What Volatility Means for a Trader

For a trader, volatility is a measurement of how far a price typically moves over a given period, usually expressed as an annualized standard deviation of returns. It sets the scale for the rest: how wide a stop must be to sit outside normal noise, how many shares fit a fixed dollar risk, and what size of move counts as ordinary.

Key takeaways

Realized volatility is the standard deviation of past log returns, scaled to a year with √252. For the S&P 500 it was 9.5% over the 21 trading days to October 13, 2026, against 18.1% over ten years.
Implied volatility is read from option prices; the VIX is Cboe's 30-day measure for the S&P 500, with a median close since 1990 of 17.56.
Dividing an annual figure by 16 gives a typical daily move: a VIX of 15.27 points to about 0.95% a day.
Against one 10-year standard deviation, 1.43% of S&P 500 days moved more than 3σ, against 0.27% on a normal curve.
With a fixed dollar risk, a stop set at twice the average true range (ATR) produces smaller positions when volatility rises.

How realized volatility is measured

Realized, or historical, volatility describes moves that have already happened. The standard method takes the standard deviation of daily log returns and scales it to a year. If daily returns are independent, variances add across days, so standard deviations grow with the square root of the number of days.

r_t = ln(P_t / P_t-1)
daily σ = standard deviation of r_t over N days
annual σ = daily σ × √252

P_t  = closing price on day t
N    = lookback window in trading days (for example 21, 63 or 252)
252  = approximate number of US trading days in a year

Applied to the S&P 500 daily closes published by FRED (a price index that excludes dividends), the answer depends heavily on the window:

Window ending October 13, 2026	Daily σ	Annualized σ
Last 21 trading days	0.60%	9.5%
Last 63 trading days	0.69%	10.9%
Last 252 trading days	0.80%	12.7%
Oct 2016 – Oct 2026 (2,511 days)	1.14%	18.1%

Calendar years ranged from 6.7% in 2017 to 34.7% in 2020, so a volatility figure means little without its window and dates.

Average true range: a price-unit measure

Average true range, introduced by J. Welles Wilder in his 1978 book New Concepts in Technical Trading Systems, measures range rather than returns and is expressed in price units (dollars per share, points per contract), which suits stop distances. ICE's chart-study documentation sets out the calculation:

TR_t  = max( H_t − L_t , |H_t − C_t-1| , |L_t − C_t-1| )
ATR_1 = average of the first n true ranges
ATR_t = ((n − 1) × ATR_t-1 + TR_t) / n      (n is commonly 14)

H, L, C = the period's high, low and close; C_t-1 = the prior close

Illustrative example: after a $99.00 close, a stock trades between $98.50 and $101.00; the true range is $2.50, the same as the high-low range. On another day, after a $100.00 close, it opens higher and trades between $104.50 and $106.00. The high-low range is $1.50, but the true range is $6.00, because it counts the gap from the prior close. With a 14-day ATR of $2.50 and a next true range of $4.00, the new ATR is (13 × 2.50 + 4.00) ÷ 14 = $2.61: it moves gradually.

Implied volatility and the VIX

Implied volatility runs the other way: it is the volatility figure that makes an option pricing model match the option's market price. When traders pay more for options, other inputs unchanged, the implied figure rises. It is a price, not a guaranteed forecast.

According to Cboe's VIX methodology, the VIX measures 30-day expected S&P 500 volatility from mid-quote prices of SPX puts and calls across many strikes, interpolating near- and next-term expirations to a constant 30 days, with each option's contribution inversely proportional to its strike squared. The VIX is quoted as an annualized percentage, so a reading of 20 means about 20% per year.

In the FRED VIX series, 9,294 daily closes from January 1990 to October 13, 2026 have a median of 17.56 and a mean of 19.42, pulled up by short spikes such as the record 82.69 on March 16, 2020. The October 13, 2026 close was 15.27.

Implied usually exceeds realized

Comparing each VIX close from October 2016 with the S&P 500's realized volatility over the next 21 trading days, the VIX was higher 84% of the time, and the median gap across all periods was 4.7 points. Carr and Wu (Review of Financial Studies, 2009) measure this gap as a variance risk premium, and the paper's 2004 working version finds it strongly negative on average for the S&P 500 and S&P 100, which it reads as variance buyers accepting a cost to hedge upward moves in index variance.

In the other 16% of periods realized volatility overshot, usually by a few points (median 3.7) but not always: the VIX closed at 14.38 on February 19, 2020, and the next 21 trading days realized 84.3%. A gap in volatility points is not a trading return: it ignores costs, margin and those losing periods.

From annual figures to daily moves

Because √252 is 15.87, dividing an annualized volatility by 16 estimates one daily standard deviation. This "rule of 16" puts a VIX of 15.27 at about 0.95% a day, roughly 74 S&P 500 points at the October 13 level of 7,805. Realized daily σ over the prior 21 days was lower, at 0.60%.

Longer horizons use the same scaling. Cboe's mathematics methodology measures time in minutes of a 365-day year (525,600 minutes), so the VIX's own convention is calendar time:

expected 1σ move = price × (IV / 100) × √(days / 365)

IV   = implied volatility in annual percent (for example the VIX)
days = calendar days to the horizon

Illustrative example: with the index at 7,805 and the VIX at 15.27, the 30-day factor is √(30/365) = 0.287, so one standard deviation is 4.38%, about ±342 points; two are about ±683 points. Trading days (√(21/252) = 0.289) give nearly the same answer. For one day the conventions split: VIX ÷ √365 gives 0.80%, but that spreads variance over weekends when the index does not trade. Spreading the same 30-day variance over the roughly 21 trading days in it gives 0.95%, the rule-of-16 figure, which is the one to compare with daily closes.

