Relative Volatility Index (RVI): Dorsey’s Formula, the 50 Line, and How It Compares With RSI

The Relative Volatility Index (RVI) is an oscillator from 0 to 100 that tells whether recent volatility came mostly on up days or on down days. Donald Dorsey introduced it in Technical Analysis of Stocks & Commodities in June 1993, as a confirming indicator: TradingView reports that Dorsey saw it as supplying information that the Relative Strength Index lacks. Dorsey is also the author of the Mass Index. The RSI averages the size of up and down closes; the RVI averages the standard deviation of prices on up and down closes. Two indicators share the RVI acronym: this one, and the Relative Vigor Index, a momentum oscillator built on the close relative to the open.

How the Relative Volatility Index is calculated

LuxAlgo describes the original steps. Compute a rolling standard deviation of closes, 10 periods in Dorsey’s article. On a bar where the close rose, assign that deviation to the up bucket and 0 to the down bucket; on a bar where it fell, the reverse. Smooth both buckets over 14 periods with RSI-style averaging, and plot the up share times 100. LuxAlgo writes the averaging as an exponential average; TA-Lib, the open-source library behind many platforms, uses Wilder's smoothing instead, with the same 10 and 14. TA-Lib, under the name RVIR, and LuxAlgo describe Dorsey’s 1995 revision as the average of the index computed on highs and the index computed on lows; we could not read the 1995 article itself. TradingView describes its RVI as using the standard deviation of highs and lows.

The deviation itself has no sign: a calm rise and a calm fall both give a small value. The direction comes only from which bucket receives it, so a large deviation on a down close pulls the index down as hard as the same deviation on an up close pushes it up.

How to read the RVI: the 50 line

Dorsey used the RVI as a gate rather than a signal. The summary of his 1993 article describes it as a confirmation for a crossover of two moving averages, and LuxAlgo writes that his rules only allowed trades in the direction the RVI endorsed. His September 1995 article, “Refining the Relative Volatility Index”, adds a long-term trend filter. TradingView gives the plain version: above 50, volatility is to the upside and a potential buy signal is confirmed; below 50, to the downside, confirming a potential sell signal. TradingView adds that, like the RSI, the RVI can be used to spot overbought and oversold conditions, and LuxAlgo reads RVI lower highs while price makes higher highs as a bearish divergence in the volatility backing the move.

RVI on a real chart

The chart shows RVI(10, 14) on the daily S&P 500 ETF from March 26 to October 28, 2026, with dividend-adjusted prices. It made its low of the window, 23.5, on March 30, the day the ETF made its low close, 628.78. It crossed above 50 on April 2, at 652.52, and reached its high, 85.3, on April 17, at 706.56. It stayed above 50 until June 10, at 721.77, while the ETF gained 10.6%.

After that, the rule gave little guidance. From June 10 to the end of the window, RVI crossed 50 another 25 times, often for a day or two, while the ETF rose 7.0%, from 721.77 to 772.31. It ended the window at 59.3.

Does it work?

Our bench tests chart patterns detected by rule against random entries, not indicators (see our methodology). We measured two things on the daily S&P 500 ETF from 1993 to 2026 and on Bitcoin from 2017 to 2026, without costs. First, how close the RVI is to the RSI: the correlation of their daily values was 0.729 on the ETF and 0.676 on Bitcoin. Second, whether Dorsey’s gate improved a simple long signal, the first close above the 20-day simple average.

One ETF signal, on March 2, 1993, came the day before the RVI’s first reading; it counts only in the first row. Neither market rewarded the gate: signals the RVI confirmed did no better than those it rejected. These are descriptive averages, not a trading test.

The smoothing matters at the margin. With Wilder's smoothing in place of the exponential average, the two versions of the RVI had a correlation of 0.946 on the ETF and 0.943 on Bitcoin, but sat on different sides of 50 on 10.0% and 10.9% of the days, and that is enough to move signals from one row to the other:

With either smoothing, the rejected signals did at least as well as the confirmed ones on average; on Bitcoin, with Wilder's smoothing, the rejected ones did clearly better. A gate that changes its verdict with the averaging method is not one to rely on without testing it on your own market.

The Average True Range measures the size of volatility without a direction, Bollinger Bands draw the standard deviation around the price, and the Choppiness Index says whether the market trends. See all volatility indicators.