Long VX1: Volatility, Weekends and Overnight Sessions

Update 2023-03-19: this study has been updated with refreshed charts, tables and content after correcting an error with the roll-yield calculations.
Before we jump in to today’s post about long VX1 trades, The Ultimate Guide to Taxes in Early Retirement has been updated with refreshed visuals and calculations for 2023 tax brackets (for filing in April 2024), updated tools for understanding and estimating current and future tax liabilities, and updated tools for understanding how early retirement impacts social security benefits. Check it out if you’re looking to optimize your unique scenario or simply want to refresh your understanding of early-retirement tax mechanics.
In January 2020, we explored the performance of trading SPY during distinct trading sessions (overnight, intraday, price return). The following month we explored the performance of trading SPY during specific days of the week. In this study we combine these ideas and explore the performance of trading something other than SPY: front-month, S&P 500 VIX futures (long VX1).
We begin by acquiring the full set of freely-available VIX futures data, which has a start date of January 1 2013, from CBOE and pivot it into a front-month price series.
On the Friday before contract expiration (eg: three business days, or five calendar days, before expiration), we roll the entire position to the next month. This is to avoid liquidity issues inherent with the last few days of the /VX contract lifecycle.
We then adjust the spliced, rolled price series to account for roll yield in accordance with futures methodology standard procedure (mirror).
The performance of a long VX1 position is then recorded and binned across all 18 mutually-exclusive configurations of session and day-of-week combinations in accordance with the binning methodology.
This backtest assumes that 100% of the account is allocated when a position is open (eg: fractional shares of /VX futures). This of course isn’t feasible in practice and is one of the reasons why ETFs may not always achieve the performance of their benchmark index: there will almost always be a residual amount of unallocated cash. As the portfolio grows and the residual cash becomes a smaller and smaller fraction of the portfolio, deviations that this assumption cause will approach zero. Nevertheless, it’s a reasonable performance approximation that a fund may realize if it implemented this strategy (before expense ratios/fees).
There are 18 backtests in this study evaluating over 15,100 long VX1 futures trades.
Let’s dive in!
Contents
Summary
A long VX1 position held during the weekend overnight session (open at 4:00pm ET on Friday and close at 9:30am ET Monday) outperformed regard to total return vs all other long VX1 strategies.
A long VX1 position held during the Tuesday overnight session (open at 4:00pm ET on Tuesday and close at 9:30am ET Wednesday) outperformed regard to risk-adjusted return and max drawdown vs all other long VX1 strategies.
Methodology
Strategy Details
- Symbol: /VX Front Month (VX1) rolled the Friday before expiration
- Strategy: Long Futures
- Days Till Expiration: N/A
- Start Date: 2013-01-01
- End Date: 2022-12-31
- Positions opened per trade: 1
- Entry Days:
- session
- overnight
- intraday
- pice return
- day of week
- Monday
- Tuesday
- Wednesday
- Thursday
- Friday
- every trading day
- session
- Entry Signal: N/A
- Timing:
- overnight: entry at 4:00pm ET, exit at 9:30am ET the next trading day
- intraday: entry at 9:30am ET, exit at 4:00pm ET the same trading day
- price return: entry at 9:30am ET, exit at 9:30am ET the next trading day
- Strike Selection: N/A
- Trade Entry: daily
- Trade Exit: daily
- Max Margin Utilization Target (short option strats only): N/A | 1x constant leverage
- Max Drawdown Target: 99% | account value shall not go negative
Assumptions
- Margin requirement for short CALL and PUT positions is 20% of notional
- Margin requirement for short STRADDLE and STRANGLE positions is 20% of the larger strike
- Margin requirement for short VERTICAL SPREAD positions is the difference between the strikes
- Margin requirement for short CALENDAR SPREAD positions, where the short option expires after the long option, is 20% of the short option
- Margin requirement for long CALENDAR SPREAD positions, where the short options expires before the long option, is the net cost of the spread
- Early assignment never occurs
- There is ample liquidity at all times
- Margin calls never occur (starting capital is set using hindsight bias so that max margin utilization never exceeds 100%)
- Apply a 20% discount to displayed results. For example, if a strat depicts a CAGR of 10%, assume that it’ll yield 8% in practice.
