The Value Perspective Podcast – with Joe Peta

AE: Joe Peta, welcome to The Value Perspective Podcast. It is a pleasure to have you here. How are you?

JP: Thank you, Andy. It is a pleasure to be here. I am really anxious for our conversation.

AE: Brilliant. Maybe we could start with you introducing yourself – your background, your career and how you got to the point of writing a book?

JP: Sure. I started in the financial industry in the mid-1990s. I was actually an accountant by trade – a CPA out of university – but I wanted to go up to Wall Street as I was always fascinated by stocks. I didn’t really know what that meant – to work on Wall Street – but realised the best way in the US to have the Wall Street firms come calling on you is to get an advanced degree. So I got my Master’s, an MBA, and I started working for Lehman Brothers. I actually met CEO Dick Fuld in my first week of graduate school and I was instantly captivated by his story about what he was trying to do at Lehman Brothers – and it was a great fit for me.

I started as a NASDAQ market-maker and, after eight years, Lehman asked me to move from New York to San Francisco and help launch a ‘TMT’ fund under the Neuberger Berman umbrella, which was a wholly-owned subsidiary of Lehman. That allowed me to move to the buy side from the sell side without ever leaving Lehman Brothers – I didn’t ever want to leave Lehman’s – and that lasted for about five years until the group’s bankruptcy. At that time, I really started moving around. It was not something I wanted to do – I guess I wanted to work for one employer for 25 or 30 years – but, as I started moving around, the industry was changing and data was becoming a big part of it. And whether that was involving traders or portfolio managers, I found myself starting to get immersed in data.

I had an unfortunate accident in New York City, working for a bank, where I was in a wheelchair for a period of time, with my family still on the West Coast – and that is when I got the idea to start writing down some of these thoughts. I saw a lot of critical reasoning overlap between asset management; the book and the movie Moneyball and what was happening in professional sports; and then in sports betting too. So that was really three interests of mine coming together and I wrote my first book, Trading Bases, which led to a whole new career in data analytics. I joined a company called Novus, which was a fintech in the data analytics space; from there, I joined a client; and, from there, one of the very large hedge funds in the US came calling. They were familiar with my work and asked me to solve some of the problems I write about in Moneyball for the Money Set. I thought there was a story in letting people know, Hey, here is what I found over the last 10 or 15 years doing this – I think it is applicable to a lot more places than just where I have been.

AE: That is a really interesting backstory – thank you for that. I am now going to introduce my co-presenter on the home team today. Robert Donald is an avid listener to the pod – or so he tells me! – and he is joining us because his background overlaps in a similar area. But, Robert, maybe I should let you talk about that?

RD: Thanks, Andy. I am an avid listener because I find the only time I get to enjoy podcasts that are in the market, and particularly in the firm, is when I do gardening. So it is a good time to pretend I am doing gardening in the eyes of my wife – but actually I’m listening to work! My background is that I have done 15 years on the sell side. I started in corporate finance as an equity sell-side research analyst – thinking I was going to discover the answer to all future problems by being able to predict the future with certainty – and I actually worked in Schroders’ equities in the mid-1990s for a period.

Then we were sold to Citigroup and being moved to an American institution and out to Canary Wharf was an incentive to leave for the buy side. I went and worked as an active long-short manager at GLG Partners, Soros and Brunner and then moved to Schroders to build a multi-manager, multi-sleeve, marker-neutral long-short product. On that, we use many data points to appraise and select managers and we often think about some of the touch points I believe Joe has made in his book and his analysis at Novus. So I am very happy to be involved in this conversation, Andy – thank you for inviting me.

AE: Thank you very much, both of you, for joining us today. Now, Rob is going to play the role of co-commentator – or ‘colour commentator’, as I think you call them in the US, Joe, which takes us nicely onto the subject of sports. In the introduction to Moneyball for the Money Set, you talk about the combination of sport and data – and that instantly drew me in because I am very passionate about both areas in different parts of my life. Could you talk about the book and what prompted you to write it?

JP: Well, actually writing the book was a result of putting down the years of research in the field that I had done, which I thought had started to overcome the long-running industry caveat of past performance not being indicative of future returns. I thought some of the work I had done either had broken that code or was starting down a way where other researchers and practitioners could now pick up the baton and go much further with it. So I thought there was use in laying out the work I have done – I wanted it to be exposed, frankly. But that raises another question – it is not just, Why the book? It is also, Why go down that path of research?

And the idea that I could start using data came from the first buy-side firm I went to within Lehman Brothers – like I said, I transferred out West, where I was working with a portfolio manager (PM), who was an engineer by education. He was an expert in the semiconductor and hardware spaces – so a very logical thinker. And I started to realise – as you would from a trading desk when you start processing all someone’s decisions, and you see some of the behavioural finance issues that might come up in their decision-making – that, while he was very skilled in some ways, he was eroding his skill by undertaking certain actions. So I knew I needed a way to communicate with him. I had come from a market-making background – and, on those desks, we just yell at each other, which I knew wasn’t going to be effective with my PM! So I needed a data-driven approach to try to get him to think about his decision-making and the whole process around decision-making and to really identify, Hey, sir, this is what you are good at so let’s stick to doing that stuff.

