We started WU around our Hedge algorithm and then provided other indicators to help navigate the market, like the Bandbreaker for oversold conditions and the VIX Ribbons and their stats for playing volatility (something we haven’t done recently, but I will come back to this soon). Last winter, we added to this list our Option Model, which was the result of an idea I had four years ago to create a predictive indicator that would simulate the price movement of the S&P 500 if the only forces acting on it were the options market activities. When we released the model, here’s what I said about where it fits in WU's indicator portfolio.
For me as an investor, my primary need was to have a hedge signal that would help me protect my capital in a massive drop. While designing such a hedge algorithm, I came to one realization: it's impossible to hedge successfully against any small drop in the market and get a return that outperforms, shielding only our capital from the big drops. The way I had envisioned WU was that we would offer a hedge signal that just hedges the big drops for patient investors, but we would also have another offer that would be more for the active trader. But no matter what I tried, I never really outperformed the original hedge signal in trying to remove all bumps from the market. Ok, no... Actually, I did by something like a 5% extra return over a long period of time, but this almost negligible bonus was coming at the price of a very high volume of transactions and a much lower hit rate.
After developing a strategy around this new Option model, I am not sure it perfectly fits the role of being a proper signal to hedge all the bumps. That being said, I think this could be a wonderful tool that, if used in conjunction with other indicators, could greatly help make the right assessment that it's time to raise cash on some individual stock at the right moment. This would be particularly advantageous for low cap, high beta stocks that usually lead SPY in their movements. Indeed, the main hedge signal is primarily made to avoid massive drawdowns, but in between, there are some playable corrections that could be worth avoiding on some high beta stocks.
Almost a year after creating the first beta version for myself, the option model has become one of my favorite tools to help understand where we are in the market. I'm still surprised by how rarely this model gets it wrong and also in about 70% of the time, it actually leads in and out of a correction, which is incredibly powerful. However, the issue I have with it is its usability. The hedge algorithm is simple: it’s a binary signal—hedge on, hedge off. The option model functions in a similar way, but should I sell everything when it goes red and buy back when it turns green or blue? Data suggests that this wouldn’t be a losing strategy, but it would be far from achieving the hedge algorithm’s returns. So, while the hedge has better stats, it tends to ignore smaller drops. How do we combine them?
One of our members actually looked at the stats of using it in conjunction with the hedge, which was an awesome idea that was on our to-do list. However, such member initiative also highlights that its role in relation to the hedge is not entirely clear. The recent correction underscored this issue, as the option model triggered considerably early. Following it would have put us in a better position for that occasion, but in other circumstances, it could mean hedging a complete portfolio for no real move down at the index level. (I mention this because when it triggered in January of this year, although the index only went sideways, several tech stocks went into a real downtrend.)
To make it more usable, we decided to create a super Risk Index that combines all of our option-related indicators, providing a more progressive indication of risk while maximizing our chances of not missing a drop. No indicator is perfect, and having redundancy in our indicators is one way to protect ourselves. For example, although the option model started to decline before most legs down we had in 2023 and 2024, it fell in sync with the S&P 500 in the recent one. However, we have other indicators that rose early and others that lagged during that correction. On the other hand, the one that was the latest in the recent correction became bearish seven candles before the move down we saw in March of this year. So, since all these indicators are reliable but shine at different moments, why not unite them into a super indicator and give each a vote to maximize our chances of filtering out small corrections from the bigger ones? When I say "each," I mean:
1. Option Model
2. NYSE and Nasdaq Derivative volume indicator
3. Implied Correlation Signal
I know this collegial mode of working, which resembles a Jedi council, wasn't particularly successful in Star Wars at predicting the “dark swan event” that was Palpatine, but that was just a movie. In real life, pooling the decisions of several indicators usually leads to greater wisdom and typically increases the success rate while providing a more nuanced assessment of an algorithm's conviction. In fact, several AI algorithms operate this way. This approach is also very useful in real life, such as in an investment portfolio, where decisions are not always binary and our willingness to hold a given stock can vary from one stock to another.
