Trading the VIX, the Fear Index: Development and Optimization of a Trading System

In this article, we delve into the concept of volatility and explore intriguing opportunities to use it to our advantage. Generally, volatility refers to the magnitude of price fluctuations relative to their average value. Several tools can measure volatility, with the most well-known being standard deviation. Other tools, such as the Average True Range (ATR), are frequently employed in trading system development, as well as Bollinger Bands.

Regardless of the unique characteristics of these tools, it is essential to note that they all measure historical volatility, meaning they analyze the volatility exhibited by an asset over a specific time frame.

Understanding Implied Volatility

Here, we introduce a new concept: implied volatility. What does it entail? Unlike historical volatility, which looks backward, implied volatility estimates future price fluctuations of an asset. Essentially, it reflects the market’s and market makers' expectations of the asset's movements in the near future.

Technically, implied volatility is derived from the prices of complex instruments like options on the S&P 500, one of the world’s largest and most influential markets. The index that measures this is known as the VIX Index, commonly referred to as the “Fear Index.” This nickname stems from the index’s nature; since it is inversely correlated with the stock market, it tends to rise when increased volatility—and often a market downturn—is anticipated. Conversely, it decreases during periods of market calm.

Figure 1. Historical trends of the VIX Index ($VIX.X) from 2010 to present

As shown in Figure 1, the VIX Index has a unique behavior: it remains stationary most of the time, oscillating around average values, but experiences brief, explosive spikes during market crashes.

Broadly speaking, there are two main approaches to leveraging market volatility. The first involves using implied volatility, such as the VIX Index, as a filter for trading decisions. This is similar to how historical volatility tools—like standard deviation, average true range, and Bollinger Bands—are utilized. This approach aims to limit trading activity, especially in equity futures, to focus on trades with a higher probability of success.

The second approach, which we'll explore in this article, focuses on developing a trading strategy specifically targeting implied volatility. This involves utilizing financial instruments designed to replicate implied volatility.

Trading Strategy on VIX Futures (@VX)

This strategy focuses on the VIX Futures (@VX), traded on the CBOE in Chicago, with the VIX Index as its underlying asset.

As previously mentioned, the VIX Index is characterized by its strong mean-reverting behavior. To test this trait, we designed a straightforward trading system based on Bollinger Bands and applied it to a 30-minute chart. For our analysis, we backtested the strategy from January 2010 to November 2023, using the standard trading session hours (5:00 PM to 4:00 PM).

The trading system follows these rules: a short position is entered when the price breaks below the upper Bollinger Band (“UpperBand”), and a long position is entered when the price breaks above the lower Bollinger Band (“LowerBand”). An initial stop loss of $2,000 is applied to manage risk effectively.

Here's the EasyLanguage code for the strategy:

inputs: MyStop(2000);
inputs: Length( 20 ), NumDevs( 2);
variables: UpperBand( 0 ),LowerBand( 0 );
UpperBand = BollingerBand( c, Length, +NumDevs ) ;
LowerBand = BollingerBand( c, Length, –NumDevs ) ;
if C crosses under UpperBand then Sellshort next bar at market ;
if C crosses over LowerBand then Buy next bar at market ;
if MyStop > 0 then setstoploss(MyStop);
setstopcontract;

Figures 2 and 3 provide examples of trades and the backtest results.

Figure 2. Example of a short trade executed using the “Bollinger Bands” trading system on VIX Futures.

Figure 3. Strategy Performance Summary for the “Bollinger Bands” trading system on VIX Futures.

Upon analyzing the results, it's evident that the strategy seems to work effectively only on the short side. But why does this significant difference in behavior occur? To understand this better, take a look at Figure 4, which shows the long-term charts of the VIX Index and the VIX Futures.

Figure 4. Comparing VIX Index (@VIX.X) and VIX Futures (@VX).

The correlation between the two charts is apparent. For instance, every spike in the VIX Index corresponds to a spike in the VIX Futures, albeit of varying magnitudes. However, one detail immediately stands out: the continuous loss of value in the VIX Futures over time. Why does this happen?

The answer lies in how futures are priced. Since the futures trade at a higher price than the index, in the absence of market anomalies or crashes, they gradually depreciate each day until they converge with the index's value at expiration.

This creates a strong long-term bias, making the strategy ineffective on the long side when employing a mean-reverting approach.

Given this, our next step is clear: we will refine the development of this strategy by focusing exclusively on the short side.