Mechanics
- Prices are in USD
- Prices are nominal (not adjusted for inflation)
- All statistics are pre-tax, where applicable
- Margin collateral is invested in 3mo US treasuries and earns interest daily
- Option positions are opened at 3:46pm ET
- Option positions are closed at 3:46pm ET (4:00pm if closed on the date of expiration)
- Commission to open, close early, or expire ITM is 1.00 USD per non-index underlying (eg: SPY, IWM, AAPL, etc.) contract
- Commission to open, close early, or expire ITM is 1.32 USD per index underlying (eg: SPX, RUT, etc.) contract
- Commission to expire worthless is 0.00 USD per contract (non-index and index)
- Commission to open or close non-option positions, if applicable, is 0.00 USD
- Slippage is calculated according to the slippage table
- Starting capital for short option backtests is adjusted in $1000 increments such that max margin utilization is between 80-100%, closest to 100%, of max margin utilization target
- Starting capital for long option backtests is adjusted in $1000 increments such that max drawdown is between 80-100%, closest to 100%, of max drawdown target
- Positions that have an exit date beyond the backtest end date are excluded
- For comprehensive details, visit the methodology page
Results
Starting Capital


Win Rate


Monthly Returns
An active subscription is needed to view this content. Already a subscriber? Log in. Not a subscriber? Sign up for just 99 USD a year. Learn more.
Max Drawdown
An active subscription is needed to view this content. Already a subscriber? Log in. Not a subscriber? Sign up for just 99 USD a year. Learn more.
Number of Trades

Compound Annual Growth Rate
An active subscription is needed to view this content. Already a subscriber? Log in. Not a subscriber? Sign up for just 99 USD a year. Learn more.
Annual Volatility
An active subscription is needed to view this content. Already a subscriber? Log in. Not a subscriber? Sign up for just 99 USD a year. Learn more.
Sharpe Ratio
An active subscription is needed to view this content. Already a subscriber? Log in. Not a subscriber? Sign up for just 99 USD a year. Learn more.
Capital Efficiency
An active subscription is needed to view this content. Already a subscriber? Log in. Not a subscriber? Sign up for just 99 USD a year. Learn more.
Total PnL
An active subscription is needed to view this content. Already a subscriber? Log in. Not a subscriber? Sign up for just 99 USD a year. Learn more.
Overall



Discussion
Volatility is generally inversely correlated with its underlying. In the case of VIX, when SPX/SPY goes down, VIX goes up and vice versa. We can see this demonstrated in the chart below, which depicts continuously holding a long VX1 position rolled 3 trading days before expiration.
The VIX futures term structure is generally in a state of contango, where the future-dated volatility (/VX) is higher than the spot volatility (VIX).
source: VIX Central
If a trader was to hold a long VX1 position and roll it each month, they would incur a cost to maintain an identical position size since the future-dated contract has a higher value than the current contract. This cost is essentially an implied fee to hold the position. Here’s the same chart when accounting for roll costs/credits (ie: contango/backwardation).
Despite changes in volatility, the general state of contango accounts for the lion share of PnL over most time horizons greater than 30 days, grinding a long position to almost zero after 10 years.
The question goes: does contango behave consistently across time or are there times when contango is amplified, muted, or reversed?
At first glance, based on the Monday Overnight backtest, it appears that expectations of future volatility increased over the weekends. A possible supporting narrative may be:
When financial markets are closed, life goes on. Uncertainties in the world continue to influence markets by way of systematic risk. A geopolitical event can occur after US-market hours, or over the weekend, and cause concerns for market participants. To express uncertainty about this and other potential risks, long VX1 positions have experienced increases in value over weekends as measured by the closing price of the position on Friday and the opening price of the position on Monday. In other words, the futures term structure is generally in contango when measured across months, but when measured over weekends, at least as far as the front month is concerned, it has generally been in backwardation.