It was very crude metrics but it got me thinking about that process and that is kind of what the book is about. It is the result of starting from that very crude idea that there are data queries that can be solved and the problems are very similar to those in sports – so a data-rich environment and highly compensated star performers within a large pool of other performers. And because the data gets captured we can use some of the same techniques. Oh, and of course, the biggest one is there is so much noise that goes into the final results, right? So using sports – and specifically baseball, where I had the most knowledge at the time – I was able to separate that noise from results by focusing on skills. That was really the process I started down and I guess I thought this could be a helpful story and hopefully interesting to others in the field.

Separating luck and skill – in golf and money

AE: That’s great, thank you. Separating skill from luck is something we have discussed a lot on this podcast. Could you maybe talk about some of the key takeaways from your book – in particular, some of the methods you use to separate skill from luck or noise when assessing performance?

JP: I’ll tell you what my overriding template became: when I spoke with a head of one of the very large multi-manager platforms, he said to me he had three problems that his quants were unable to answer – and you will recognise these as noise and skill issues. The first one, he said, was anytime a PM has a good year, they come to me and ask for more money. Now, in this case, he wasn’t talking about pay – he was talking about buying power, or AUM, which of course the PM hopes, translates into more pay. And he asked, but how do I know if their prior year was repeatable? By the same token – he was very frank – he said, We have let go or cut back some PMs on buying power and, worse, let go some PMs who went on to have very strong careers elsewhere. So his second problem was, he said, how do I avoid making that mistake when my risk department tells me, Hey, this guy just blew us up for a year – we have got to get rid of them?

And his final question was, we are all trying to steal each other’s talent – so how do I know what a PM is worth going forward? I know what they have made me before, he said – but what are they worth? And those three questions, they hit to the heart of what sports analysts attempt to do. Again, I will use a lot of baseball analogies because this is what I was most familiar with initially. And on that noise and skill idea, if you think about American baseball, you know, the pitcher is trying to stop runs from being scored – and that result you can look at, at the end of the year. And the pitcher who allowed the fewest amount of runs, just mechanically, had the most value to his team.

Yet that result is subject to a whole bunch of different variables the pitcher cannot control. He cannot control his opponent; he cannot control the ballparks he pitches in, compared to others in the league; he cannot control the quality of the defence behind him, which has a lot to do with not giving up runs; and, of course, in turning his ‘runs allowed’ into victories, he is dependent on the offence of his teammates. So that is why, while looking at ‘runs allowed’ will enable you to look back and say, This was the value, it tends to be very non-repeatable from one year to the next – very noisy results.

So what baseball analysts did was say, OK, we are just going to look at what the pitcher can control. All he can control is his strike-out rate, his walk rate and his ground ball rate – and, fortunately, those tend to be very persistent and stabilise very quickly, in terms of figuring out what his future ground ball, strike-out and walk rate are going to be. And it turns out, through multi regression, variable regression, you can come up with a prediction of what his future ‘runs allowed’ are going to be – and that process is much more predictive of future ‘runs allowed’ than his prior ‘runs allowed’, which have so much noise in it.

So I set out to look at PMs and find the ground ball rate, the strike-out rate and the walk rate equivalents for a portfolio manager – and that is what I lay out in the book. The first half of the book is about my framework – and the framework really takes stock selection attribution and breaks it into skills. It breaks it into the ability of a PM to consistently hold outperformers and then it looks at the quality of their outperformers, or underperformers – and all of these benchmarks are versus a skill-neutral performer. That is very important and differs from a lot of the software packages that exist in the industry and kick out stacks on PMs. Finally, there are some other skills too – there is measuring of the sizing skill and there is measuring of the PM’s ability to manage net exposure and the PM’s ability to manage gross exposure.

But it is really the consistency, the explosiveness and the sizing that are the three skills that can be used to determine the rate of alpha production of a PM. And while I used American baseball there, it turns out, as I mention in the book, the much better analogy is professional golf – and the ‘strokes-gained’ concept is what analysts use in golf. A professor at Columbia, Mark Brody, has invented ‘strokes gained’ and it breaks down prior performance into, essentially, three skills. There is a fourth, but the three main skills are driving ball, approach and putting and, honestly, it was that framework that was the most similar to what I did in explaining what skills a PM has in producing or detracting alpha.

RD: I like the golf analogy. When you describe it, though, I can understand the logic of sequencing from a driver to a fairway to a green – but how does that correspond to the actions of a portfolio manager? Is that the initiation? Is that the sizing? Is that the length of holding the position? How do you take that sequencing from golf into the PM world?

