Relative Strength Index

The Relative Strength Index is the most over-used and most misunderstood momentum indicator in retail trading. It is also the indicator most likely to be cited in a course, app, or screener that does not work. That is not an indictment of RSI - the math is fine. It is an indictment of how it gets taught: a green dot when RSI hits 30, a red dot when it hits 70, and an account that bleeds out trying to fade strong trends.

The actual value of RSI is more specific and more limited than the marketing suggests. This page walks through what RSI measures, the canonical formula, a worked example you can verify by hand, the two or three setups where it provides real edge, and the trap that costs newer traders the most money.

What RSI actually measures

RSI was developed by J. Welles Wilder Jr. and published in his 1978 book New Concepts in Technical Trading Systems. The premise is simple: if recent up-bars dominate, the market is in a buying mode and RSI rises toward 100. If recent down-bars dominate, RSI falls toward 0.

It is critical to read RSI for what it actually says: it tells you the dominance of buying pressure relative to selling pressure over the lookback window. It does not tell you that the move is "too big" or "about to reverse." Those are interpretations layered on top of the math - interpretations that hold up in ranges and fall apart in trends.

The 0-to-100 scale is bounded, which makes RSI visually easy to read at a glance. That accessibility is also its biggest UX trap. A bounded oscillator looks like it should reverse at the boundaries. In a trending market it does not.

The formula

The textbook formula is:

RSI = 100 - (100 / (1 + RS))

where RS = Average Gain over N periods / Average Loss over N periods

The two averages are computed using Wilder's smoothing, which is a modified exponential moving average. The first calculation uses a simple average of gains and losses over the first N periods. Each subsequent bar updates using:

Average Gain (new) = ((Average Gain (prior) × (N - 1)) + Current Gain) / N
Average Loss (new) = ((Average Loss (prior) × (N - 1)) + Current Loss) / N

A "gain" on a bar is the close-to-close change when positive; otherwise the gain is 0. A "loss" is the absolute close-to-close change when negative; otherwise the loss is 0. The loss is always recorded as a positive number.

A worked example

Take a 14-period RSI on a hypothetical instrument. The closing prices for the last 15 bars are:

50, 51, 52, 50, 49, 51, 52, 53, 52, 54, 55, 53, 54, 56, 58

Step 1 - compute the close-to-close changes:

+1, +1, -2, -1, +2, +1, +1, -1, +2, +1, -2, +1, +2, +2

Step 2 - separate gains and losses across the first 14 bars:

Total gain = 1+1+2+1+1+2+1+1+2+2 = 14
Total loss = 2+1+1+2 = 6

Step 3 - compute averages over 14 periods:

Average gain = 14 / 14 = 1.00
Average loss = 6 / 14 ≈ 0.43

Step 4 - relative strength and RSI:

RS = 1.00 / 0.43 ≈ 2.33
RSI = 100 - (100 / (1 + 2.33))
    = 100 - (100 / 3.33)
    = 100 - 30.03
    ≈ 69.97

So on bar 15 our RSI reads around 70 - the canonical "overbought" boundary. The next bar will use Wilder's smoothing rather than a fresh 14-period average.

You can do this exact computation in a spreadsheet and verify it against the RSI on your platform. They should match to within rounding.

How traders actually use RSI

There are three setups where RSI provides edge that holds up across thousands of trades. Everything else is noise.

1. Divergence

Bullish divergence: price makes a lower low while RSI makes a higher low. The market is making fresh lows but with less downward force behind each one. This is RSI's strongest single signal. It works because momentum tends to lead price at trend transitions.

Bearish divergence is the inverse: price makes a higher high, RSI makes a lower high. The new high lacks the conviction of the prior one.

Caveat: divergence by itself is not an entry. It is a setup that must be confirmed by structure - a break of the prior swing low (for bearish divergence) or swing high (for bullish divergence). Without confirmation, divergence can persist for weeks while the trend continues.

2. The 50-line as a trend filter

The midpoint of RSI is 50. When RSI holds above 50 on pullbacks, the trend is intact. When it loses 50 and stays below, momentum has flipped. This is a much cleaner trend filter than the 70/30 lines.

Many systematic strategies use "RSI > 50" or "RSI < 50" as a binary regime switch rather than as an overbought / oversold signal.

