VIX-TA-Macro-MP Extreme: Advanced Trading Signals for Market Timing

Algorithmic Trading Model Description

Overview and Implementation

For maximum responsiveness with 23/5 trading, this model exclusively trades the S&P 500 Futures Contract. However, it can be implemented fairly accurately through any ETF or other instrument that tracks the S&P 500 such as SPY or VOO. The model is always either 100% long SPX or 100% cash. Each trade alternates between these two positions. In practice, many investors implement our models through a hedging system instead. For example, on a buy signal, you would hold your regular diversified portfolio of ETFs and stocks as you do now. During a sell signal, you would sell S&P 500 Futures contracts against the value of your holdings for a net market neutral overall position. This method has the added benefit of lower overall tax liability.

Trade Frequency

VIX-TA-Macro-MP Extreme is an average frequency trading model with 210 total trades over the 15.65 year period. On average, the model made one trade per 18.78 trading days. However, the trades are not uniformly distributed; during periods of higher market volatility, more trades were made compared to relatively calm periods.

VIX-TA-Macro-MP Extreme is best for medium/long term investors and swing traders who want to both avoid large drawdowns and substantially outperform the benchmark. At 81.43% trade win rate, traders following this model will occasionally run into false positives. Think of these as insurance premiums that are well worth paying to avoid the potentially massive drawdowns that are bound to happen from time to time (such as the -36.01% thrashing the SPX took during the COVID-19 crisis).

Compared to our VIX-TA-Macro Advanced model, VIX-TA-Macro-MP Extreme makes a few more overall trades while substantially improving both CAGR and max drawdown. It's important to realize that while VIX-TA-Macro-MP Extreme may have around the same number of total trades per year as some of our less advanced models, if you look closely, you'll see that there is a major difference in when each model chooses to trade. This model is not merely VIX-TA-Macro Advanced with some tacked on extra trades; it has an entirely different evaluation system in place to better pick times to execute trades.

Key Indicators

This model focuses on the Cboe Volatility Index (VIX), an index that represents the market's expectations for the relative strength of near-term price changes of the S&P 500 index (SPX), the overall macroeconomic picture, current monetary policy and several TA indicators.

Indicator Overview

A few of the indicators this model uses behind the scenes are:

1. Absolute VIX level
2. VIX futures curve shape (degree of backwardation vs contango)
3. Realized volatility vs implied volatility trend
4. VIX sensitivity
5. Derivative analysis of change in VIX futures structure
6. Bollinger Bands
7. Chande Momentum Oscillator (CMO)
8. Unemployment rate
9. Continued Claims Seasonally Adjusted
10. Several housing market health metrics
11. Current EFFR and implied future EFFR
12. Yield curve trends
13. Fed balance sheet trends (degree of Quantitive Easing or Tightening)

The model uses these and other minor indicators under varying and adaptive time parameters to identify medium-term trend changes. As with all our models, we run the backtest through our proprietary anti-overfitting machine learning system to ensure we are identifying real and repeatable patterns.

VIX-TA-Macro-MP Extreme is not merely a combination of our VIX-TA-Macro Advanced model with added monetary policy indicators. Even though it uses many of the same core indicators, it is much more sophisticated in how it applies them in decision-making. This sophistication leads to substantially better performance on both an absolute and risk-adjusted basis.

Indicator Analysis

Many savvy traders have been using the VIX to their advantage in identifying local minima and maxima on the daily or weekly charts for a long time. VIX-TA-Macro-MP Extreme uses several derivative analyses of underlying VIX movement. Some examples:

• SPX and VIX divergence. Most of the time VIX and SPX are inversely correlated. The rare times when they are not (i.e. both SPX and VIX increasing or both decreasing), is a tip to our model that we may be near a trend shift.
• VIX sensitivity to movements in the SPX. Usually the VIX moves inversely to the SPX with a sort of beta multiple of around 3-5. i.e. if the SPX is down 1% for the day, VIX will usually be up 3-5%. When the VIX responds well outside that normal range (i.e. VIX up 10%+ or flat), that's a sign of a potential trend reversal on the horizon.
• Shape of the VIX futures curve. Traders of VIX futures are well-known to be very sophisticated market makers and institutional investors that are making intelligent plays over the long run. 85% of the time, the VIX futures curve is in contango which means that longer dated contracts are priced higher than nearer dated contracts. This make sense intuitively as just like option premium, the longer the time between now and expiration, the greater risk of some disaster upending the system. 15% of the time, the VIX futures curve falls into backwardation. This is when nearer dated contracts are priced higher than longer dated contracts. This signals to the model that a near-term bottom is likely imminent, as sophisticated traders believe that the current risk event does not have lasting power.