JP: The nice thing about the golf analogy is, those three skills are largely independent of one another, which intuitively makes sense, right? There are guys who are great drivers of the ball and not good putters, for example, and all of those are independent – you know, once the ball gets on the green, it doesn’t matter whether it took the pro golfer four strokes to get it there, six strokes to get it there, now the measuring of putting is going to be independent from everything that came before it.

So my framework is similar, Robert. These are daily measurements, by the way – it takes a look at a PMs portfolio on a single day – and gets everything benchmarked correctly. It determines that, for every PM you would be measuring, if it were a random PM – or, you know, a million monkeys with a million iPhones with a million Robinhood apps on it – the skill reading would be zero across everyone. So the first thing it measures is, Is the PM holding an outperforming stock within the universe they are either forced to play in or choose to play in? That way, it is market-neutral, sector-neutral, size-neutral and so on. That is the first reading. And what you find is it becomes a fairly strong leading indicator for alpha production. It doesn’t guarantee there is going to be alpha but the PM who can consistently ... and it is small amounts. It is almost like card-counting in blackjack – having a 1% edge every day. If you had 100 holdings, the average within the universe was 50 and you are holding 51 – you do that on a daily basis and you are going to create alpha that compounds.

Then the second step is, OK, now we put all the PMs back on an even footing and say, given the amount of outperformers or underperformers they have, what is the quality of those outperformers and what is the quality of the underperformers? In other words, are they identifying the best of the outperformers and avoiding the worst? That is a second reading and, while I was sceptical when doing that, it turns out that is the most important skill. It has the most value in terms of ‘meat on the bone’ because there are only so many outperformers you can hold. Nobody can hold 100% outperformers – nobody can even hold 60%, you know, on daily readings – but holding the correct, you know, ‘explosiveness’ turned out to be a key point.

And, when you think about the process of a PM – this seems to be the consensus of the people I worked with – the first step is sort of measuring the ‘science’ of portfolio management; and the second step is sort of measuring the ‘art’. So there is some indication in the explosiveness rating that a PM recognises the correct timing – whether it is sentiment or market factors or whatever. You know, there is an art to being a PM and it seems to be that that skill is somewhat captured by the explosiveness reading. And then, finally, there is a sizing component, which is largely mechanical – frankly – although the findings are interesting.

Then that arrives at the rate of alpha production – and, as you say, it is simple. It is elegant, in that it is additive from one period to the next – it is the same formula for whatever period you choose. The most important part, however, is once you have those three skills measured and identified – because, frankly, this is just another attribution system – it turns out, since they are skills like golf, now we can look at the persistency. And, as you may know from golf data, a number of golfers can finish a tournament with the same score of, say, eight strokes under par – all beating the field by four strokes, because the average was minus four.

Well, it turns out the way you get to minus eight is very important in determining how you are going to project that golfer will do going forward. If they got to minus eight by having the hottest putter – you know, if six of those eight strokes were contributed by putting – and six strokes of another golfer’s contribution came from iron play, well, the latter is much, much more likely to repeat that performance next week than the person with the hot putter. And then the driver fits somewhere in between those two. That really was the breakthrough insight that turned my framework into a predictive model.

RD: Moving away from golf for a second, how do you think about the sizing element in the context of stocks? If a manager looks at an individual stock, it has a standalone perceived risk profile because of its volatility. So, while the manager may have high conviction towards it among a collection of stocks and may think it could be explosive, that standalone volatility means it could go against them as much as for them. How do you think about adjusting the sizing, because they may have chosen to have a low sizing to what is a high-volatility stock, even though it is potentially very explosive – because they would be carrying too much risk in that name if they oversized it and it went wrong?

JP: That is absolutely right, Robert – and I acknowledge that in the book. I don’t go a lot into it but, in practice, I have looked at that a lot. And you are absolutely right: part of the PM sizing decision is not just conviction – it is a risk management decision as well. And, without question, they are effective at that. How I have measured that – at a number of different places across dozens, if not hundreds, of PMs – is in relation to the skill-less person. If the skill-less person had a 40-position portfolio, they would just make everything 2.5%, right? So that is where you can say, Hey, did your sizing help or did you sizing hinder? And it is true that, if you equally weight as a population all of the PMs I have looked at, the risk-adjusted attractiveness of their sizing does mildly increase their Sharpe ratio. That is true. So the risk-management component that is going into sizing is valid – and, to that end, I just had a PM give me the example of holding Microsoft versus Black Box, right? They cannot possibly have those equal because it would change the risk profile.

RD: It is stating the obvious but one of the motives of managers is to survive and retain client capital – and, going back to the analogy of your manager who wants to get paid, he needs to know he has a job next year and that he has a reasonably ‘sticky’ amount of capital. So the risk management is also about their career.

