In the beginning….there was Options Premium
This is the 1st in our “Journey to Vol Trading” series that chronicles the evolution of our trading. Keep reading for updates.
Our trading journey actually originates in the early 2000s, starting with a traditional stock trading approach. Very quickly, we decide that we want to focus on repeatable patterns that we can model (at that time with simple math and an Excel spreadsheet). These patterns should not be technical, but rather based in some underlying fundamental behavior that drives the technical charts.
It didn’t take long of course before we discovered options, the “greeks” (variables which model how options price changes) and immediately focused on vega, which measures the change in option price based on a change in the IV (implied volatility).
Out of all the greeks, vega is the only one that is arbitrary. It depends exclusively on IV, which by its very definition (“implied volatility”) is derived from the price of the option itself. When we factor out, from the option’s price, the other known price factors (such as the current stock price, risk free rate of return and days to expiration as per the Black-Scholes model) we are left with IV. In simple terms, the only real difference between the same options on one stock and another is each stock’s IV.
The first mathematical observation that becomes evident from studying vega is that different stocks can have radically different levels of IV / vega. This is generally a function of whether the stock is volatile (eg TSLA) or stable (eg T). The second observation is that most stocks have a base level of IV that stays fairly constant, but then experience large changes in IV around specific events, most often earnings.
The origin of our first trading style then became the concept that IV becomes elevated before earnings (due to possibility of large moves up or down), and then falls significantly right after, when the “surprise” of earnings has waned. This was the first repeatable pattern that we could model, and led to our first mistake.
Mistake #1: Options on Individual Stocks
Armed with our brilliant trading idea, we immediately begin selling straddles or strangles (short put, short call) on stocks a few days before earnings. This turns out to be very profitable. After all, we are siding with the house (market makers) who are pricing the options’ IV such that the actual move after earnings is unlikely to be outside the straddle’s net credit. We quickly generate returns, taking a small $25k account to $55k in just a few months. Then, the inevitable happens, a stock (the ticker RMBS will forever be etched in our memory), moves well out of the straddle range, resulting in massive losses. In retrospect, it was a unique situation with high IV due to a pending lawsuit, rather than earnings, and didn’t actually invalidate our thesis. Yet, the potential of a single stock behaving well outside of the modeled range takes us back to the drawing board.
Our theory turns to indexes, which are naturally insulated from single stocks’ erratic movement. The idea still centers around volatility. In the S&P 500 index, volatility tends to revert to a baseline, since the overall trend of markets is upwards over a long enough timeline. We begin formulating and testing index based trading in late 2014, with live trading starting in early 2015.
Just as we sold volatility before earnings on individual stocks, our mathematical minds now turn to a similar concept but for the S&P 500 index. The premise is simple, sell volatility when volatility is high. When the market goes down and VIX spikes, selling puts and calls becomes very profitable. If the market continues to fall, rolling short puts further down or out (in time) and recapturing the repurchased premium results in successful trading. While stress during downturns, such as August 2015, is very high, the strategy performs well and recovers as expected. Yet, it’s not the downturns that end up hurting us, which leads us to our second mistake….
Mistake #2: The Market Usually Goes Up
While our option premium strategy does well in choppy, and even downwards markets, we quickly run into a major issue. Historically, markets spend about 80% of the time in uptrends, a normal economic fact considering overall growth trends. As market trend up, volatility declines sharply and stays low. As a result, option premium selling becomes challenging. Call options that are sold routinely go ITM (In the Money), requiring them to be rolled up and out in time, delaying profits.
On the other hand, Put options, while usually expiring worthless, become extremely dangerous in any market pullback. The coupled effect of delta (increase in Put price as underlying S&P 500 price clines) and vega (increase in Put price as volatility increases, which it almost always does when S&P 500 price clines) result in outsize losses that cannot be mitigated by the typical risk management strategy.
We pivot once again, with a focus on a strategy that can adequately model market uptrends, as well as exit and shift to a long volatility position in market downtrends.
Our initial work, which we will discuss more in detail in the next post, focused on a very simplified interpretation of volatility, as expressed by VIX futures pricing versus the VIX itself. Futures, just like options, expire and settle at a price relative to the underlying. Thus if currently the VIX is at 30, and the next month’s future is at $25, if nothing changes the future will rise +$5 and settle at $30.
In simple terms, the difference between the VIX futures current price and the current VIX is the major determinant of VIX futures future price. When VIX Futures < VIX (referred to as backwardation) , we buy VIX futures and when VIX Futures > VIX (referred to as contango) we sell the VIX futures.
This can be modeled by the VIX/VXF ratio calculation which we used to give us our initial strategy signal.
vix_vxf_ratio = vix_futures_price / vix_price
if (vix_vxf_ratio > 1):
trade_direction = long_vol
if (vix_vxf_ratio < 1):
trade_direction = short_volThis simplistic view, unfortunately, leads us to our third mistake…
Mistake #3: Volatility Often Reverses Quickly
In it simplest form, our initial long/short volatility trading strategy assumes that the current situation (backwardation or contango) will persist. This is often true, but when it is not, the reversals can be staggering and difficult to manage.
Two scenarios illustrate this clearly:
- The VIX as at 10, a low number and the futures are clearly in contango. Over several weeks, the futures decrease in price towards the low VIX. All of a sudden, market uncertainty causes the VIX to increase, but VIX futures do so as well. While contango lessens, futures increase substantially before flipping to backwardation. We experience significant losses before we get the signal (ratio > 1) to exit the short vol trade and go long vol.
- The VIX is at 30, a typically high number. Futures are in backwardation, expected to keep rising towards the current VIX. The market experiences a rally and fundamental fears that drove the high VIX quickly subside. The VIX crashes quickly down to 20, well below the previous price of the futures. The long VIX Futures position is at a loss.
After a year or so of trading the original simple version of the system, it became clear that we needed to refine it. With some optimization and backtesting, we were able to identify a ratio value (below the original ratio of 1:1 that we had been using) that more consistently predicted short volatility. From 2016 onward, this resulted in consistently profitable trades in periods of low volatility, but did poorly in years like 2018 where markets were choppy and there was very few period of a month or longer where volatility was constantly decreasing. We knew that a long volatility model was necessary to capture those years where markets are erratic.
Unfortunately, the goal of modeling long volatility was extremely difficult because whatever ratio we were using either missed out on the big VIX jumps, or went long vol too often, resulting in a long series of small losses that dragged down on performance. This was in the end simply the result of our fourth mistake…..
Mistake #4: Using one-dimensional algorithmic trading
Although we were able to trade profitably for several years on the short vol side, and even occasionally capture some long vol moves, we knew that there were opportunities to improve. The biggest problem we quickly identifeid was that using the VIX/VXF ratio was simply too one dimensional. It did often predict sustained moves over time, but performed terribly in markets without direction. It became clear that we needed to look at other factors affecting the markets to make more consistent predictions.
At this point, quickly ingesting, testing and eventually using a large number of such factors became impossible with traditional algorithmic trading. The next step in the evolution was naturally Machine Learning. Our team already had experience in data analytics for large datasets in various industries and we had immersed ourselves in ML research for several years.
We will delve into these ML models in subsequent posts, but first, in the next series, we will discuss the details of the volatility market to provide the core foundational knowledge required to effectively trade it. Stay tuned!
*** As always, we make it very clear that this is not investment advice, none of us are allowed to provide investment advice, this is research work product only and any trading decisions should be made on your own based on your own research and trade ideas **********
