Last week I published the quarterly update on the TILS model portfolio: it did 7.41% in just three months, a great result (yes yes, it can give it all back even faster but let me enjoy it for a bit).
A few days later, it was confirmation bias time. The Man Institute published this post, validating not only the use of Trend Following along stocks and bonds, but also the “stacking” way of adding it on top.
Sure, TILS model portfolio adds only 30% of trend and “wastes” some return with $TAIL but still, the above picture should reflect our reality as well. If you do not know MAN, they pioneered the application of systematic trading within quantitative strategies since the ’80s and now manage $170bn in assets.
Smooth sailing ahead?
“Ahahahah you fool! that’s the past, it will never work in the future.“
That’s what the little man on my shoulder whispered in my ear after I read below tweets.
probably the biggest thing i changed my mind on over the last 10 years or so…i think the diversified portfolio of lots of “strategies that makes sense that used to work” is much more fragile than one high performing thing you know inside out— Robot James 🤖🏖 (@therobotjames) April 4, 2024
Robot James and Agustin Lebron are two well-respected fintwit personalities. If they talk, I listen.
Obviously, what they talk about is not material for us common mortals. None of us can even aspire to have “a high-performing strategy that we know inside out” or “a single strategy with an expected Sharpe Ratio of 2”. But this is a great prompt to reflect on the potential weaknesses of our model portfolio.
Putting together some meh strategies (each with a Sharpe Ratio of 0.3) to reach a way higher result (Sharpe > 0.7) might sound magical. Diversification/uncorrelation is doing a lot of heavy lifting but:
diversification works because investors cannot manage to look past single line items. They hate to be allocated to something that is not working, especially if this underperformance lasts more than 3 years (and I am generous here). Doesn’t matter if the whole portfolio is performing as expected, the rebuttal is “ok, but if I remove the underperforming asset I would have even better results!”. Performance chasing allows diversification to work.
perfect uncorrelation is a unicorn but there are plain reasons why some assets perform in opposite ways in some circumstances. Sorry but you have to allow me some simplifications here. If the economy is not growing and inflation is under control, Central Banks will take stimulative actions: stocks down, bonds up. If inflation SURPRISES on the upside: bonds down, commodities up.
The fact that trend following can go long AND short stocks and bonds allows the strategy to be, by definition, inversely correlated at times. You only need a trend (and a strategy that correctly identifies that trend).
Diversification helps up to a certain point. We need uncorrelation to kick in when it is most needed, when stocks go down. The past might offer guidance but it is not written anywhere that the same patterns would repeat in the future (I would rather bet on this than not but this is my personal preference).
Convexity is also an important piece of the puzzle. An asset that can generate “accelerating” positive returns while stocks dive is more attractive than one that simply generates positive returns: we can allocate less to the former while getting all the benefits we need…and lose less (since our stock allocation is higher) when the sailing is smooth. Convexity is always beneficial when we deal with margins: it does not matter where the “+” and “-” are in the portfolio, as long as the overall value is not down a lot. What is really hard is to extract the rebalancing bonus out of convex assets: rebalance too early and the whole point of having a convex asset is lost, rebalance too late and…the bonus is gone.
The issue with Trend Following
As I wrote many times in the past, there is no trend following beta. Trend is an active strategy. Outside trend managers that use their performance against stock and bond betas to show their diversification benefit inside a 3-leg portfolio, other researchers use indexes (Man Institute in the above analysis employs BTOP50) to highlight trend thaumaturgical properties. Unfortunately, we cannot invest in that index. So that whole analysis can be as useful as the coffee cup bottom I just threw in the bin. Indexes are just averages of some funds’ performances and the funds included in an index change all the time. Their style might change: a manager might realise that reducing the fund volatility and including exposure to stocks and/or carry might improve the fund performance, at the detriment of the fund diversification benefits in a stock/bond portfolio. Guess what they will choose? Yes, whatever maximises their fee intake (as of today, this means adopting a shittier portfolio-wise strategy).
If I invest in a specific trend following fund, I ride along these issues:
the fund I chose does not have anything to do with the backtest I just saw. Might be better, might be worse, who knows.
the fund’s past performance might have been due to luck. Or their strategy might be arb-ed away (by front running I guess?).
the fund manager(s) can move to another fund, or worse, stop paying attention because they found better things to do in their life, but still choose to cash their fund check for as long as possible.
strategies with a positive but low Sharpe (<0.4) can deviate (underperform) from their target for a decade and that would not be a definitive sign that the strategy stopped working (think about lost decades for equities).
The easiest way to mitigate these problems is to pick many managers in your portfolio trend sleeve; yep, diversification again. But his might be akin to mo managers, mo problems.
There are TF funds that try to replicate TF indexes, like $DBMF, but they have issues as well:
if trend as a strategy is by definition backwards looking, a pure replicating model is a rear-view mirror looking into a rear-view mirror….not exactly great?
if the funds included in the index trade markets that the replicator does not trade, the replication might be impossible.
if the replicator trades too many markets, their regression model might hallucinate. Plus fees, costs, margin management and all other sorts of admin nightmares.
There are some funds that try a hybrid solution between a replication and a proprietary strategy, what $RSST and $RSBT call top-down and bottom-up approaches, but this might simply add the problems of one model to the ones of the other.
What do I do?
I do not have any strategy with a great Sharpe so…I am stuck with asset allocation.
I diversify.
There are behavioural aspects behind trend returns. And low Sharpe trend strategies require more the ability to stick with them (the periods where the strategy loses are way longer than the periods where the strategy works) than a complex system. At the end of the day, a portfolio with a Sharpe below 1 does not attract too many snipers, its mediocrity might represent its best feat for survival.
And then I diversify more inside the trend sleeve.
I try to choose managers that I know I follow, who are transparent about their process and have skin in their game (Once I interviewed Elisabetta Basilico, who collaborates with Wes Gray. In one of our chats, she mentioned “now Wes moved to Puerto Rico” and in my mind I was “gurl, I know the size of the emergency generator in his house in Puerto Rico”. Maybe I follow them too much).
Someone recently posted on Twitter a correlation matrix between the outstanding TF ETFs. Putting aside the fact that are all quite new, so the data is not that informative. I am still not sure if I am keener on having three replicators, trying to achieve at least a good replication by average, or three very uncorrelated strategies, with the risk that I would never understand what is working and what has stopped. I definitely like the Stacking element of $RSST/$RSBT, so those are probably the last ones I would exclude, absent any other leveraged alternative.
What I am reading now:
Follow me on Twitter @nprotasoni