JP: No question, Robert. And that is where, unfortunately, there is a tension between what is best for the LPs [limited partners] and what is best for the actual manager, in terms of career risk. I do include an actual case in the book about two PMs – they were young, this was going to be their first pay day – that illustrates that conflict in terms of allocating capital. They wanted to protect their lead. That is a very real situation and it is something everyone needs to be aware of – from management to the PMs themselves who, as you say, are managing their careers.

RD: I know Andy has a burning question but just one thing you have stimulated my thinking around is – while I come from a long-short background and have a great deal of respect for long-short as a skill set – I think the jury of the market can be quite punishing for a long-only manager. In the sense that, not only do they have to select well, they also have to be mindful of what they haven’t chosen to own and that is going to punish them because it is in their benchmark – whereas the long-short manager, is only going to be punished for what they do own on the long side and what they are short on the other side. For them, it is all about dollars they are making – but, for the long-only manager, not only are they mindful of how well they are doing on their sizing and their selection and their hit rates on their longs, it is also the unintended consequences of what they don’t own. And we have seen recently, for example, some very significant bear squeezes in the market, where 20 or 30 basis-point names in an index have gone up 300% in a month because a lot of social media influencers have pushed certain electric vehicle or battery names in certain regional markets – and the long-only manager can just look like a turkey for that when, in fact, it was never on the radar screen of their selectable universe.

JP: Yes. Again, you are talking about inefficiencies within the investing universe that are not the PM’s fault – but you are absolutely right. I will tell you, all my skill measures are based on the PM’s universe – and that is not necessarily fair to them. But what this framework hopefully does is, in the hands of a skilled allocator or a skilled investor, they would know not to punish the PM with that. Still, you are right – it doesn’t make the PM feel better about their career risk if their investors are not that understanding.

Meaningful ways to compare apples and pears

AE: That actually ties in nicely with a question that brings us back to sports analytics: presumably you would not analyse the performance of a baseball pitcher the same way you would a batter – or in football, for a goalkeeper versus a midfielder versus a striker? To Robert’s point, then, if analysing long-short could be very different to analysing, say, a long-only value manager, have you thought about how appropriate some of your skills assessments are for different sorts of managers?

JP: That is a great question. Let me go to the sports analogy part of it first – and I will try to keep it to football worldwide. You are correct in that the metrics you would use to evaluate midfielders versus strikers versus goalies are very, very different – but all of the metrics you are using are designed to be able to compare each of the players or each of the positions to their value in winning a game. In other words, yes, Kevin de Bruyne is an elite midfielder because of his ability to pass the ball and complete passes, whereas Harry Kane and Erling Haaland are elite at finishing while, say, Jordan Pickford’s value comes from preventing goals.

So you would only measure de Bruyne against midfielders, your strikers against other strikers and your goalies against other goalies – and in some ways, there’s no comparison. But what you do come out with in the end is how much a midfielder contributes to a team’s goal differential at the end of the year, how much a striker does and how much a goalie does and so, whatever final metric you are coming to, in that sense, is a very efficient metric – and not necessarily in getting each player right, but getting the value of each position right. And we know that because the contract market and the transfer market is going to be efficient over time – because there is a lot of money going into it. And we know, for example, the best goalie is not going to get paid what the best striker is because the striker contributes more to the win.

So I think my work is consistent with that. At the beginning of the book, I really talk about the importance of benchmarking – and about how the standard software you can buy, whether it is batting average or slugging percentage, is worthless. I hope I do manage to explain in the book why it is virtually worthless as a data point – because, to what you are saying, it is not comparable, right? It is not comparable to the unique universe that each PM might be residing in. Yes, the majority of my work has occurred within multi-strategy firms, and specifically on the long-short equity platforms – and fortunately so, because that is where the majority of the industry’s data is too. Now, you mentioned value long-only managers and I don’t spend a lot of time dwelling on such areas in the book – but, at one point, I do point out that with each PM on a multi-manager platform, I am doing separate frameworks and projections on both the long and the short book.

So, say there are 50 PMs, I actually have 100 projections – so, you know, there are 50 long-only projections and there are absolutely value managers within long-short market-neutral hedge funds. There are absolutely managers that run value long. There are sceptical PMs everywhere and they are value-long and growth and momentum-short and there is no difference – as long as you have them properly benchmarked, you can get different skill readings that then become comparable to other PMs. And, in that sense, a utilities PM, who is equally skilled to a consumer PM, will never be as valuable – just because of the universe and similar to strikers and goalies.

RD: And is that down to the lack of dispersion within utilities?

JP: Dispersion is the major one but I think you would also say there is more volatility in consumer as well, right? Because essentially the growth element is not the same for utilities, which is a regulated industry and also more of an interest-rate play. And I have a whole chapter in the book, called ‘Hidden Figures’, because I do think that dispersion is not studied enough or calculated enough at the specific sector or universe levels – but that is all in the model and the framework. So, yes, I believe it works just now. I will say this about a value-focused long-only manager, however – my work depends on capturing PM decisions, right?

