Logica Capital August 2022 Commentary
Logica Capital commentary for the month ended August 31, 2022.
Summary
As the S&P 500 charged upward, poised to break through its 200-day moving average to the upside, we instead observed a short-term failure at that level, and significant retreat to follow.
Concurrently, volatility markets exhibited a little more normalcy in August (vs. prior months) as evidenced by a VIX/Implied Volatility move (VIX +4.5 pts) that was fairly commensurate with the S&P 500’s downturn (-4.2%).
Get The Full Ray Dalio Series in PDF
Get the entire 10-part series on Ray Dalio in PDF. Save it to your desktop, read it on your tablet, or email to your colleagues
Q2 2022 hedge fund letters, conferences and more
Commentary & Portfolio Return Attribution
As mentioned above, August gave us a fairly predictable month in terms of S&P 500 Implied Volatility’s relationship with its underlying. Our red dot below shows the relationship in August -- a breath of fresh air.
Given this mild return to normalcy, we don’t have many surprises in our attribution for August. Our Macro Overlay continues to drag on return a bit, although buoyed by exposure to the US Dollar. Elsewhere, successful delta/gamma scalping provided a tailwind for both strategies.
Specifically, we illustrate the success of our scalping in the charts below.
We can see a proliferation of dots in the good, approaching “very good” quadrants/sections. Even when we got it “wrong” from a delta exposure point of view, the S&P 500 cooperated by experiencing a minimal return in either direction, leaving us fairly far away from any “very bad” outcomes.
This is, of course, what we expect from our models! But, as we all know, the market doesn’t always conform to our model’s expectations.
“The difference between good and bad architecture is the time you spend on it.”
- David Chipperfield
With respect to vega scalping, there’s an interesting juxtaposition here that may allow us to expound on our philosophy a bit more. Here’s a look:
We can see that this outcome looks far less favorable (in fact, it’s almost the opposite) than the delta scalping chart.
In August, this was by design. As there are two main variables (among several more that optionality has to offer!) that we can work with (delta and vega), there are of course 4 general categories/combinations of positioning in terms of risk control:
Some range of high/low delta combined with some range of high/low vega, with the two most risky combinations being the two that do not share “directionality” (e.g. high delta combined with low vega, and low delta combined with high vega).
To clarify for explanatory purposes (specifically, translating Greek letters into English concepts!), let’s take a moment to understand why high/low delta/vega and low/high delta/vega are in fact risky, but in simpler terms.
Delta, broadly, is market direction influence on an option, and Vega, broadly, is the volatility influence on an option. Thus, high delta/low vega, translates to more long exposure, less vol exposure; and since market direction and volatility are historically negatively correlated, this would result in a market drop (losing on heavy delta/long exposure) getting minimal long vol payoff (little to make on low vega/volatility exposure).
Similarly, for low delta/high vega, a strong market bounce or recovery would lead to minimal participation (low long exposure/delta) and massive volatility crush (high long vol/vega exposure), as Vol retreats.
With all this in mind, and because of the uncertainty in recent months with respect to the implied vol/S&P 500 relationship (we have not seen IV appreciate as much as historical, given a similar magnitude drawdown in the S&P 500, and saw too many results outside of expectations), our models have been putting the reins on our risk a bit, and essentially limiting us to either high delta+high vega (let’s just call this “high”), or, low delta+low vega (and this “low”).
And, of note, both of these variants stand in stark contrast to the two most risky ones described in the above paragraph, as these lower risk positions have delta/vega sharing in directionality (high/high and low/low).
“So what do we do? Anything. Something. So long as we just don’t sit there. If we screw it up, start over. Try something else. If we wait until we’ve satisfied all the uncertainties, it may be too late.” - Lee Iacocca
The point is, with higher uncertainty, as has been the case in 2022, our positioning has generally been more “hedged,” meaning that if our model wants to take a larger directional bet using a higher delta exposure, it also adds additional vega (remember that, generally speaking, these two components will be negatively correlated:
On a day when the excess positive delta exposure pays well, it’s very likely the excess vega exposure will detract from overall PnL). The same applies to the “low” exposure: if our model wants to take a much smaller directional bet using a lower delta exposure, it removes vega exposure, thereby removing the risk of the potential “double whammy” loss on both delta and vega if the S&P 500 were to violently bounce upward.
One of the advantages we have at Logica is the ability to analyze what-if scenarios instantaneously, so that we know the precise effects of manipulating delta and vega (alongside other variables).
This may sound straightforward, but because option payouts are non-linear, and option variables have complex interrelationships, it means that while two portfolio scenarios may look nearly exactly the same on “normal” days, if we look at their outcomes at the “tail,” the two portfolios can vary quite dramatically.
As a brief example, here’s a look at before and after one of our trades, where we kept delta exposure the same, but added vega exposure:
We see that, as foreshadowed above, these 2 scenarios are very similar “on average”, e.g. between +/- 1% return for the S&P 500 (the approximate mean daily move of the S&P500 at its long term realized volatility of 16% annualized).
But, we see that the “more vega” portfolio diverges further and further, in a good way, from the “less vega” portfolio as outcomes get more extreme.
“Don’t let your mouth write a check that your tail can’t cash.” - Bo Diddley
The illusion here is informative: it almost seems a no brainer to choose the “more vega” portfolio. But the trick here is that, of course, most of the time the S&P 500 will return between +/- 1%. So most of the time the “more vega” portfolio will be the inferior choice, as evidenced by the blue line being slightly higher than the black line at the valley of the straddle.
Another way to view this is something we keep an eye on consistently, which is what we think of as a “balancing” of the straddle, or, basically, given a distribution of different S&P 500 return scenarios, at each point, how many calls-to-puts should one own in order to break even, and then, in calibrated steps, how many to own in order to not lose x%, while making y%, etc...?
Again, the chart is very instructive, especially in the middle. Outside of the fact that ratios get a bit out of whack there because we are dealing with small absolute dollar amounts, one of the things the wonkiness in the middle displays is that, in that zone, the portfolio PnL is essentially just theta and vega drag, and there isn’t much delta movement to make up for it.
E.g, if the S&P 500 is +10bps, one must have more than 1.5 calls for every 1 put to make up for “cost” of the paltry move, and if the S&P 500 is -10bps, one can only have about 1 call for every 1 put, given the assumptions one makes(5).
Accordingly, this wonkiness in the middle is partly portraying what exactly the implications are for the “valley of a straddle” and what work we must do to overcome the cost of ownership.
5) Our models must of course make some assumptions in order to estimate both projected PnL and breakeven ratios. Integral to both are the assumed Implied Volatility behaviors, as well as the associated co-movement of both the S&P and non-S&P 500 positions we hold in the portfolio.
Given that both relationships are non-stationary and can be convex, the accuracy of our assumptions is a highly proprietary area of our modeling as well as a primary way that we generate alpha.
Finally, taking a look at the daily movement of our strategies for the month, we see LTR with a nicely negative correlation, and LAR generally uncorrelated over time.
Follow Wayne on Twitter @WayneHimelsein
Source valuewalk