Now I see you are probably asking yourself : "What the heck is the “Implied Correlation Signal”?
To answer this question, although I’ve referenced the Option Model and the NYSE and Nasdaq Derivative Volume Indicator several times in my other analysis (I'll refer to them again later below), I know I’ve never mentioned the third one. Some of the underlying data from that indicator was used in the hedge strategy, but the reason we never disclosed it (and I’m sure you’ll understand this) is that we don’t want to reveal all the methodology behind our indicators. One of the most time-consuming parts of creating our indicators or strategies was finding the datasets that have predictive value in relation to the market.
I think explaining this model in detail would make us lose focus in this text. Therefore, Zackary Stephenson, who has worked on backtesting and developing a signal from these datasets, wrote a blog "Introducting Implied Correlation Signal" that explains in detail how this indicator works and what its statistics are.
The S&P500 Risk Index
Like we said above, our new WU S&P 500 Risk Index is the superposition of three option-related indicators. These metrics will usually not miss a pullback, but it all comes down to how much they lead or lag into that correction and how often they flag very small moves.
First, there is the Option Model that most of you probably know. This model tends to lead into a downtrend in about 70% of cases. This indicator will never really miss a considerable pullback, but it often flags very small pullbacks, like at the beginning of January 2023. If the Option Model flips to red, our WU Risk Index will jump by 2 points and will retrace 1 point on the green bullish signal and another point when we get the blue Safe signal.
The second indicator is our NYSE and Nasdaq Derivative Volume, which is a statistical analysis of all types of derivatives related to the S&P 500. This indicator, which I have often used in my analysis, will more frequently lag going into a pullback, but in some instances, like in March of this year, it can lead the price action. If the ribbon of this indicator flips to red, we will add 1 point to our WU Risk Index, and we will subtract 1 point when it flips back to green.
The last signal of our Risk Index is our freshly released Implied Correlation Signal, which is philosophically similar to an options-world equivalent of a market breadth indicator, although it doesn’t use any market breadth-related data. This signal almost never leads, but it usually raises a red flag steadily within the first few candles of a real pullback. The algorithm for it to raise a flag is relatively complex, as it involves more than one dataset, which makes it hard to visualize the conditions. Nevertheless, know that a red signal will add +2 points to our WU Risk Index, and a green signal will shed 2 points.
This superposition of signals gives an index that increases in steps. This step-like function made it easy to optimize through our homemade optimizer to define an in-and-out function.
Our optimizer allowed us to find two different optimal configurations, which we will explain in the next section.
You can also view all three underlying models separately by going into the parameters section,
selecting the drop-down list, and choosing which one you want to view.
The default view is the WU S&P500 Risk Index.
One other thing I like about this new index is how it complements our Color Signal of our Margin Risk Indicator. So far, the Color Signal of that indicator, which is the sum of six different signals explained here and not at all related to any of the one in our new WU Risk Index, has performed exceptionally well in highlighting the level of overheating in an uptrend. Although it would be hard to time the market with this, similar to how the CNN Fear and Greed Index can remain in greed for a long period, the Color Signal of our Margin Indicator can remain at 14 or 15 for an extended time. However, it is useful for knowing how overheated we are and maybe at least restraining ourselves from buying when it is in the red and also deleveraging our portfolio. That being said, as soon as a correction starts, this indicator falls rapidly to a null reading. This is where our new Risk Index picks up and starts to rise (with some overlap), providing guidance during a downtrend.
I am just starting to work with this new indicator, but I already anticipate that I will probably use these two in tandem as they seem to complement each other. I should have more to say on this topic in future post.
The Results
This simple method of giving a vote to all indicators allows for building a straightforward strategy that is difficult to overfit due to the near absence of tuning parameters. We backtested this index from 2017 to 2021 and validated it from 2021 to now. We couldn’t go much further back in time, unlike with the hedge strategy, for two reasons.