But this is where we might have a difference in terms of what PMs do on a daily basis. In a hard, market-neutral, long-short platform, generally, the PMs are very active in terms of their decision-making. They trade a lot, the positions change and I am capturing each of those, which enables a more robust look at what the PM actually does. If there is a long-only value manager who simply holds 30 names at the beginning of the year and holds the same 30 at the end of it, there is really not much to read there, right? I will get the same readings but they almost certainly, I would expect, not be as robust in terms of persistency from one period to the next.

RD: But isn’t it all about looking at the value manager versus a value universe and seeing how good they have been relative to what that value universe was, rather than the broader market? Actually, just before coming to do this pod, a colleague of mine highlighted – and I don’t know if this data point is correct but out of respect to him, I’ll assume it is – that apparently Warren Buffett has quite a low ‘hit’ rate with his ideas. But what he has benefitted from are the passage of time and the quantum of the returns on the ones that do come right. So he may have a less than 50% ‘hit’ rate but the sheer length of time for which he holds and the compounding effect of that and the quantum of the compounding means he is quids-in over the performance.

JP: I actually touched on that exact claim about George Soros in the book. You do hear similar things about Warren Buffett but I specifically address the claim – and it has been made by an ex-CIO of his – that Soros has a 30% ‘hit’ rate. And it is not possible for that to be even remotely true – and I lay out why in the book mathematically. I also suspect that same narrative gets applied to Buffett – and Buffett is a better stockpicker than hitting under 50%. I would love to do that work – you know, Buffett is obviously constrained from measuring his sizing for a lot of reasons but, as I show in the book, it is very hard to have a consistently and meaningfully below-average ‘hit’ rate and make up for that with either the passage of time or explosiveness.

RD: I respect that comment greatly. When I was at GLG running a long-short strategy, we had all this kind of analytics done on us in the mid-2000s and my strategy was compared with that of the person I worked next too. The analysis concluded I was a ‘boring fundamental stockpicker’, with a particularly successful ‘hit’ rate, but I was too scared about running my winners too long – although I was quite quick at taking down my losers. I ended up with a very nice return but I could have done a lot better if I just ran my winners longer. The other chap, who was a Goldman Sachs trader who came onto the platform, had a less than 50% hit rate but he was super-quick at taking off his losers and he was super-good at sizing and adding to the sizing of his winners. Still, his average holding period was very short around specific catalysts so he made enough on those wins and sized them up as they started to traction but he was very quick at taking them off. So I have seen people, certainly over a three to six-year period – I am not saying beyond that – do actually quite well with a low ‘hit’ rate.

JP: How much AUM were you running?

RD: Well, that is a great point – he ran hundreds of millions, not billions. And just to add a little bit of colour to the story, I used to invite him to company meetings the whole time and he never came to any of them. And one day I asked him, Why is it I invite you to all these meetings but you never come? Where is your edge? Aren’t you going to miss out on the colour of the companies? And he said, No, no, no – all my edge is sitting in this room because what I do is I watch all of your positions and I see how you’re behaving. And I’ll exploit that. So he was very good – and those were the days, which isn’t the case now, when we had a lot of transparency within all the different sleeves. His edge was watching other people in the building and how they were trading and where the crowding was and he was exploiting that. I didn’t have time to go and look what other people did. I didn’t really care what other people did. I just wanted to build my quarterly model and understand who was going to beat the next results meeting. Anyway, his process was different and it worked for him – but, yes, he didn’t have a lot of capital, but he was very mobile with what he had.

JP: I have had this discussion in-depth with Michael Mauboussin in terms of the claims about Soros and we certainly agreed that – and you touched on it right there – ‘watering your flowers’ and ‘cutting your weeds’ is a very wise way to invest. And there are certainly ‘swing’ traders – and, like you say, your colleague may not have even known what the product was of the companies he was trading, right? He was trading stock symbols and charts and technicals. No question – that is an absolute formula that exists to make money. It is usually not scalable to the billion-dollar firms and, within the platforms now, there is almost no-one that does practice that approach – and I do take pains to talk about institutional-quality investing that exists today, versus exactly what you described or what somebody could do with their own $20m or $40m or $60m portfolio.

Goodhart’s Law – or ‘When a measure becomes a target ...’

AE: This tangentially brings me on to something I was thinking about as I read your book and that is Goodhart’s Law – the idea that, when a measure becomes a target, it stops being a good measure. So, to come back to Buffett, he talks in terms of watching the scoreboard rather than actually playing the game. So I did wonder, if you start getting judged by these metrics, do you then start changing what you are doing? Take Robert’s colleague – if you told him to get his ‘hit’ rate up, that could destroy the whole way he makes his money. So could you talk a bit around this idea that, if this were to become the way you judge people, it could suddenly destroy the way somebody will make their money?

