Logica Capital commentary for the month ended September 30, 2023.
Summary
Markets retreated in September, and with that came a gain in VIX/Implied Volatility (“IV”), albeit not as vigorous a gain as we would expect starting from such lower levels and in the context of historical S&P declines of similar magnitude. In relative terms, the IV move was uneventful and fairly muted. On the equities market side, it was generally a broad drawdown, grabbing most sectors and indices for the ride, inclusive of the technology darlings that led earlier this year. The exception was the Energy sector, which generated the only positive return in September, and which was one of only two advancing sectors in Q3 overall (with Communication Services being the other).
Q3 2023 hedge fund letters, conferences and more
1) Returns are net of fees and represent the returns of Logica Absolute Return Fund, LP and Logica Tail Risk Fund, LP, respectively. Past performance is not indicative of future results.
2) Naïve Straddle Return: a 1.5 month out, S&P 500 at-the-money put and call bought on the final trading day of prior month and sold on the final trading day of current month. This return on premium is divided by a factor of 6 to be comparable to Logica’s typical AUM-to-premium ratio. For illustration purposes only.
3) Naïve Ratio Straddle Return: a 1.5 month out, S&P 500 at-the-money put and at-the-money call (divided by 2) bought on the final trading day of prior month and sold on the final trading day of current month. This return on premium is divided by a factor of 6 to be comparable to Logica’s typical AUM-to-premium ratio. For illustration purposes only.
4) S&P 500. The index measures the performance of the large-cap segment of the U.S. market. Considered to be a proxy of the U.S. equity market, the index is composed of 500+ constituent companies.
5) The Nasdaq-100 is a stock market index made up of 101 equity securities issued by 100 of the largest non-financial companies listed on the Nasdaq stock exchange.
6) The Dow Jones Industrial Average is a stock market index of 30 prominent companies listed on stock exchanges in the United States.
7) The Russell 2000 Index is a small-cap U.S. stock market index that makes up the smallest 2,000 stocks in the Russell 3000 Index.
The Portfolio: Looking Inside
Commentary & Portfolio Return Attribution
8) For illustration purposes only. Attribution returns are composed of daily returns, gross of fees
“Timing is everything. That’s a cliche now. If I’d said that a long time ago, it would’ve been better timed.” - Demetri Martin
Our strategies in September faced two independent disappointing outcomes; one from the macro-overlay, and the second, from the results of our core models, i.e. the buy/sell timing across our Delta and Vega scalping. While our average Delta tilt and Vega exposures were in line with our long-term average exposures, the day-to-day scalping underperformed our average positive expectancy, which can unfortunately occur over short time frames. While we came into September with a heavier than normal Vega load and lighter Delta load (i.e. more “short the market”), as the month moved forward and the S&P inched downward alongside IV mildly climbing, our models called for mean reverting trades. That is, adding Delta (for the S&P bounce) and unloading some Vega (for the IV pull-back). But then late into the month, as the S&P fell further, and the IV extended higher, our increased Delta and lighter Vega levels caused more pain.
While we have often talk about low IV being a coiled spring, and we target loading up as the spring contracts, there are standard components to our models that forecast “more of the same,” i.e., a persistence of realized volatility which suggest mean reversion is more likely as it begins its spike. Even in the face of a relatively low level of IV, if the Vol market is not signaling a continuance, the higher probability bet is reversion. Broadly, our daily scalping process relies on the general mean reversionary tendency of Vol markets, a well-established and widely papered phenomenon. This is precisely the mechanism we depend on to carry the otherwise costly straddle through “normal” times.
Of course, if there were a material event that triggered breakout signals (such as the outsized magnitude moves or gaps in the S&P and/or IV that have so often preceded a stress event), our models are built to halt their mean reversionary process and lock down in preparation for an expansion. But September did not reveal any such anomalous behavior. In fact, the methodical and tempered step down in the S&P through September 20th, alongside the somewhat gentle climb of IV during the same period, indicated quite the opposite. But as markets are wont to humble us, our model’s reversal timing couldn’t have been worse, as we saw the S&P fall further, and IV start to rise more aggressively on September 21st, and into the end of the month.
For a visual look at how things played out, see the grids below (with some of the “worst” points having occurred later in the month):
“The facts are always friendly, every bit of evidence one can acquire, in any area, leads one that much closer to what is true.” - Carl Rogers
While September is certainly disappointing, a recent deep dive into our historical scalping performance leaves us highly encouraged by the fact that both our Delta and Vega scalping systems have clearly demonstrated positive expectancy (hit rate times magnitude) over time, providing both a respectful hit/win rate at 53.8% and larger wins than losses, at 0.60 positive skew. Most importantly, this alpha has improved over time. Yes, it ebbs and flows, as any statistical process naturally does, but confidently, it flows more than it ebbs. To this end, not only are we not perturbed by September’s result, but given our confirmation of the long-term positive trend of persistent alpha from our scalping, we have instead committed to increasing our exposure to these models in exchange for reduced allocations to what appears to be a more clearly deteriorating exposure – namely, our Macro Overlay (more on this below).
