# An Illustrated Guide to my Flagship Portfolio

### Risk on-Risk off

I am not a fan of the momentum portfolio I shared yesterday because it is riskier than the market portfolio and I am exclusively interested in portfolios that are demonstrably *less* risky than the market — even if just by a little bit. At the heart of my business plan is a methodological innovation that can be described as “process control” over systematic risk. The idea is to isolate and control our exposure to systematic risk. For our purposes, we can say that I am in exclusive possession of some technology that allows me to de-risk portfolios and more efficiently harvest the signal alpha arising from some market microstructure or constrained intermediary behavior. Our flagship portfolio applies this technology to harvest the premium associated with the overnight drift. This is the robust pattern whereby market returns are higher overnight than during trading hours. The pattern is an artifact of inventory risk management by dealers — so it is a sort of intermediary premium.

It is known that the associated premium cannot be harvested profitably through daily round-trip trading, as a new ETF is trying to do. We harvest it, instead, through the cross-section of blue-chip US equities. We talked about the construction of “the market portfolio” yesterday. The “derisked portfolio” is a monthly-rebalanced, longonly portfolio whose weights are, roughly speaking, *conditionally-optimal* — given what we have learnt about systematic risk and the signal as of the last trading trading day of each month. The “longshort portfolio” is a classic 130-30 longshort (ie, 130 dollars on the long leg, 30 on the short leg) with the market portfolio as the short leg and the derisked portfolio as the long leg. We ensure that there is no look-ahead bias, announcement lags, or survivorship bias.

Our unconditional estimate of the expected return on the market portfolio is about 11%, to be compared with 14% on the derisked portfolio, and 15% on the longshort. So, our estimate of signal alpha is about 3-4% per annum. These are compounded mean returns and they cumulate impressively. A buck invested in the market portfolio in October 1994 would’ve cumulated to $18 by May 20, 2022; whereas a dollar invested in the derisked portfolio would’ve cumulated to $38, one in the 130-30 longshort would’ve cumulated to $45. So, 3-4% alpha is pretty good, especially for portfolios that are less risky than the market portfolio.

In risk-adjusted terms, the past performance of the derisked and longshort is even more impressive. Our unconditional estimate for the Sharpe ratio on the market portfolio is 0.60. The Sharpe ratio of the derisked portfolio is 0.99, 65% larger than that on the market portfolio. The longshort has a marginally higher Sharpe ratio than the longonly derisked portfolio. The max drawdowns on the derisked and longshort portfolios are smaller than that on the market portfolio.

Those are the unconditional estimates. More recently, the market portfolio has taken a beating. It is up just 1% over the past year. Meanwhile, the derisked portfolio is up 8% year-on-year, and the longshort is up 10%. The year-to-date cumulative return is -12.9% on the market portfolio, -2.6% on the derisked portfolio, and +0.5% on the longshort. So, this may be one place to hide from the harsh gales of the present market winter.

One way to summarize a lot of information on the returns of these portfolios is to look at calendar year metrics. We can summarize this information even further by looking at the unconditional means of the metrics. The mean of means can be thought of as a robust estimator of the underlying parameter. We estimate that, consistent with our unconditional estimates, the expected returns on the market, derisked and longshort are 11%, 14% and 15% respectively.

Expected returns are only a third of the battle, however. What really matters is risk. We estimate a volatility of 17% per annum on the market portfolio, and 13% for both the derisked portfolio and the longshort. This is a very significant difference. Close inspection of the graph reveals that the difference is means is broadly representative of calendar year observations. Market volatility is usually a head above.

Max drawdowns reveal the underlying pattern. The derisked portfolio and the longonly portfolio sport lower volatility than the market portfolio because when the market falls dramatically, they fall by much less. The derisked portfolio thus earns its name.

What really matters, of course, is risk-adjusted performance. We find a mean Sharpe ratio of 1.00 for the market portfolio, which overstates the Sharpe ratio by two-thirds. The derisked portfolio has a mean Sharpe ratio of 1.44, while the longshort has a mean Sharpe ratio of 1.48 — these are overestimated by a half. So, this measure understates the risk-adjusted premium — eg, the longshort’s unconditional Sharpe ratio is 72%, instead of 48%, larger that the market’s.

