Revisiting VIX Call hedging - 2025 data update and new strategies

Revisiting OTM VIX calls

Back in 2021, I published a report about using OTM VIX calls as a way to hedge a portfolio, and even benefit from market crashes. It’s been five years, so I want to look back on the strategy and update a few points, including:

1. How did the strategy perform in the bull market of 2021-2025? It will obviously underperfom, but by how much?
2. There was a lot of timing luck in capturing the COVID crash. Can that be made more robust?
3. Calling the R code that I used to do the backtest “academic” would be generous, let’s have an agent reproduce the analysis in python.

Refresher on the strategy

The VIX Tail Hedge (VXTH) index is a hypothetical portfolio that buys 30-delta VIX calls with a variable allocation depending on the forward value of VIX:

In my original report, I extended the base strategy by looking at different option deltas, monetization rules, a 30/60/90 day ladder strategy, and more.

Do VIX hedged portfolios survive into 2025?

My VIX option data extends from 2006-2025, so we can backtest a strategy any point in that period. The original report ended in 2021. How do the single call and ladder strategies do over the whole period, and in 2021-2025? First the equity curves, then the statistics.

Growth of $1 (log scale) for each portfolio vs SPX unhedged, in the full period and out-of-sample extension.

Interpretation

The main conclusions are unchanged. The low-delta VIX calls function like a lottery ticket, paying off exponentially during the COVID crash but dragging the most in calm years. 30/60/90 day ladder strategies remain superior to the single-call strategies. I was most surprised by the 50Δ ladder - it was the only strategy that raised the sharpe ratio and returns over SPX. The 50Δ ladder also bled the least during 2021-2025, making it more likely that an investor would actually follow the strategy during calm years when it’s a drag on returns.

Crash payoffs: where does each portfolio benefit the most?

VIX calls are most effective in fast market crashes, such as GFC and COVID. However, the strategies provide little to no protection in “slower” bear markets.

Total return of the SPX + 1% VIX call strategies through select crash windows.

Volatility/return frontier

Here I’ve plotted the volatility and returns of the different strategies over the full and OOS period. This is where the 50Δ ladder strategy really shines. It lowers volatility compared to unhedged SPX in every period. Alternatively, you can use a modest amount of leverage on the portfolio to boost returns while maintaining SPX vol.

CAGR vs annualized volatility of different strategies, toggle the period above.

Measuring and reducing timing luck

The COVID payoff of any VIX call strategy looked great, but how much was structural and how much was luck in when the position happened to be on? Three types of timing are present in this strategy:

• Entry-phase: the day you first put the hedge on.
• Roll-phase: which day inside each option’s life you roll the position.
• The allocation signal: the entire strategy depends on having the hedge in place before VIX spikes.

I evaluated the impact of each type of timing luck across the full 2006-2025 period.

Entry-phase luck: doesn’t matter

Re-running every strategy from 21 staggered start only moved the full-sample CAGR by ~0.13 percentage points across the cohorts.

Roll-phase luck: the ladder earns its keep

This is where it gets interesting. VXTH buys only the monthly option expirations, but weeklies are available too. We can simulate the impact of COVID happening a week earlier or later by buying the weekly expirations instead of monthly.

Shifting the roll by ±1–2 weeks across a window spanning the COVID crash reveals that single 30-day call is almost entirely at the mercy of the roll calendar. The COVID crash payoff swings from +29% to −22% - a 51-point spread - depending on which week it happened to roll. The 30/60/90-day ladder reduces the COVID crash payoff range to 7 points.

The allocation signal: don’t read VX1

The official VXTH strategy sets the hedge allocation off VX1, the front-month VIX future. VX1 decays toward spot VIX as it nears its own expiry, then jumps back up when it rolls to the next month. This disjoint behavior means that the allocation function can whipsaw over the decision boundary for purely technical reasons. This is exactly what happens during the calm period before the COVID crash: reading VX1 one day earlier would have left the strategy unhedged into the crash.

Instead of VX1, I calculated a 30-day constant-maturity VIX forward by interpolating the front- and second-month VX futures on their days to settlement, which I call CMF30. CMF30 represents a point exactly 30 days out, so it agrees with VX1 on the roll day but stays smooth in between instead of jumping across the allocation line. To study this effect in detail, I backtested the 50Δ ladder with the allocation function read ±1–2 days from VX1 or CMF30. The COVID crash returns varied over 34 percentage points when reading VX1, and didn’t budge for CMF30.

Conclusions

The VIX call hedging strategy performs about as well as I would have expected in the 2021-2025 out of sample period. Volatility is probably fairly priced, and the gap in returns compared to an unhedged portfolio in the out-of-sample period is evidence of that. VXTH got lucky in March 2020, but there are a number of improvements that can be made to reduce the impact of luck in a future market crash. Those include using a 30/60/90 day ladder instead of a single date, and reading the allocation signal off a 30-day constant-maturity VIX forward. Finally, the 50Δ ladder strategy is starting to look like the most appealing if you aren’t predicting another COVID-like event in the near term. It bleeds the least in the calm years, and actually increases both Sharpe and returns over the whole 2006-2025 period.

Limitations

Throughout this work I have maintained a few assumptions that change how you should think about this work. In order of possible severity:

1. The performance of the VIX call part of the portfolio is dependent on a very small number of crash events. I’ve put a lot of effort into reducing the timing luck of the strategy, and it should perform in other “fast crash” events in the future, but we have no way of testing that.
2. All VIX call transactions take place at the mid price; I don’t model the spread at all. During calm times, VIX calls are liquid and you can often get filled near the mid, so I expect the impact to be small most of the time. Spreads widen during a crash when liquidity dries up, so the impact will be greater then.
3. Commissions are also not included in my model. They are small relative to the dollar size transacted.

New backtesting repo in python

I’ve updated the vix_hedge repo, which previously held 5-year old R code, to a new python implementation. Index values can be downloaded automatically, but you’ll have to provide your own VIX option series if you want to fully reproduce these results.