An indicator doesn’t know whether a currency or stock is “too expensive.” The relative strength index, developed by J. Welles Wilder, summarizes the balance of recent average gains and losses into an oscillator between zero and 100. It is a measure of recent price momentum, not intrinsic value. Both Fidelity’s indicator guide and Schwab’s technical explanation describe the common use of 70 as an overbought convention and 30 as an oversold one.
What the calculation is trying to capture
At a high level, RSI compares smoothed average upward moves with smoothed average downward moves over a specified lookback, often 14 periods. Strong recent gains push it higher; strong losses push it lower. The actual reading depends on the sampling interval: 14 five-minute bars and 14 daily bars can tell entirely different stories. Calling a value “the RSI” without naming the instrument, timeframe and lookback is incomplete.
A simplified mathematical outline is: relative strength equals average gain divided by average loss; RSI then maps that ratio to the zero-to-100 range. Wilder’s standard smoothing uses prior averages, so two platforms initialized with different histories can show small differences. Compare formulas and warm-up histories before assuming a discrepancy proves an error.
Overbought can persist during a trend
A common mistake is treating RSI above 70 as an automatic sell order. The Fidelity explanation explicitly notes that RSI can stay overbought or oversold during prolonged trends. If price keeps rising, successive strong bars can sustain high RSI even while short sellers absorb losses.
The inverse holds in downtrends: below 30 describes stretched negative momentum, not a guarantee of a quick rebound. Never make a risk decision solely from one number.
How an analyst might use it without pretending certainty
Use the reading as context and ask a question it could help test. For example, is momentum broadening or weakening as price approaches a previously studied level? Does a change persist across multiple bar intervals? Did the observed relationship hold in older data after realistic transaction costs?
A divergence—price reaching a new high while RSI does not—may suggest momentum has weakened, but it does not prove a reversal will happen at any particular bar. A stop or limit instruction entered from a signal still obeys the broker’s execution mechanics, explained in our order guide.
A small, reproducible research exercise
Choose a market with enough historical bar data. Specify the exact RSI calculation, lookback, thresholds and position entry/exit rules before inspecting the later period. Record all threshold events, not only memorable successes. Separate an initial sample used to design the rule from a later sample used to evaluate it.
Measure winning and losing outcomes, average holding period, drawdown, fees, spread and missed fills. Compare against a simple baseline, such as entering at fixed intervals or applying no trades. A higher hit rate can still produce losses if losing trades are larger than wins.
Common mistakes and limits
RSI does not measure fundamental value, guarantee reversals or account for execution costs. Its familiar 70/30 levels are conventions, not universal parameters. Technical patterns evaluated after selecting successful charts are susceptible to hindsight and survivorship bias. The visual accompanying this article is abstract artwork not an RSI plot of a real instrument. Use trading journal guidance for a testing record rather than relying on an attractive historical example.