Optimization of a Mean-Reverting Strategy for the Volatility Futures Market (@VX)

First, we assessed whether the default Bollinger Bands settings—20-bar length and 2-period standard deviation—are optimal for the chosen instrument. To identify better settings, we optimized the inputs, varying the “Length” from 10 to 30 bars in steps of 5 and “NumDevs” (number of standard deviations) from 1 to 3 periods in steps of 0.5. Additionally, as a precaution against weekend overnight risks, we implemented a condition to close all positions by Friday evening.

Figure 5. Optimization results for the “Bollinger Bands short only” trading system on VIX Futures.

The results revealed that reducing the length of Bollinger Bands from 20 bars to 15, while keeping the standard deviation at 2, improved performance. Net profit increased from $145,220 to $158,350, and average trade value rose from $160.82 to $172.60. Based on these metrics, we selected these parameters.

Figure 6. Detailed Equity Curve of the “Bollinger Bands short only” trading system on VIX Futures.

Until now, we've operated under the assumption of leveraging the full trading session from 5:00 PM to 4:00 PM. However, let's explore whether a more restricted time window could yield better results. Additionally, we'll examine the maximum number of days a position should remain open for optimal performance, keeping in mind our initial rule to close all trades by Friday afternoon.

Using a proprietary function, we incorporated these time constraints into the system's code and ran a parameter optimization.

Figure 7. Time window optimization for the "Bollinger Bands short only" trading system on VIX Futures.

Figure 8. MaxDaysInTrade optimization for the "Bollinger Bands short only" trading system on VIX Futures.

The optimization of the time window revealed that the most profitable operating hours aren't the standard 23-hour period (5:00 PM–4:00 PM), but a shorter window between 6:00 PM and 2:00 PM. Furthermore, the analysis showed that the ideal maximum duration for holding a trade is five days. Extending beyond this limit provides no significant benefit and, in fact, increases exposure to market risk.

Figure 9. Strategy Performance Summary and Total Trade Analysis after optimizing the time window and MaxDaysInTrade.

At this stage of development, we've started drawing some initial conclusions about the system. We've seen clear improvements in metrics, both in terms of net profit and average trade. However, when focusing on average trade, it's important to note that the current value remains insufficient for the type of instrument under analysis. The @VX Future has a tick value of $50, so for live trading, an average trade of at least $250-$300 is required. Moreover, the maximum drawdown is still too high (over $40,000). Although this figure largely reflects the market shock caused by the COVID-19 crisis in 2020, it's crucial to improve it in anticipation of future market corrections, which are cyclical.

What steps can we take to enhance the performance of our trading system?

One possible approach is to test for specific price patterns. We'll use a proprietary list of scenarios and optimize the system to identify the best-performing pattern. Adding this filter, the optimization process highlighted some interesting possibilities, and we selected Pattern 39 due to its significant impact on reducing drawdown (from $44,160 to $23,970). This choice did involve sacrificing some net profit compared to the top options on the list. Here's the rule for Pattern 39: the difference between today's high and open must be smaller than the difference between yesterday's high and open. Essentially, this pattern indicates uncertainty and compression, suggesting that the market is likely to soon take a more defined direction—in this case, a downward one.

Figure 10. Optimization of the price patterns in the “Bollinger Bands short-only” trading system on VIX Futures.

As a final step, we could implement a Breakeven Stop. This stop triggers when an open position reaches a certain profit level, preventing a winning trade from turning into a loss. While this type of stop doesn't always enhance the overall performance of a strategy, it provides peace of mind for traders, especially in a market like this one. We'll experiment with this alongside optimizing the initial stop loss, which was initially set at $2,000.

Figure 11. Optimization of stop loss and breakeven stop in the “Bollinger Bands short-only” trading system on VIX Futures.

Figure 12. Detailed Equity Curve after optimizing patterns, stop loss, and breakeven stop in the “Bollinger Bands short-only” trading system on VIX Futures.

Figure 13. Strategy Performance Summary and Total Trade Analysis after optimizing patterns, stop loss, and breakeven stop in the “Bollinger Bands short-only” trading system on VIX Futures.

Conclusions: Mean-Reverting Trading System on VIX Futures (@VX)

As we wrap up this article, we can draw some key insights. The latest version of the strategy, which achieved a net profit of $173,300 and an average trade exceeding $240, appears to be a solid foundation for developing a live trading system for VIX Futures (@VX). The mean-reverting logic has proven to be an effective approach for navigating volatility futures trading. Moreover, this strategy has the potential to be adapted for similar instruments, such as ETFs or CFDs that track VIX performance, with the necessary adjustments.

I encourage readers to explore this approach further, test its application, and perhaps refine it to improve results even more. The realm of trading is one of constant experimentation and learning, and this strategy offers an excellent starting point.

Until next time, happy trading!

Andrea Unger