After extensive analysis, we have taken the 3rd reduction in exposure to our Macro Overlay, marking our most significant reduction in this module over the past few years. In particular, the story of the interest rate headwind and its adverse effect on Long-Term Treasuries is well known by now. For decades prior, this asset has provided somewhat of a “free Put”, that is, a costless or even positive carry asset, alongside a somewhat reliable performer in times of crisis. But in a different interest rate environment and era of inflation, the former is certainly no longer the case. And while we can still make a good case for the latter -- that “flight-to-safety” may prevail during material stress events -- the greater likelihood of headwinds the rest of the time, and the greater unknown around when inflation subsides (and the Fed’s related decision-making), poses too much uncertainty around the carry cost.
Similarly, Gold has failed to provide negative correlation to S&P’s declines for some time now. In fact, in September, it dropped heavily alongside the S&P. We hypothesize that some of the disinterest in Gold may be tied to the new interest in cryptocurrency as a disconnected storage of wealth and efficient method of global exchange. Will Bitcoin replace Gold as a safe haven? Or at the very least, is money flowing into Crypto in lieu of the “old school” shiny metal? Maybe. But hypothesizing aside, we clearly observe a deterioration in Gold’s conditional behavior over the last few years, and so whatever the explanation, we see a rational decision to make given the consequences of negative carry cost.
“The question to ask when you look at security is not whether this makes us safer, but whether it’s worth the trade-off.” - Bruce Schneier
More importantly, cost of carry is not in a vacuum, as the tightly related issue ties to reliability (i.e. basis risk). From this standpoint, the macro risk-off assets require a “correlation” assumption; one must believe that they will pay off during crisis. In stark contrast, S&P Put Options are of course perfectly reliable -- without question. Understandably, the original benefit of utilizing positive carry risk-off assets as a “Put Proxy” was a fair trade-off, they were far cheaper to carry – even profitable -- but they came with some basis risk. Said simply, income + strong potential was an acceptable exchange for the known bleed of pure long Vol + absolute reliability. But now we’re at a crossroads; after a multi-year decline in IV, Vol itself has gotten “cheap enough” to enable carrying larger inventory, and especially given our empirically confirmed positive edge in Gamma/Vol scalping. In an ironic twist, at these recent IV levels, we can carry more Put options at potentially lower cost than the recent “bleed” of flight-to-quality assets!
Observed in this light, the choice to shift from potential negative correlation to the wholly reliable negative correlation and convexity of S&P Put options becomes obvious. Given the performance of our Delta and Vega scalping systems overcoming the bleed of safe haven assets, the pure Put clearly beats the proxy Put. Accordingly, we are excited to reduce our basis risk in the Macro bucket for a similar scale up in the Vol/Gamma trading side of the portfolio.
All that said, we still recognize that there is a place for these assets. To some degree, and as regularly demonstrated across 50+ years of historical behavior, we believe that they will still provide some flight to quality behavior during crisis events. The problem, as we’ve outlined above, is of course the cost of holding them alongside the more prominent basis risk of a changing economic environment while we wait for a crisis, and particularly, relative to a fully reliable Put option at now cheaper pricing. But this last point is the key differentiation; the cheaper IV pricing makes the trade-off clear, but given a potentially higher IV hurdle to overcome, such as IV levels in the 30’s and above, the trade-off starts to look interesting again.
To this end, we will continue to utilize the Put proxies as a replacement down-capture asset as we monetize our Put holdings into the heat of a stress event, i.e. as IV climbs to higher heights. In the most extreme of examples, at the top of the ’08 GFC or ’20 Covid-19, IV approaching the 100 level was highly unattractive to own on the long side, and per the monetization methodology of our scalping process, we will have stepped our way out of significant inventory along this “spiking” path. But in still wanting to hold defensive positioning, we will leg into the safe haven assets to fill the void. Specifically, and as is already built into our model, we will continue our process of adding the replacement as IV levels get increasingly overpriced, or more importantly, as IV “crush” becomes a greater risk. Thus, the macro assets will still serve the Put proxy function, but only when the relative pricing screams loudly in favor of the exchange.
"A world which sees art and engineering as divided is not seeing the world as a whole.” - Professor Sir Edmund Happold
Finally, our Sector & Single Stock Calls kept pace with the S&P 500 in September, even slightly outperformed. While wholly acceptable, and another confirmed source of alpha over the years, the level of results of this module have also been the subject of evaluation over the course of 2023. As the core model relies on longer term trends (in momentum and anti-momentum), there is a reasonably wide cycle of over/under performance; it takes time to “play out,” so to speak. Over the long run, we clearly see alpha, but we have encountered enough ebbing periods to question the cycle width as it relates to basis risk. While 2022 saw strong excess performance with an overweighting in Energy, 2023 has seen some underperformance with an underweighting in Technology. Accordingly, we are currently researching optimal trend look-back windows as well as expanding/contracting these exposures relative to their cyclic behavior.
