RSI: The Relative Strength Index

How the Number Gets Built

RSI compares the magnitude of recent up closes to the magnitude of recent down closes over a lookback period. The default Wilder set is 14 periods. The output runs from 0 to 100, with three zones traders watch: above 70 is overbought, where price has run hot relative to its recent range; below 30 is oversold, where price has dumped hard relative to recent range; around 50 is neutral, mid-range, no edge from RSI alone.

The 14-period default is a compromise between responsiveness and noise. Drop it to 7 and the line whips around, throwing signals every other day. Push it to 21 and you'll be late on every reversal. I run 14 on daily charts and I don't touch it. If you're scalping the 5-minute, a shorter lookback can make sense. For swings, keep Wilder's number.

The Overbought/Oversold Trap

Here's the mistake that kills more accounts than any other RSI error: treating 70 as a sell button and 30 as a buy button. It's the first thing every book teaches and the first thing that costs you money in a real trend.

NVDA in 2023 is the textbook case. The stock ripped from the 140s in January to over 480 by year-end. RSI on the daily was above 70 for huge stretches of that move — sometimes weeks at a time. I watched two guys at the desk short every print of 70. They got run over again and again. Same logic in reverse killed people fading down moves in 2022. Strong trends keep RSI pinned at the extreme. That's not a sell signal. That's the tell that the trend is real.

Let me put a specific trade on it. NVDA, May 24, 2023, the night of the Q1 earnings beat. The stock gapped from 305 to 380 pre-market. RSI on the daily was already at 78 the day before. Every textbook reading said don't touch it long. I bought 200 shares at 381.40 the morning after the gap, stop at 366, risk a touch over $3,000. RSI stayed pinned in the 75-82 range for the next four weeks. I trimmed half at 410, moved the stop to break-even, and rode the rest to 470 before getting shaken out on a wick in late June. Roughly 2.6R on the position. If I had used RSI 70 as a sell signal, I'd have been short into a 25% rip. The signal that works is the signal that fits the tape you're actually trading.

RSI's overbought/oversold reads work best in chop — range-bound tape where price is bouncing between [[Support and Resistance|support and resistance]]. In a clean range, RSI 70 near the top of the box is a reasonable fade and RSI 30 near the bottom is a reasonable bounce play. The mistake is using the same playbook in a trending market. Always check the trend first. RSI second.

Divergence: Where the Real Edge Is

If you only take one thing from this article, take this: the most useful RSI signal isn't the level. It's divergence.

Divergence is when price and RSI disagree. Bearish divergence shows up when price prints a higher high while RSI prints a lower high — momentum is bleeding out under the surface even as the chart still looks bullish. That's the classic warning at the right shoulder of a [[Head and Shoulders Pattern|head and shoulders]] top. Bullish divergence is the inverse: price prints a lower low, RSI prints a higher low. Selling pressure is fading. The downtrend is running out of fuel and a base is forming somewhere near.

Divergence is not a timing tool. It's an alert. The actual reversal can come days or weeks after the divergence shows up. Sometimes price keeps grinding while RSI keeps weakening, then the break finally happens. Treat divergence as the moment to tighten your stop, trim into strength, and start watching for a real trigger from [[MACD: Moving Average Convergence Divergence|MACD]] or [[Understanding Volume|volume]]. Don't short on the divergence alone. Wait for the structure to break.

RSI Setups That Actually Work

Trend Confirmation Bands

In an uptrend, RSI tends to oscillate between roughly 40 and 80. Pullbacks to RSI 40-50 are buy zones inside the trend — not RSI 30. If you're waiting for RSI 30 in a clean uptrend, you'll wait forever, because the stock will never get that weak before resuming. If RSI breaks below 40, the trend is in trouble. In a downtrend, RSI oscillates between roughly 20 and 60. Rallies into 50-60 are the spots to fade or short, not the spots to chase long.

Failure Swings

A failure swing is a reversal signal that lives entirely on the RSI panel, without reference to the price chart. Bearish version: RSI pushes above 70, pulls back, rallies again but fails to make a new high, then breaks below the prior RSI low. Bullish failure swing is the same thing inverted below 30. Wilder considered these among the cleanest signals the indicator produces. He was right. They're rare, but when they show up they tend to mark real turns.

Stacking RSI With Other Tools

RSI alone is mediocre. RSI plus context is good. The stacks I run on my own book: RSI plus [[Support and Resistance]] — long when RSI is oversold AND price is sitting on a defined support level. Two confirmations together are worth a lot more than either alone. RSI plus [[Moving Averages Explained|Moving Averages]] — in an uptrend, look for RSI to pull back to 40-50 while price taps the 50-day MA. That's a high-quality dip buy. RSI plus [[Understanding Volume|Volume]] — bullish divergence on RSI followed by a rally on expanding volume (call it 150% of the 30-day average) is one of the cleaner reversal triggers you'll find on a daily chart.

Think about it the way you'd read offshore water before dropping a line. The RSI level is one piece of information — the surface temperature, say. Support, the moving average, the volume, the broader trend — those are the color of the water, the bait scattering, the birds working. You don't drop bait because the temperature looks right. You drop because three or four signs line up at the same edge. Same idea on the chart. Stack reads until the math is in your favor.

RSI on Penny Names

RSI on [[Understanding Penny Stocks|penny stocks]] does not behave like RSI on large-caps. Pennies are thinner, more volatile, and the tape moves on smaller orders. RSI prints of 85 or 90 are routine on a penny rally that lasts three sessions. Readings under 20 happen on every halfway-decent flush.

For sub-$5 names, widen the bands. Use 80/20 instead of 70/30. The standard Wilder levels generate too many premature triggers on volatile names — you'll be flagging “overbought” on day two of a five-day rip. The wider bands filter out the noise and only fire on the genuinely extreme readings that tend to mark real exhaustion. I traded a sub-dollar runner in late 2020 (won't name it because the float was a joke) where the daily RSI hit 78 on day two and stayed there through day six. Anyone shorting the 70 print got their face peeled off. By the time RSI tagged 88 on day seven, the print on the offer was already fading. That's the read the wider band gives you.

Practical Rules

If you take RSI seriously, three things matter more than the level itself. The trend you're trading in. The structure on the price chart. The other indicators in agreement or disagreement with the read. Size is the knob that matters most of all — but that's a different article.

RSI is a tool, not a system. It tells you when momentum is stretched and when momentum is diverging from price. That's it. Anyone selling you a fully mechanical RSI strategy is selling a backtest, not a method. Use it as one input. Pair it with structure. Respect the trend. The chart will tell you which read matters; the indicator only tells you to look.

See also: [[MACD: Moving Average Convergence Divergence|MACD]] · [[Moving Averages Explained]] · [[Support and Resistance]] · [[Head and Shoulders Pattern]]