RSI Indicator: How It Works and How It Is Misread

RSI puts recent price movement on a scale from 0 to 100. It tells you how one-sided that movement has been. That is the whole job. Nearly every mistake made with it comes from wanting a different job done.

J. Welles Wilder published it in 1978. He used it carefully. What gets taught now is a coarser thing altogether.

The calculation

RSI = 100 − (100 ÷ (1 + RS)), where RS = average gain ÷ average loss over the period.

Run the default 14 periods and the indicator looks back over the last 14 bars, averages the gains on up days against the losses on down days, and drops that ratio onto a 0 to 100 scale.

Three readings show what the arithmetic actually means. RSI 100 would need 14 consecutive up periods and nothing down at all. RSI 50 says gains and losses have been roughly even. RSI 70 says recent gains have substantially outweighed recent losses. That is genuinely all it says.

Not a word about value. Nothing about whether the move has gone too far, nothing about what comes next. It reports what already happened.

Why “overbought” loses money

Sell above 70, buy below 30. That is where most people land. In a ranging market it works acceptably well.

Trends are where it falls apart. Strong uptrends park RSI above 70 for weeks at a stretch, which is the indicator doing its job correctly, reporting that gains have consistently outweighed losses, which happens to be the definition of an uptrend. Sell every overbought print in a market like that and you end up shorting strength over and over, each trade placed against the dominant flow, each loss arriving for the same reason as the one before it and none of them teaching you anything new.

Market condition	RSI above 70 means	Correct read
Strong uptrend	Buyers are firmly in control	Confirmation, not a warning
Established range	Price is near the top of the range	Potential fade, with a level
Late in an extended move	Momentum is stretched	Tighten stops, stop adding

Which row you are in depends on one variable: whether a trend exists. You settle that from price structure, before the oscillator gets a vote.

Divergence: the signal worth learning

Divergence happens when price and RSI disagree. It carries far more information than any threshold reading.

Bearish divergence: price makes a higher high while RSI makes a lower high. The new high arrived with less momentum behind it. Fewer participants. Less conviction. The bullish case mirrors it, with price making a lower low while RSI makes a higher low, which means the fresh low came with less selling force behind it than the one before.

Failure swings

Wilder’s own preferred signal. Almost nobody uses it.

The bearish version runs like this. RSI climbs above 70, pulls back, rallies again, fails to clear its previous peak, then breaks below the low of that pullback, and only at that final break does the signal exist. Everything there is generated inside the indicator itself, and the requirement is that RSI break its own structure, which asks considerably more of it than touching a number on the way past.

The bullish version mirrors all of that below 30. Both are stricter than a threshold cross. That strictness is why they produce fewer signals of better quality.

Using RSI sensibly

1. Establish the trend first, from price. RSI reads differently in a trend than in a range.
2. In a range, fade an RSI extreme that lands on a level you already marked. The level is the reason. RSI corroborates.
3. In a trend, use RSI for divergence and for timing pullbacks. RSI dipping to 40 inside an uptrend is often a better entry than RSI at 30 inside a downtrend, which is the sort of comparison a threshold rule cannot make.
4. Never trade the number alone. RSI 72 with no level and no structure around it is arithmetic.

Next: MACD for a trend-following counterpart, and moving averages for the simplest trend filter going. Put every trade that results through the position size calculator.