In prior letters, we have discussed (and explained) our thesis on the focus on long term trends, targeting a “mean expansionary” element in the portfolio to counterbalance the day-to-day “mean reversionary” focus of our scalping models. In this framework, our research found that short term momentum resulted in too much “whipsaw” – and would often exhibit mean reversionary characteristics. Concurrently, our approach has generated considerable alpha over the years. We have thus stuck with longer term trends for quite some time. However, as mentioned earlier, we are now identifying concerns related to the breadth of the trend cycle’s trough. This observation is not made in isolation; instead, it must be considered within the broader context of our primary emphasis on defensive positioning and the basis risk posed by broader cycles when seeking to maintain a constant state of readiness for defense.
The questions, conceptually, become: 1) how much alpha is provided? 2) how wide are the time cycles/periods that such alpha is above/below expectation? And 3) to what degree of above/below the average expectation does that alpha vacillate? And all of this, to be clear, is with the goal of constantly being in defensive mode. If, for example, a model can trend below its average positive alpha, at some uncomfortable magnitude, and do so at random precisely during times that the broader market is drawing down, then it poses a much larger risk of “is the longer-term alpha worth the momentary pain at, potentially, the worst of times?” When the primary goal is to defend, alpha cannot be viewed in a vacuum. Its “wavelength” and “amplitude” become critical characteristics to calibrate.
We thus continue evaluating the merits of this alpha engine (i.e. its long-term outperformance and its cyclical behavior) in exchange for its short-term basis risk. While we have not come to a conclusive determination on this, we continue to analyze all these facets and look forward to sharing more in future letters.
Finally, we can see LAR was weighed down by its slightly positive Delta exposure throughout the month, while LTR avoided that headwind and opposed the S&P 500 nicely (and expected, and by design):
The Volatility Market: Looking Outside
“MAGNITUDE, noun: Size that is purely relative. If everything in the universe were increased 1,000 meters nothing would be any larger that it was before, but if one thing remained the unchanged all the others would be larger than they had been.” - Ambrose Bierce
As we can see in the chart below, the VIX/IV response in September was seemingly within expectation.
However, when we take a closer look and start to incorporate where IV was starting from, we can see that the result in September was in fact well below expectation. One might wonder why this would matter for a strategy that is long volatility. After all, any positive result for volatility should simply be a net positive, right? Usually, yes. But not always. Among several methods we use to generate carry for holding long volatility (such as Gamma/Vega scalping), our models are also designed to increase total Vega exposure when IV is “cheap” (and via scalping, monetize some portion of the excess IV we hold following a pop). However, “loading up” on IV (put more simply, increasing option inventory and gross notional exposure) comes at a cost: an increase in total Theta, or time decay. This is built into our models, of course, but when we experience a disproportionately weak move in IV given a specific S&P drawdown while have that larger inventory, the increase in Theta becomes a heavier drag against the lackluster Vega gain the portfolio experiences.
Let’s walk through an example: Looking at the first chart above we see that a general model expectation (per historical behavior) for a -5% move in the S&P is for VIX to be +5 points. But this is without respect to VIX’s absolute starting level coming into the month. However, looking below, with a VIX starting from a lower level of 13.88, we would actually expect a 7-8 point move in VIX given a -5% drawdown in the S&P (and hence our model’s inclination toward larger inventory as IV gets relatively cheaper). While this is not precisely what our models do/project, a naïve strategy might consider increasing their exposure to IV by 1.25x-1.50x to take advantage of this more attractive IV-to-S&P relationship. So, circling back to the point above, without the expected more powerful leap in IV, the higher cost of Theta overtakes the gains in Vega. We can see the red dot in the graph below and note the large difference between realized and expectation.
* Methodology: VIX point change over 21 day rolling periods vs S&P 500 % change over 21 day rolling periods, given VIX coming from 13.88 +/- 1 point. 1990-Present.
All of this is to explain that when evaluating the “performance” of IV, we always need to be cognizant of what level we are coming from. As we illustrate, the expectation for IV is dramatically different depending on its starting level. If we look at a hypothetical situation (below) where VIX is coming from a level 15 points higher than it was at the beginning of September (28.88 instead of 13.88), we can see the expectation for a -5% down move in the S&P 500 is only about +2 points in VIX, rather than the +7-8 points we expected from a level of 13.88 above. While these observations make our increasing exposure the logical and probabilistically preferred action, as we saw in September, they can also cause increased pain.
* Methodology: VIX point change over 21 day rolling periods vs S&P 500 % change over 21 day rolling periods, given VIX coming from 28.88 +/- 1 point. 1990-Present.
“Don’t be irresponsible in your risks but keep trying; keep taking the best shots.” - Dean Kamen
In sum, IV behavior is never in a vacuum. At the same time, while our strategies will suffer if our IV forecast is below expectations, the right thing for us to do is to keep taking high probability bets that we observe have historically provided a positive edge. If we keep swinging at fat pitches, eventually we’ll not only get a hit, but also perhaps an extra base hit/home run/grand slam that is the raison d’être of a long volatility strategy.