JP: Absolutely, Andy. Again, that is another very thoughtful question – and something I hope I have been thoughtful about myself over the years. In fact, in the very first book I wrote, I talked about ... and I’m not exactly sure of the wording of Goodhart’s Law but there is also Heisenberg’s uncertainty principle that what gets measured gets changed, right? Or you can even bring in Jane Goodall – that once you start observing nature, you are changing nature. So I spent a whole chapter in my first book, Trading Bases, on this and the chapter was called ‘Managing to the wrong metric’.

I took a whole bunch of different examples in different industries and some that didn’t even involve data per se – for example, Barron’s infamous ‘Amazon.bomb’ cover in 1999, where I argued they were looking at the wrong metric. At the same time, baseball managers in America were shunning the use of their best pitcher unless it was a ‘save situation’. Now, a ‘save’ is a metric that gets attached to each pitcher – so, instead of trying to win the game and using your best pitcher at the highest-leverage point of the game, they were always saving that pitcher for the last inning, because then it was a ‘save situation’. So I am very aware of that pitfall.

Still, here is a point of pride in the framework I have developed – I don’t think it can be gamed. And this is why I said that ‘hit’ rate and ‘slugging percentage’, as traditionally calculated, are worthless. For one thing, they can be easily gamed just like we talked about with your colleague earlier – you absolutely would have an incentive to not realise losses until they became ‘a couple of basis point’ winners and then sell them, right? Sell something anytime it is up and you are going to help your ‘hit’ rate. At the same time, slugging percentage is even worse – there is an even more perverse incentive, if you are trying to managing that because slugging percentage traditionally weighs the average of each winner over the average of each loser.

So you can tell – if you want to improve that ratio and therefore your slugging percentage – it would be helpful to move a winner into the loser category, as long as it is a very small loser, because now you have a small denominator, which would increase the ratio and therefore increase the slugging percentage. So, as I say, it is a point of pride that you can’t mess with my framework. I give acronyms based on American baseball players to the readings in the book – ‘Carew’ is the first reading and ‘Aaron’ is the second reading and the only way to increase your ‘Carew’ or ‘Aaron’ is to pick more winners – so, if you are trying to game it, you are going to have to be better at your job! You are going to have to exhibit more skill to game that. So, in that sense, I don’t think that problem exists with my framework.

RD: We observe a lot of managers in this building – how they trade and their performance – and one area we have spent a lot of time looking at is their execution of new positions versus their existing portfolio. We find the ‘hit’ rate on new positions tends to be consistently higher than on the residual portfolio – and there is an opportunity set to take advantage of that. Our experience suggests it may be human instinct to be cautious at first and therefore size up moderately and wait for confirmation signals and catalysts to reaffirm and then size up further. So there is quite a lot of P&L left on the table by the manager who is unaware that a) their hit rate is disproportionately better on their new positions and b) their sizing is slower than it could be so they might then actually add to their performance.

Also, to build on your gardening analogy, sometimes it is good to prune the residual portfolio to create that freshness. So I guess my point is you do want to share some data with the manager so they are aware of those things, because that is constructive. But then there are other things we look at that we don’t want to tell the managers about because we can’t use it or take advantage of it consistently – and, if we show them certain things, then they might change their process, which could dilute their ability. So I do believe there is value in this issue of Goodhart’s Law and measuring changing behaviour. It could actually lose your edge – but, in some cases, you need to let the manager know what their missed opportunity is and how they could do better because that is only accretive to them and to you. Equally, there are some mistakes they make that we don’t want them to change because we can take advantage.

AE: That probably ties in with something you write about in the book – that, if you look at sizing decisions or scaling decisions across the managers you have seen, it doesn’t actually add anything and you would actually recommend most people to equal-weight. Is that what you have found?

JP: Yes – but that is a bit simplified so I will amplify a little. And all I can say to what Robert just said is I was nodding vigorously at each of his points. Those views are consistent with what I have found – and, Robert, I sat in an internal-alpha capture division doing that work. So, of course, we were very interested in what we could capitalise on and there is no question about ‘alpha decay’ in terms of the initial starting point. As for the sizing of the position – and this gets right to sizing – the strongest signal from talented PMs is the inclusion in the portfolio. That is the money-making signal – so the idea of scaling into a position is folly.

Robert, I have a story in the book about how, fortunately, I learned that lesson at the hands of Dr William Sharpe – of Sharpe ratio fame – who was a professor of mine when I was in grad school. I won’t ruin it because I think it is a fun story in the book but he absolutely made all of us in the class, including me, look foolish with that exact decision to size into a position – in his example, it was an inheritance. I think it is a fun example in the book and it certainly changed my behaviour and the way I thought about things.