Probability ranges and fat tails

Under a normal distribution, 68.3% of outcomes fall within one standard deviation and 95.4% within two. Market returns have fatter tails. In the FRED S&P 500 data from October 2016 to October 2026, against the full-period standard deviation:

Daily move larger than	Days observed	Share observed	Share if normal
1σ (1.14%)	481	19.2%	31.7%
2σ (2.28%)	112	4.5%	4.6%
3σ (3.42%)	36	1.43%	0.27%
4σ (4.57%)	19	0.76%	0.006%

Quiet days were more common than a normal curve predicts, and extreme days far more common: 36 beyond 3σ against about 7 expected. The largest was March 16, 2020, when the index fell 12.0% (a log return of −12.8%), about 11 standard deviations. Part of this excess comes from using one standard deviation for calm and turbulent years alike: measured against the prior day's VIX ÷ √252, only 9 days (0.36%) moved more than 3σ.

Volatility clusters and reverts

Volatility is not constant. The 2003 Nobel press release for Robert Engle's ARCH models describes turbulent periods with large fluctuations followed by calmer periods with small ones. In the FRED data, the size of a day's S&P 500 move had an autocorrelation of about 0.37 with the previous day's size, so large moves tended to arrive in groups.

Volatility also tends to drift back toward its long-run level. Of the 207 VIX closes above 40 since 1990 with 63 trading days of data after them, 98.6% were followed by a lower reading three months later; from readings below 12, the VIX was higher three months later 77% of the time. That is history, not a forecast. One consequence: a stop set during a calm stretch can sit inside ordinary noise once volatility rises, and the position sized from that tight stop is larger than the same dollar risk would allow at the new volatility.

Why volatility products decay

The VIX itself cannot be bought. FINRA's investor guidance notes that volatility-linked exchange-traded products track VIX futures, generally aren't designed as buy-and-hold investments, and can lose some or all of their value in a very short time.

VIX futures prices have tended to rise with time to expiration, a shape called contango, so a product holding a constant short-term exposure keeps selling nearer contracts and buying later, pricier ones. FINRA Regulatory Notice 17-32 (October 16, 2017) reminded firms that this rolling can produce a negative roll yield, that many volatility-linked ETPs are highly likely to lose value over time, and that they may be unsuitable for some retail investors, particularly those who plan to use them as buy-and-hold investments. The notice also explains that when the curve is in backwardation, with nearer contracts priced higher, rolling produces a positive roll yield.

Inverse products carry a different tail risk. On February 5, 2018, the intraday indicative value of the VelocityShares Daily Inverse VIX Short-Term ETN (XIV) fell to 20% or less of the prior day's closing indicative value of $108.3681, a drop of at least 80%. In its February 6, 2018 notice filed with the SEC, issuer Credit Suisse stated that an acceleration event had occurred and gave February 20 as the expected last trading day.

Volatility in position sizing

A volatility measure converts a fixed dollar risk into a position size. With the stop set as a multiple of ATR (or of daily σ in price terms), size adjusts to how much the instrument moves:

dollar risk     = account value × risk per trade
stop distance   = k × ATR             (k = chosen multiple, e.g. 2)
position (units) = dollar risk ÷ stop distance

Illustrative example: a $50,000 account with a 1% risk limit has $500 at risk. For a hypothetical $100 stock with a 14-day ATR of $2.50, a 2×ATR stop is $5.00 away, so the position is 500 ÷ 5 = 100 shares, worth $10,000, or 20% of the account. If ATR doubles to $5.00, the $10.00 stop allows 50 shares worth $5,000. The site's Position Size & Risk calculator applies the same arithmetic.

A stop limits loss only at prices the market actually trades. Investor.gov notes that a stop order becomes a market order when the stop price is reached. If the stock in the first case closed at $100 and opened the next morning at $88, the order would fill at the next available price, near $88, for a loss of about $1,200 (2.4% of the account) rather than $500.

Frequently asked questions

Which lookback window should historical volatility use?

There is no single correct window; 21, 63 and 252 trading days roughly match a month, a quarter and a year. Short windows react fast but are noisy; long ones are stable but slow, so state the window with the result.

Why use log returns instead of percentage changes?

Log returns add across days, which keeps square-root-of-time scaling consistent. For small moves the two are nearly identical (a 1% gain is a log return of 0.995%). For large ones they diverge: +10% then −10% leaves a price 1% lower, and the log returns sum to −1.005%, recording that loss; the percentages sum to zero.

Is the square root of 252 right for every market?

The factor should match the number of return periods a year in the data. About 252 fits US stock exchanges, closed on weekends and holidays; a market quoted every calendar day, such as most crypto venues, would use √365, and intraday data needs the number of intervals in a trading year.

Does the VIX usually rise when stocks fall?

Yes. Cboe's methodology calls the negative correlation between volatility and stock returns well documented. In the 2016–2026 FRED data, the correlation between the S&P 500's daily log return and the VIX's daily change was about −0.79, and the VIX rose on 75% of the days the index fell. On the other 25% of down days it was flat or lower.

How often has the VIX closed above 30?

On 7.9% of trading days from January 1990 to October 13, 2026 in the FRED VIXCLS series; above 20 on 36.9%, above 40 on 2.2% and above 50 on 0.8%. The readings above 30 clustered: 190 of the 735 came in 2008–2009 alone.

Checking a volatility figure

Type: realized (past returns), range-based (ATR, in price units) or implied (option prices).
Window and end date: a 21-day and a 10-year figure for the same index can differ by a factor of two.
Time convention: √252 for trading days, √365 for calendar days as in the VIX.
Horizon: price × σ × √(t) gives a typical move, not a limit; tails and overnight gaps can go well past it.
Position size: recomputed when ATR or σ changes materially, with the dollar risk held fixed.