As it turns out, there is a material amount of timing luck in this backtest and the above statement doesn’t quite hold. Just like with options backtests that do not open a new position daily but instead open a single position then wait till that position expires before opening another, the same applies to futures backtests that span multiple contracts (unless otherwise noted in the methodology section, all options backtests here at spintwig use a daily-entry methodology in order to mitigate timing luck).
The methodology in this long VX1 futures backtest uses a roll event 3 trading days prior to contract expiration. Luck can play a material role in the cost or credit associated with a roll event. The roll event could have occurred, say, 5 trading days before contract expiration. Or 10 days before. Or on expiration day. Each roll date choice can yield a different PnL.
What about options? A trader can indirectly trade volatility (/VX) by trading options on /VX’s underlying: the S&P 500. It’s worth noting that trading options on /VX futures would be trading volatility of volatility – ticker: VVIX – which is not the same thing. If a front-month /VX contract has an expiration date 28 days away, an SPX option of matching duration (DTE) would be the proxy vehicle. If attempting to capture the weekend long VX1 performance through S&P 500 options, the one approach would be to open a long put or long put spread at Friday’s close and close the position at Monday’s open.
Trading options to capture long VX1 performance comes with its own sets of challenges and complexities, such as having to be concerned with directionality and magnitude of moves in the underlying, breakdowns or amplifications of correlations, differing margin mechanics (SPAN vs Reg-T or portfolio), differing trade commissions and frictions, etc.
Additional Resources
Private, Custom Backtests
Discover your edge with private, custom backtests for as little as 99 USD. Learn more or contact us for a quote.
Trade Logs
Visit the trade log store and download the data used in this and other backtests.
s1 signal
Trade S&P 500 options more efficiently using the s1 signal. Learn more.
s2 signal
Trade GLD options more efficiently using the s2 signal. Learn more.
s4 signal
Trade Russell 2000 options more efficiently using the s4 signal. Learn more.
mREIT Preferred Share Dashboard
High-yield, low-beta alternatives to cash or treasury bills. Learn more.
Consultations
Schedule a consultation to review your specific scenario and get direct answers to your questions.
February 24, 2023 @ 9:25 am
well well, look at this
must’n everybody and the world be trading this for ages if it’s that obvious?
seems too good to be true..?
February 24, 2023 @ 9:42 am
Similar clams have been made for years about SPY overnight trading. Despite this, ticker:NSPY was released only a few months ago. Give it more time. An ETF for overnight vol may be launched in due time.
February 27, 2023 @ 11:07 am
ETFs are created for the purpose of generating fee revenue for the emittent not for the prospect of making alpha for traders. What is the message of your statement?
March 4, 2023 @ 11:05 am
I knew a guy who used to buy VIX calls on Friday and sell them on Monday morning. As he put it, “Calls for Sabbash”. He didn’t backtest his strat, but I guess it was on to something.
What was your thesis going into this project? Were you surprised with the results?
March 5, 2023 @ 11:20 am
Some of the research starts with a thesis and other studies are purely an exercise in exploratory curiosity. This particular study falls into the latter category.
I’ve been curious about trading vol directly, rather than indirectly / partially through options, and haven’t seen any comprehensive work on /VX to my liking. This seemed like a relatively easy and straight forward series of backtests to explore and, should they be viable, implement.
I’m not sure I’d say there was any surprise. I walked in with a blank slate and no expectations. I had no idea what I was going to find. This is a good first step in exploring trades in this space.
The effect of the roll yield occurring 3 trading days before contract expiration is a key risk re: timing luck. It would be the equivalent of opening a single 30DTE option and rolling 3 days before expiration. I can generate ~20 different PnL curves based on the day in which the roll event occurs.
With options, are two ways to solve for timing luck: open a 30DTE position daily or run 20 different backtests for the same strat then calculate a +/- variance band of performance a.k.a. portfolio tranching (I do the former). For futures, I’m figuring out what the “roll daily” methodology would look like. Not only would there a different roll yield, the effective average duration of the futures position would be different, which is different exposure to /VX futures curve and consequently a different trade.
