. Overbought describes what has already happened; it carries no information about when it will stop. Treat a high reading as a reason to manage an existing position, never as a reason to open a new one against the trend.
What RSI settings should I use?
Start with the 14-period default that Welles Wilder published in 1978 and change it only for a stated reason. A shorter lookback such as 7 reaches the extremes more often and suits scalping; a longer one such as 21 is steadier and suits swing time frames. There is no universally correct setting, and cycling through settings until one fits the last month of chart is how traders manufacture edges that do not exist.
Does RSI divergence work?
Divergence is a real observation with poor timing. When price makes a higher high while RSI makes a lower high, the second push was genuinely weaker in momentum terms. But a trend can produce three or four divergences before it turns, and each one that fails costs money if traded directly. Use divergence as a reason to tighten risk or take partial profits, not as a standalone reversal entry.
Bottom line
RSI is a well-built momentum gauge that most traders point in the wrong direction. Stop reading 70 and 30 as instructions, start reading which band the indicator is living in, and let a break of that band tell you the regime has changed. Use divergence to protect profit rather than to call tops. Above all, put the indicator behind the level: the trade is decided by a price with a defined invalidation, and RSI is only there to tell you how the market got to it. Read technical analysis basics for how the pieces fit together, what makes a breakout real for the price side of the same question, and how the room's callouts are structured if you want to see what we actually trade from.