Volatility Targeting: What the Backtests Don't Show
Placeholder article. This is seed content so the site renders populated. Replace with real research before launch.
Volatility targeting has quietly become the default risk overlay for systematic strategies. Scale exposure inversely to recent realized vol, and most equity strategies show higher Sharpe, shallower drawdowns, and prettier equity curves. The mechanism is real: volatility clusters, returns don't, so de-levering into high-vol regimes trims the left tail without symmetric cost to the right.
What the backtests rarely show is estimator sensitivity. A 20-day realized vol estimate, a 60-day estimate, and an EWMA with a 0.94 decay produce meaningfully different position paths through the same market. In the 2020 crash, the 20-day estimator cut exposure roughly a week faster than the 60-day — worth several hundred basis points — but it also whipsawed through 2019's low-vol grind, paying that advantage back in turnover.
| Estimator | Sharpe | Max DD | Ann. turnover |
|---|---|---|---|
| Realized 20d | 1.34 | −11.2% | 4.1x |
| Realized 60d | 1.21 | −14.8% | 2.3x |
| EWMA (0.94) | 1.38 | −10.9% | 3.6x |
The honest framing: vol targeting is not free Sharpe. It is a systematic short-gamma-like bet that recent vol predicts near-term vol, funded by turnover. Most of the time that bet pays. The article walks through when it doesn't, and why the target level (10% vs 15%) matters far less than the estimator's responsiveness.
the dispatch
Get the next piece of research when it ships. No schedule filler.
want the full notebooks behind articles like this? go premium →