With a professional tennis career spanning over 20 years, Roger Federer won 20 Grand Slam titles, more than 100 ATP single tournaments and north of 1250 matches, spending 310 weeks at the top of the ATP Tour ranking1. One of the best, most recognised and esteemed tennis players of all time, he managed to earn $1.1bn through his career. Not too bad, uh? And yet, the Swiss won only 80% of the matches he played and, even more incredibly, he barely won about 53% of the individual points he played2 over the two decades (that adds up to more than 230.000). Not what the average person would guess, I’m sure. Successful investing, much like tennis, is about consistency and repeatability of good decision-making, rather than spurious (albeit potentially large) victories, and it’s our job to research managers and identify those that we think have a higher chance of delivering outperformance through time.
Literally last week, we were looking at an equity manager that follows a top-down approach, where macroeconomic and thematic thinking informs sector, country and style allocation which ultimately drives stock-selection. I am being simplistic here, but this is the opposite of what we typically look for, which instead is bottom-up stock selectors with a clear philosophy and a distinct style exposure (value, quality, growth, momentum etc…), where the process is structured to identify stocks that should outperform no matter what the macro environment ends up being, and where country and sector allocation is a consequence of stock selection rather than an input. We are evidence-driven investors, and data (alongside our experience) suggest that the latter approach is more likely to lead to stable outperformance than the former. Now, the jury’s out on whether that’s really the case, or simply anecdotal. It might be that macroeconomic information is more efficiently spread than stock-level fundamentals, hence leaving less room for alpha generation. Or it might be that macro cycles are less impactful on stock prices than one would think. Or it might be that the data we’ve looked at is simply wrong and biased.
We like to challenge our assumptions quite regularly, keeping ourselves honest, and this time this specific manager seems to be pretty good at what they do. They have a well-structured process, some decent people, and most importantly, they have provided us with data supporting their skills at macro-driven alpha generation. Their track record is long and fairly successful, which might convince a superficial eye, but we are sceptical and we’re trying to disprove it. Or rather, we’re working on understanding if the successful track record is due to luck or skill, if it is driven by macro-allocation or other factors, and whether it’s replicable or not. If at the end of our research we can’t prove them wrong, then it might well be that a whole new world of investment opportunities opens to us, but if we can’t find any evidence then we’ll simply move on to the next item on the research agenda.
Now, the ball is in our court. We have all the data we need, and we’ll apply the scientific method to analysing them: formulate some hypotheses, make a few assumptions, agree on a test statistic, and test the hypotheses. For a macro-driven process, the assumptions we make are that the decisions a manager makes are expressed through relative positions (vs benchmark) in terms of country allocation, sector allocation and style allocation, but we’ll check stock selection too, for completeness. Both direction (over/underweight) and magnitude (big/small deviation) matter for alpha generation, so we’ll have to test both types of decisions, separately. Another assumption is that each decision is taken monthly, i.e. increasing geographical weight to France in May, and keeping the same relative position in June are two distinct decisions. The hypotheses are four, distinct: the manager is good at country/sector/style/stock allocation. The test statistic, or the metric we’ll use to evaluate whether the hypotheses are true, is the ‘Hit Ratio’, which measures the percentage of calls that were right (a right call is being over/underweight something that out/underperforms and having a larger/smaller relative position in something that moves more/less). Naturally, we’ll have to divide the entire history into a few non-overlapping blocks to be able to calculate the Hit Ratio’s t-stat and evaluate whether the hypotheses are accepted or rejected.
What’s all of this to do with Federer then? Well, if my good friend Roger managed to win so much by having a hit ratio of 53%, then certainly that’s good enough also for our macro manager, so we’ll accept the hypotheses if the hit ratio is significantly (in a statistical sense) higher than 50%. But also, Federer has been number 1 for so many years because he kept his hit ratio consistent throughout the years, over hundreds of thousands of points played. This is the concept of “breadth”, i.e. the number of independent calls a manager makes (or the number of points, that in tennis are all independent, that a player plays). By having many observations (many years of monthly data) and many dimensions (there are 10 sectors, 20+ countries, 2 styles and 1500+ stocks) we have artificially ensured that breadth is large enough, within our assumptions.
If you want to know the results of this research project, check out our factsheets and newsletters in the coming months and you may just see a new addition in our manager lineup.
Sources:
120 Grand Slams, $1.1 Billion In Earnings: Roger Federer’s Legendary Career By The Numbers (forbes.com). 2Roger Federer's Success: A Surprising Lesson for Your Investments (forbes.com)