Technical analysis (TA) is a trading discipline used to identify trading opportunities by analyzing statistical trends gathered from trading activity, such as price movement and volume. In aggregate, TA is about analyzing the psychology of market participants through their transactions. Some examples of TA used in this model:

• Bollinger Bands with RSI and SMA synchronicity on multiple time frames. Bollinger Bands (BB) draw two lines around a simple moving average (SMA) based on a positive and negative standard deviation. During times of volatility, the bands widen and then contract when volatility subsides. When price moves well outside the bands in either direction, it can be a a warning sign that the price action has been unsustainable and that price is likely to mean-revert back inside the bands soon. Combining these extreme movements with the relative strength index (RSI) as well as where the current price rests in relation to important SMAs (such as 20 day, 50 day and 200 day) can be a tip to the model that we are likely near a local minima or maxima.
• Chande Momentum Oscillator (CMO) uses momentum to identify relative strength or weakness, classifying such periods by degree of overbought / oversold. By itself, CMO usually does not generate fantastic signals, but when combined with RSI, SMA, BB width and other TA indicators and analyzed across multiple time frames for synchronicity, it can be a valuable tool in guiding the model to certain signals.

A macro strategy is one that bases its holdings primarily on the overall economic and political views of various countries or their macroeconomic principles. By themselves, macro indicators are useful in identifying upcoming recessions with a broad time span, but they really start to shine when used in combination with sentiment and the TA and VIX factors we've discussed above. Some examples of how we apply macro indicators:

• The unemployment rate is the percent of the labor force that is jobless. When this ratio is rising, and especially when it its second derivative is also rising over multiple timeframes, it is a strong indication that the good times are coming to an end and a recession is likely to be around the corner. By itself, this is not a good timing indicator because the timeframe suggested is too broad. However, it assists us in an input to our proprietary market "zone" classification system. These zones help us apply more appropriate heuristics to TA, VIX and other data to avoid false positive signals.
• The health of the housing market is reported by metrics such as housing starts, new home sales, construction spending, Shiller home price index, etc. This model combines a multitude of these metrics into a proprietary real estate health score. Like unemployment and continued claims, this score is used as an input in our market "zone" classification system described above. Historically, housing market health has had strong predictive power in predicting imminent bear markets such as the catastrophic Great Recession that led to a 58% drawdown of the S&P 500. Although our historical trades begin in April 2009, we've built our proprietary real estate health score with decades of data; so you can be sure we'll be prepared if a crisis like this ever comes again.

Monetary policy is a set of tools that a nation's central bank has available to promote sustainable economic growth by controlling the overall supply of money that is available to the nation's banks, its consumers, and its businesses. Since the Great Recession, the U.S. central bank, the Federal Reserve (Fed), has been using more experimental and controversial policies to support the economy such as Quantitative Easing and Zero Interest Rate Policy (ZIRP). Since the dawn of this era of experimentation, market movements have been far more correlated with monetary policy than in the past, making it one of the most important tools in this model's playbook. A few examples of how we apply monetary policy in the model's signal generator:

• First and second derivative movement of the Fed's balance sheet. As part of its Quantitative Easing (and occasional reverse known as Quantitative Tightening that we last saw in 2018), the Fed's balance sheet grows and shrinks dramatically depending on the feelings of a small group of bankers. This expansion or contraction is a key indicator in classifying the current market "zone", described above. Correctly identifying the zone with more sophisticated parameters helps this model outperform others by applying new and better heuristics on a more specific basis.
• The yield curve is a line that plots yields (interest rates) of bonds having equal credit quality but differing maturity dates. The slope of the yield curve gives an idea of future interest rate changes and economic activity, and like many macro indicators, is a powerful predictor of upcoming recessions. This model takes a variety of yield curves such as 10-year to 3-month and 2-year to 3-month and analyses not only whether they are normal or inverted, but also the degree and pace of any such inversion. Two yield curves may both be inverted and signaling recessions, but one may be rapidly normalizing, while the other is roughly flat when looked at from a second derivative perspective. These two indicators would have very different implications in the model's signal generator. Its this kind of complex analysis of commonly used indicators that gives our model its edge.

When building the model, we took these many others into consideration and carefully tested the patterns using machine learning and heuristics to distinguish those with predictive power against those that merely seemed like they had predictive power. This is what we call our proprietary anti-overfitting engine, and although no model can ever perfectly predict market movements, with this engine you can rest easy knowing the model is backed by the latest science and engineering. We leave human emotion at the door; the signals are generated 100% mechanically.

Performance

VIX-TA-Macro-MP Extreme achieved an annual CAGR of 51.80% vs. 13.56% for SPX which we use as the benchmark. This represents a substantial outperformance of 38.24% per year. $100,000 invested in the model at inception would have grown to $56,633,009. Following the rules of all investment professionals, we must of course disclaim that past performance does not guarantee similar results in the future. However, our testing period covers many different market conditions, from the uber turbulent COVID-19 drawdown in Feb-Mar 2020 to the incredibly complacent 2013 and 2017 years. This lengthy period of analysis, combined with our proprietary anti-overfitting system, gives us confidence that our model has alpha.

Drawdown

Historically, the model had a maximum drawdown of -16.89% vs -36.01% for the SPX. This represents a substantial reduction in drawdown vs. the benchmark of 53.10%. When looking for a great model or market timing system, it's very important to not only consider absolute performance but also drawdown because investors should value risk-adjusted returns above absolute. The model's combination of absolute outperformance and smaller drawdown is exactly what savvy investors should be looking for in an actionable market timing system.

Get Started Free