Staying with the sizing discussion – and sort of like when I talked about putting before, when of course there are skilled putters – it is not that sizing cannot contribute to alpha, right? It will always be part of the formula – but it will almost always have the smallest impact because institutional-quality investing generally does not allow for a PM to make a position a 25% position. That is what it might take to truly add a tremendous amount of alpha to the equation yet, in general, when you are talking about sizing in an institutional-quality portfolio, you are talking about a PM who has positions ranging from 1.5% to maybe 7.5%, right?

So there are two findings that came out of it and they are so strong across so many PMs – the first of which is that whether they added or detracted with their sizing decision versus equally weighted one year had no effect on prior years. In other words, it strongly reverts back to the mean of zero. So it is not that they can’t create alpha with sizing – it is that, if they did, that is not the repeatable portion going forward. And the same holds for the framework we built – you find a PM who makes two basis points a day in Carew and two in Aaron and is losing on sizing, that next one will be additive the next year when you start looking at a projection model.

The caveat to that is this, though: if the PM has less than two dozen positions – and that is a bit arbitrary; it is based a little too much on an eye test of mine, as opposed to rigorous statistical analysis – but, in general, the PMs I did see, who had sticky persistency in terms of sizing being an additive element of their alpha, they had very concentrated books. And I believe that is a factor of the human mind – the human mind cannot rank 70 ideas because too much art and too much noise comes into equity returns. That is why it is mean-reverting.

In the book, I borrow a quote from someone at a competitor firm I didn’t want to give away my secrets to – so I was gently saying that sizing tended to revert and he kind of cut me off. He laughed and said, Joe, we tell our PMs that if they spend 90 seconds thinking about sizing, they have wasted 60 seconds. And Robert, to the way my competitor’s firm looks at things, they don’t care if a PM is suddenly adding a ton of beta to the portfolio because they undo that at the top. So what they want the PMs to do is essentially, Go ahead and equally weight your Microsoft and your Black Box – we will take care of what that does to your expected volatility and your expected beta because all we are paying you on is your idiosyncratic alpha, right?

So that seems to be a consistent finding. Also, underscoring that idea that too many positions mess up the ability to rank efficiently – I did not do this work but it was done within my firm. They had access to what we called ‘analyst ranks’, right? The analysts would rank their universe – the analysts tend to only model six to eight companies and they tended to be in-depth and they tended to be very good in terms of their conviction by position. And then that would be another measurement rating we had for the PM – how good were they at converting their analyst work into what actually went into a portfolio? But, again, that underscored how it is the number of positions that inhibits the human mind from having any sort of effectiveness in ranking.

RD: If I could just add two quick things to what Joe was saying there – one is, we have done a lot of work looking at the success rate of analysts versus the follow-through from PMs and there are some very interesting insights, which we have shared with a number of the heads of desks. There is a lot of rich value in what Joe has just referred to and we do see some interesting paths that differ between analyst prowess versus manager follow-through. The other thing I wanted to mention was – and I am just smiling because I guess this podcast will not go too much to the outside world – but, when I worked at Soros, I was a bit taken aback at first by the lack of sophistication of the risk models because the portfolio model would only allow you to have five positions. I had been a PM, in equity land, with 35 to 40 positions – so how was I going to know what my risk and my sizing was?

I didn’t really understand why the system was so narrow. Later, though, somebody pointed out that George really thought there were only two or three things at any point in time that really mattered. So his whole view was that you should only have a handful of high-conviction ideas in the portfolio at any point in time and everything else is noise – and you should be forced to hold only two or three ideas and focus in on them because that is how you will make big returns. Then, in the firm, it became clear you would lose your job for either taking too much risk and making a mistake, or being right and not taking enough risk – so you suffered either way. Anyway, the point of the story was that actually being forced to focus in on those two or three things at one point in time is where you got 80% of your returns – and he forced people to do that by narrowing how many holdings they could have in their risk view.

JP: That is a great theory. What I can tell you is, as we know, I worked for a very well-known head-of-firm and it always used to be said the most effective portfolio he could give to his limited partners (LPs) – and, of course, he would be the largest LP – would be to have his 50 managers just have one long and one short. He would then have a diversified portfolio of 50 by 50 ideas, right. But it goes to what you said earlier about career risk – that would probably be the most profitable portfolio for the firm and the LPs but it would be the most high-risk in terms of career risk for each of the PMs so you wouldn’t get people to sign on to that.

Feedback or pushback? And a book recommendation

AE: We are getting towards the end of our time but, before we finish, you wrote in the book – and you mentioned it at the top of the podcast too – that part of the reason for writing it was you wanted these ideas exposed to the outside world and to be open to constructive feedback. What has been the best or most frequent feedback you have had on the system you describe in the book?