We dig a bit further into the returns. The next graph shows a scatterplot with daily market returns on the X axis and the returns on derisked portfolio on the Y axis. We can see that returns on both the upside and the downside are tempered relative to the market.

The difference in returns between the derisked and market portfolios contains interesting information. Days when returns on the market portfolio exceed those on the derisked portfolio can be thought of as “risk on” days; conversely, for “risk off” days. By this definition, 51% of days are “risk-on”. On these days, the market goes gangbusters. The expected return on the market portfolio conditional on Risk-On is 146% per annum; conditional on Risk-Off, it is -116% per annum. By comparison, the expected return on the derisked portfolio conditional on Risk-On is 40%, conditional on Risk-Off, it is -11%. The beauty of the 130-30 longshort is that it achieves positive returns in both cases: 8% on Risk-On days and 21% on Risk-Off days. This makes the longshort portfolio particularly attractive for risk intolerant investors. It does allow us to hedge away some of the market risk in our otherwise derisked longonly portfolio.

In order to better understand what is going on with these portfolios conditional on risk on-risk off, we fit skew-*t* distributions to the returns of the two portfolios. This is a highly flexible, four-parameter family of distributions that admits both fat tails and skew, and includes the Student’s *t*-distribution and the normal distribution as special cases. Following Adrian et al., as we did with growth risk, we fit the parameters by quantile matching. The market portfolio is in blue (Risk-On) and green (Risk-Off); the derisked portfolio is in red (Risk-Off) and orange (Risk-On). Note how close the latter two are. While the market gyrates up and down, the derisked portfolio yields a considerably smoother ride.

According to the fitted distributions, the unconditional probability that the market portfolio will return more than 1% on any given day is 0.7%; while the same is 1.4% for the derisked portfolio and 2.1% for the longshort. Conversely, the probability that the market portfolio will decline by more than 1% on any given day is 14.9% for the market portfolio, 5.9% for the derisked portfolio, and 3.9% for the longshort.

Looking at the probabilities conditional on risk on-risk off is even more revealing of the characters of the portfolios. The probability that the daily market return will exceed 1% is 12.9% on days when it exceeds the return on the derisked portfolio (Risk-On) and only 0.8% on days when it falls short of the return on the derisked portfolio (Risk-Off). The derisked portfolio is much less volatile. The probability falls from 6.6% to 1.4%. The longshort is even less volatile: 4.5% to 2.1%. Conversely, the probability of daily market return will be worse than -1% rises from 0.6% to 14.9% between risk-on and risk-off; that for derisked rises from 2.3% to 5.9%; that for the longshort, from 2.8% to 3.9%. So, derisking and hedging market risk significantly temper the gyrations of the market portfolio.

All these interrogations yield a consistent picture. Our technology allows us to evade systematic risk to some degree. Not every time, surely. But systematically and consistently. Eventually, we want to have a diversified portfolio of products that harvest many different sources of signal alpha to further tame portfolio volatility. That is partly how we plan to survive and stay ahead of the competition. The other bit is methodological innovations, which, precisely because they are “merely” methodological, can be applied to any signal.

I am very excited about the launch of Systematic Portfolios LLC this summer. We should have our first rebalance in July, if not June. I can’t wait. Our general strategy for success is very simple. We aim to be the most informed shop in the market. Over the coming years, I will be looking for the sharpest minds on the planet to help me build the world’s greatest research shop, complete with a world class library. Do get in touch if you’d like to be involved in any capacity — as an investor interested in our portfolios, a market counterparty, an angel investor, if you’re interested in working for us, or if you’re a member of the media interested on my take on market developments. We’re going to be big. Before the decade is out, we’ll take [the] Citadel.

## Create your profile

## Only paid subscribers can comment on this post

Sign in## Check your email

For your security, we need to re-authenticate you.

Click the link we sent to , or click here to sign in.