Home/Systematic Trading/Volatility Forecasting

Volatility Forecasting

volatility-tradingLevel 2 — Intermediate

Prerequisites

Volatility Trading

Unlocks

Convexity StrategiesMarket Making MechanicsVariance Risk PremiumVolatility Surface Reading

What It Is

The quantitative estimation of future realized volatility using historical price data, implied volatility, and statistical models — providing the "fair value" benchmark against which to measure whether options are overpriced or underpriced and to calculate the edge of any volatility trade.

Correct Execution

Practitioner builds at least two independent vol forecasts and triangulates between them: a simple realized vol estimate (e.g., 20-day historical vol as baseline), a GARCH-family model (capturing vol persistence and clustering), and IV as an implied market forecast. The forecast is used to calculate the expected edge of the trade: "The straddle is trading at $10 implied vol; my model says $7 fair vol; my edge is approximately $3/straddle." Forecast accuracy is tracked out-of-sample against realized vol — a practitioner who cannot prove their forecast is better than a naive model should not trade on it.

Progression Levels

Diagnostic Tree

Coaching Cues

  • "Vol is the most forecastable thing in markets. Use that." — Euan Sinclair framework
  • "Your edge is the difference between your vol forecast and the market's IV. No forecast, no edge — just luck."
  • "GARCH(1,1) is often 90% as good as any more complex model. Start there." — Euan Sinclair, Outlier Podcast, 2022-11-08
  • "Track your forecast accuracy. If you can't beat a 20-day naive estimate, stop pretending you have a forecast." — Euan Sinclair

Common Errors

  1. Treating IV as the vol forecast: IV contains useful forward-looking information but is systematically biased high (variance risk premium). Using IV directly as the forecast without adjusting for VRP overestimates future vol and destroys short-vol edge.
  2. Using a single vol estimate as a point forecast: Vol forecasts have wide confidence intervals. Treating a GARCH output as a precise number rather than a range leads to false precision in edge calculation.
  3. Ignoring vol persistence: Selling vol aggressively after a calm period without accounting for the possibility that vol is entering a rising regime. Vol clusters — calm periods are more likely to persist in the near term, but trend reversals in vol are also more persistent than reversal in returns.
  4. Not tracking forecast accuracy: Building a vol model and never comparing its out-of-sample accuracy to a naive baseline. Without validation, there is no way to know if the model adds value.

Edges

⚡ Conventional Wisdom Is Wrong

Vol Is The Most Forecastable Variable In Markets — Use That

volatility-tradingvol-forecasting →

Market participants spend the majority of research effort forecasting returns (extremely difficult — near-zero autocorrelation) while underinvesting in volatility forecasting (much more tractable — high autocorrelation). A 5-day GARCH forecast of realized vol has substantial predictive power; a 5-day return forecast is barely better than random. The practitioner who has a good vol forecast has a genuine, quantifiable edge that most market participants are not even trying to generate.

What most people do
Invest research effort in return forecasting (market timing, factor rotation, entry/exit signals). Treat vol as a risk metric rather than a primary forecasted variable.
What the best do
Invest heavily in vol forecasting infrastructure. Treat vol forecasting as the primary research advantage. Use return forecasting only where vol forecasting provides the confidence (vol is low and well-forecast) to exploit a return signal.
Why it's an edge: Return forecasting is the hardest problem in markets; vol forecasting is much more tractable. Redirecting research effort to the more tractable problem generates more reliable alpha.
How to exploit: Build a GARCH(1,1) or EWMA vol forecasting model as the foundation of any options strategy. Track out-of-sample mean squared error vs. a naive 20-day historical vol baseline. Any model that reduces MSE by >15% is adding genuine forecast value and should be maintained. Kill models that don't beat the naive baseline — they aren't forecasting.
Euan Sinclair, "Find Edge and Trade Volatility," Outlier Podcast, 2022-11-08
🔑 Hidden Causal Lever

IV Is A Biased Forecast — Correct For The VRP Before Using It

volatility-tradingvol-forecasting →

Implied volatility contains genuine forward-looking market information but is systematically biased high by the variance risk premium. A practitioner who uses IV directly as their vol forecast will consistently overestimate future realized vol. The correct procedure is to treat IV as a starting point and adjust it downward by the expected VRP for that instrument. The adjusted IV is a better vol forecast than either raw IV or pure historical vol.

What most people do
Use current IV as the vol forecast input. Or use historical realized vol and ignore IV entirely. Both are suboptimal — one is biased high, the other ignores forward-looking information.
What the best do
Build a blended forecast: (1) GARCH-based historical forecast (captures persistence); (2) IV minus the historically observed VRP for that instrument (captures forward-looking signal without VRP bias). Blend the two at approximately 50/50 or optimize the blend out-of-sample.
Why it's an edge: The VRP adjustment converts a systematically biased input into an unbiased one. This directly improves edge calculation accuracy — the difference between forecast and IV is the basis for position sizing.
How to exploit: For each instrument in the vol-selling universe, calculate the historical mean IV/RV ratio over 24 months. Use this as the VRP adjustment factor. When current IV is 20% above forecast GARCH vol, and the historical VRP for this instrument is 15%, the adjusted forecast is GARCH + 2.5% (not the full 20%). This is the input to edge calculation.
Euan Sinclair, "Positional Option Trading," Flirting with Models S3E12, 2021-04-10
🔑 Hidden Causal Lever

GARCH Models Normal Days — Events Are A Separate Problem

volatility-tradingvol-forecasting →

GARCH-family models are excellent at forecasting diffusive volatility — the continuous, day-by-day fluctuation that characterizes normal markets. They are designed for this problem and perform well within it. But realized vol during a period containing a major discrete event (earnings miss, geopolitical shock, data release) is dominated by the jump component, which GARCH cannot forecast. Treating a GARCH estimate as a complete vol forecast for any period containing known discrete events dramatically underestimates actual realized vol.

What most people do
Use GARCH as the primary vol forecast for all periods, including those containing known events. Attribute forecast errors near events to "model error."
What the best do
Operate two separate vol forecast models: (1) GARCH for diffusive vol in non-event periods; (2) event vol model for periods containing earnings, macro announcements, or other known discrete events. The event vol model uses the implied vol specifically around event dates (straddle pricing near earnings) as the event component, added on top of the GARCH diffusive estimate.
Why it's an edge: Correctly scopes what each model can and cannot do. Eliminates systematic forecast errors near events, which are the most common source of unexpected losses in vol strategies.
How to exploit: Build an event calendar integration into the vol forecasting pipeline. For any period containing a scheduled event (earnings, Fed meeting, macro data release), automatically flag it as an event period and add the event vol component to the GARCH forecast. Use implied vol of straddles with the event date in their life as the event vol proxy.
Euan Sinclair, "Positional Option Trading," Flirting with Models S3E12, 2021-04-10

Sources

  • Euan Sinclair, "Positional Option Trading" (Corey Hoffstein / Flirting with Models, S3E12), 2021-04-10 — GARCH, vol persistence, forecast-to-edge calculation
  • Euan Sinclair, "Find Edge and Trade Volatility," Outlier Podcast, 2022-11-08 — vol forecasting as the foundation of vol-trading edge
  • "How to Profit Trading Implied Volatility," Predicting Alpha / YouTube, 2023-07-28 — practical vol forecasting steps: simple, GARCH, IV-based blending