March 5, 2023 @ 6:14 pm
Yeah I think the model is overfitted.
There must be some kind of real world restrictions to getting these results.
Not saying that there isn‘t an edge there by trading Fr-Mo futures.
March 7, 2023 @ 7:39 pm
There are no models in play that can be overfit (or underfit), just a basic measurement between close-close, close-open, and open-close pricing.
I’m working on a few updates to this post to remove the roll yield effect on the Fri close to Mon open trades. In offline discussions, the roll event which occurs 3 trading days before expiration appears to be materially and improperly influencing the weekend trade stats. Doing this also solves for the timing luck issue I mentioned in another response by avoiding roll events entirely. After the update, daily results will be reflective of the trade without any roll yield effects.
I’ll post a comment and a note at the top of the page with more details once the updates have been made.
March 19, 2023 @ 11:20 pm
Thank you for your patience.
The post has been updated to correct for the roll-yield calculation error. Timing luck remains a notable factor and is mentioned in the updated discussion section.
March 6, 2023 @ 10:33 am
VIX closed (4pm EST) on 3/3/23 at 18.49 and opened on 3/6/23 (9:30am EST) at 19.05.
/VX closed (4pm EST) on 3/3/23 at 19.50 and opened on 3/6/23 (9:30am EST) at 19.40.
When looking at the data from VIX it suggests a profitable trade, but when looking at /VX futures data, it is a loss. I think the products you are suggesting won’t work. There needs to be a way to trade VIX directly, which outside of options is not possible. The issue with /VX futures is that they open for trading at 6pm EST on Sunday, which allows for plenty of time for this trading opportunity to lose its edge. Would you be able to rework this strategy using data directly from /VX futures? Or do you only have access to VIX data?
March 6, 2023 @ 11:10 am
How did you jump to this conclusion just from one data point? The backtest clearly indicates a smaller than 50% win rate. This is completely in line with the strategy’s performance.
Yes, I guess directly “trading the VIX” through SPX options would be possible, but that is not something many could or would like to do, given the high maintenance and transactions costs that come with it.
March 6, 2023 @ 12:09 pm
I’d like to edit my previous comment, rereading your post you do state that you used /VX futures data, not VIX data. When you did your back tests, did you try closing the futures positions out on Sunday at open? Do you think that this would have performed any better? Or would it be better to have held them until Monday at open like you suggest? I’m also curious if purchasing a call option on VIX, like another commenter had suggested, or a put option on SPX/XSP would work as well, worse or better than trading the futures products?
March 7, 2023 @ 8:04 pm
I have not explored a 6pm ET Sunday entry/exit. The data Cboe provides is open, high, low, close (OHLC) for each “standard hours” trading day. Any after-hours price action, such as on Sundays, is not in the freely-available data. However, it wouldn’t be difficult to test if I get my hands on the data.
My purely-speculative answer is that exiting the Friday 4pm position at 6pm Sunday it would improve performance vs opening at 9:30am Monday. Reason being is the illiquidity premium of not being able to exit the position has expired. Any news events can be acted upon by 6pm Sunday and wouldn’t have to wait for a Monday morning exit.
VIX options should be roughly similar in performance if matched to the duration/expiration date of the /VX contract and opened such that delta of the option position is as close to 1 as possible. There will be alternative friction costs of trading options which may or may not exceed those of trading futures, and there will be additional imperfections due to the introduction of greeks for options trading. However, one nuance is that a trader can trade the VIX options using Reg-T or portfolio margin as opposed to SPAN margin (i.e.: a futures account wouldn’t be needed). This may be useful from a cash management optimization standpoint.
Trading on S&P500 products would be a different trade as it gives you delta exposure to the S&P500. Hypothetically, ignoring all the complexities and costs associated with hedging out delta and the other greeks so that you’re left with exposure to just vega, results may still be entirely different since vega doesn’t respond linearly to changes in IV.
March 8, 2023 @ 3:51 am
Forward testing with /VX Minis and Automation API for a year or better two is the best way to find out if this strategy really holds up.