3. Failure swings

A failure swing is RSI's most rules-based reversal pattern. Bearish failure swing:

1. RSI rallies above 70
2. RSI pulls back without falling below 70
3. RSI rallies again but fails to make a new high
4. RSI breaks below the local low formed in step 2

That break in step 4 is the trigger. Bullish failure swings are the mirror image around the 30 line. Failure swings are rare but high-conviction setups.

The trap most retail traders fall into

The default RSI lesson in 99 percent of YouTube videos, online courses, and indicator apps goes like this: "RSI > 70 means overbought, sell. RSI < 30 means oversold, buy."

This is the most expensive sentence in retail trading.

In a strong uptrend, RSI can stay above 70 for weeks at a time. NVIDIA in 2024 spent stretches of 4 to 6 weeks with RSI never closing below 70. Selling every 70 print into that move would have produced a series of losing short trades against the dominant trend.

The math is straightforward. RSI is a bounded oscillator, but the price it is computed on is not. A market that puts in twenty up-days in a row will pin RSI near 100, and the indicator has nowhere to go. The "overbought" reading is real - momentum is high - but it is not predictive of a reversal. It is predictive of momentum, which is the opposite of what the textbook signal implies.

The fix:

• Treat 70 and 30 as warnings, not entries
• Require trend context before any counter-trend RSI read (price below the 200-SMA before fading a 70 print, for example)
• Use the 50-line and divergence as the actionable RSI signals, not the bands

RSI vs other momentum indicators

vs Stochastic. Both bounded oscillators with similar visual feel. Stochastic measures where the close sits in the recent high-low range; RSI measures the ratio of average gains to average losses. Stochastic is faster and noisier; RSI is slightly smoother. Practical difference: stochastic generates more signals (more false signals in a trend); RSI generates fewer.

vs MACD. MACD is unbounded and built from the gap between two EMAs. It is built for trends, where RSI is built for momentum within a range. The two are complementary, not redundant - many traders run both: MACD for trend direction, RSI for momentum context.

vs CCI. CCI is unbounded and measures deviation from the mean. CCI is more volatile and produces extreme readings (+200, -300) that RSI cannot. CCI is favored for mean-reversion setups; RSI for general momentum reads.

Common questions

Is RSI better on stocks or futures or crypto? The math is identical across asset classes - RSI is computed the same way on a Bitcoin chart or an ES futures chart or an Apple stock chart. The interpretation changes with the volatility regime. Crypto tends to have prolonged stretches above 70 or below 30 because the underlying volatility is higher. Treat the bands more loosely on volatile instruments.

Should I change the 14-period default? For most use cases, no. The 14 setting is what every other trader looks at, which makes it a coordination point - support and resistance, divergence levels, and signals are all read by other participants using the same number. Shortening to 7 makes RSI faster and noisier (more signals, more false signals). Lengthening to 21 or 25 smooths it (cleaner signals, but later).

Does RSI work on intraday charts? Yes, with caveats. On the 1-minute and 5-minute timeframes, RSI is noisy because every gap, news catalyst, and algorithm flush pushes it. The 15-minute and 1-hour timeframes produce more reliable RSI signals. Pair with a higher-timeframe trend filter.

What is a "good" RSI value? Not a sensible question. RSI is contextual - 60 in a strong uptrend is healthy; 60 in a downtrend can be a counter-trend bounce nearing exhaustion. Treat RSI as a relative measure, not an absolute one.

How is RSI different from RS in equity analysis? Different concept. Equity "relative strength" compares a stock's performance to a benchmark (often the S&P 500). RSI is an internal momentum calculation on a single instrument. Same word, different math.

Can RSI predict a reversal? No. RSI describes recent momentum. A reversal requires a confluence of price structure breaking, momentum failing to confirm new highs (divergence), and often a catalyst. RSI is one input in that confluence, not the trigger.

When to use RSI and when not to

Use RSI when:

• You are looking for divergence in a market that is reaching new highs or lows on weakening momentum
• You need a trend filter cleaner than visual reading (the 50-line)
• You are scanning for failure-swing setups across a watchlist

Skip RSI when:

• You are in a strong directional trend and tempted to fade extreme readings
• The market is choppy and every cross of 50 produces a whipsaw
• You already have ADX, MACD, and other indicators on the chart - more indicators do not add information, they add noise

The smallest viable indicator stack for most discretionary traders is one trend filter (200-SMA or ADX) and one momentum filter (RSI is fine here). Anything beyond that needs to earn its place in a journaled, evidence-based way.