The first reason is that, although the underlying indicators like the option model provide readings further back in time, some of the data used in these models only started becoming available around the beginning of 2017. Therefore, the earlier history is only based on partial data. The second reason is that, while I don’t think market behavior at the index level has changed significantly over the last 20 years, the options market has evolved considerably. Previously, this market was mostly reserved for professional traders and investors, whereas now many more retail investors use it for both investing and speculation. At the NYSE, the volume of retail traders doubled between 2010 and 2020, reaching 48% at its peak during COVID, and it is currently down to 45%. Despite the shorter backtesting period, the higher number of instances where this strategy triggered makes it relatively similar in terms of statistical validity to the hedge strategy.
Looking at that period and trying every possible combination, we found two different optima that proved to have very similar results in our validation period. While we have a preference for one of these optima for reasons we will explain, one optimum is exiting when the index reaches a value of 3 and re-entering when the index goes back to a value of 1. This strategy yielded a higher return, with $2,678 of profit on $1,000 invested in 2017, beating the buy-and-hold strategy by 169%. This strategy generated 43 trades during that period, with a hit rate of 72% and an asymmetric average win to average loss ratio of 3.5.
The other optimum was exiting the market when the index reached a value of 4 and re-entering when it dropped back to 1. This strategy provided a slightly lower return of $2,598 of profit on $1,000 invested, but it generated about 25% fewer trades with an outstanding hit rate of 78%. Here are the complete stats.
As an investor with a strong risk tolerance, but also as someone who prefers trading less since exiting and entering the market can sometimes be stressful, I have a considerable preference for strategy no. 2, where we exit at a value of 4. By default, the strategy will operate in this mode, but I also added the option in the indicator parameters to change this threshold in case someone would like to use strategy no. 1, which will likely hedge the smaller bumps.
Beyond the statistics, the combination of all these indicators seems to be very robust during market bounces. It’s actually one of the first strategies I’ve seen that navigated the correction in fall 2023 without much flaw. It successfully ignored all the bounces, although it didn’t catch the absolute bottom since the Option model led considerably out of that drop.
It also did a pretty decent job in the March-April correction of this year, as well as the most recent one.
There is one area where it doesn’t perform incredibly well, and that’s in the 2022 bear market. It’s not doing completely bad either. That being said, I’m not surprised, as no data-driven strategy really excels in such an environment, which is why our hedging strategy has a bear market mode. It’s like when we enter these bear markets, nobody knows what the market’s next move will be. Even the Dark Pool, which tends to sell at the top in a bull market, starts selling at the bottom, and options also make incorrect bets. Out of all the strategies I’ve tried in bear markets, the only thing that really works is momentum, which is why we had great success riding it in 2022 using mostly the phase angle.
This weakness in our new Risk Index is actually something that makes me incredibly optimistic, as we already have the tools to handle these periods where it doesn’t perform as well. This is why I’m really looking forward to combining this new Risk Index with our hedge algorithm. This will definitely take some time, as both strategies involve a considerable amount of code, but we are working on it.
Conclusion
This text was longer than I intended, so here’s a brief recap in case you got lost:
We created this Novel ~WU SP500 Risk Index to give us an objective way to act on our existing models and indicators, which tend to be accurate. We also made its output variable to help strengthen our conviction when selling. We believe it’s particularly useful for a diversified portfolio made up of stocks with different betas, where the conviction of holding varies from one stock to another.
This index uses three of our option-related models and indicators, each analyzing a different aspect of the market. One of these, the Implied Correlation Signal, was just released, and we believe it is highly correlated with market movements. We use a scoring mechanism, where each of these three indicators raises or lowers the resulting Risk Index depending on their state.
The output signal is proportional to the risk seen in the options market for a real pullback. It can be used as-is, but we’ve backtested strategies that suggest exiting when the score reaches 4 and re-entering when it cools down to 1 is the most efficient way to play this signal. A variant of this strategy involves selling at 3. While this variation yielded slightly higher profits in the past, it required about 25% more trades, which could be useful for investors looking to hedge more volatility in the market.
Comments