JP: Yes – let me first touch on a certain point of that too – in terms of why I wanted to expose it. It surprised me because I grew up on a trading desk, being a market-maker for hedge funds that were whipping stuff around all the time. I didn’t know what went on inside whoever it may have been – I just traded or made markets for all of them. So it was a surprise, when I got inside the firms, to realise that, in the modern sense, they are allocators. The head of the firms are allocators – because they have so many managers, their job is essentially to try to direct capital to the most effective PMs. Though, of course, there is turnover too, right? They do get rid of the ones that don’t perform.

In that sense, therefore, I realised they were performing the same role as my favourite clients at Novus, who I marketed to, and the other hedge funds and pension funds, right? That is what they are doing – they have de facto multi-manager platforms and I was never really pleased with how they performed manager selection. So this book was aimed at them as an audience as well – like, Hey, here is what goes on inside the really sophisticated allocators, even though you think of them as multi-manager hedge funds, and I think it can help you as well. So that also touches on who I was aiming the book at.

As far as pushback goes, allocators don’t like to be told they don’t do manager selection well – because they spend so much time doing it, right? Nobody likes to be told, fear of change is a very real thing and narratives get embedded for a reason. So I think there is a little bit of pushback there, in terms of career risk – you know, We are not going to change, Joe, just because you think you have built a better mousetrap or that there is some innovation! So I have certainly gotten some pushback from allocators as to whether there is any real value here that they can apply.

At the PM level, meanwhile, the questions always come back to sizing and they always say, Hey, my biggest position did do better last year. And I am like, OK – but remember, when you’re judging that, you have to judge your smallest positions too. They have to all have done better to make that judgement. And I’m not saying you should start every position equally weighted – what I am saying is: spend a lot less time at it; start them equally weighted and just let them run a little bit. So I get pushback on sizing and I certainly get pushback on the idea of daily measurement because PMs will say, I don’t care what happens on a daily basis – judge me quarterly or yearly. And that is fair, right? You wouldn’t make any decision on a manager based on what they did in a single day.

And they will always tell a story – and I have a chapter in the book about this – about a stock of theirs that languished for nine months of the year and then either passed an FDA trial or got taken out or something else. And they will say, I was early and I was right. And that is an emotionally satisfying result for a PM but it is not repeatable – and, in the book, I try to go into why that is. That is where the daily readings come in – it is not making a decision based on how they do on a single day but it is gathering a data point that ends up in a series that starts to have use after six months, with a strong regression to the mean to fill in the missing observations. Once you get to two years, however, you now have 500 data points and many PMs are only giving their allocators a quarterly PDF, right? So I have got 500 data points versus eight data points and you can get a lot of insights from those 500 data points because hopefully, like I said, I am measuring skills and have determined the skills have some stickiness outside of noise.

So that is the pushback. I love the discussion. I love when people are sceptical about the predictive nature of the work – because then, if it is ultimately proven to be true, it has value, right? After all, if they were not sceptical to start with, then it is just something people already know. So that is the part I love. And what I tell people is, We have old data so give me the PM data from 2019 and 2020 and I will tell you how they do in 2021 and 2022. And then then we will look and see if it is any improvement on your existing work, which is still largely looking at past results – even if you are looking at Sharpe ratios. You know, there are some signals on past results – but they are still past results.

RD: Andy, I know you have more to ask Joe but I have one quick question for him – and I apologise, Joe, as I have not read your book yet. Still, I think we both share the view of reversion to the mean being a natural cyclical path – hopefully, around secular good performance of managers – but, from your experience, what would you say is the average half-life of above-mean performance you would bank for a manager before that mean reversion kicks in? I guess the answer will differ for each one but, in the round, what do think that persistency period is?

JP: In the round, it is 500 days – in other words, adding more data does not help the projection going forward. I do talk a little bit about that in the book but I can’t say why that is. I do know, if I data-fit prior information, I might find out 530 days is the best number to start at for 2023 – but then I also know that is too exact. I can tell you that other quants I have talked to also seem to support the idea there are market regimes that go on and that, over two-year periods, you are hopefully catching the end of one and the start of another so you are getting a blend of all the PMs’ skills. So that is what I found. And I have found it starts becoming predictive once you have six months of data – or, like I say, 125 observations – and then you fill in those other 375 with simply the mean and then they get replaced each time you record a new daily observation.

AE: Thank you very much, Joe. Robert and I really appreciate you not just writing the book and exposing these ideas but then being so willing to come on the pod and discuss them. Now, we always like to finish up by asking our guests if there are any books – apart from your own, of course! – which you like to recommend or which have been particularly pertinent to your career?

JP: There will be a recency bias to my answer – but hopefully that is also most helpful to listeners. I imagine most people who talk about what we talk about will be familiar with [former TVP Pod guest] Michael Mauboussin and The Success Equation – which is a ‘must’ starting point for anyone on this topic – and Annie Duke, who has written a lot now on decision-making [and has, twice, been a TVP Pod guest]. But the recent one I read that is of that realm is The Biggest Bluff by [another former TVP Pod guest] Maria Konnikova.

She is a professional writer, so she is good at writing, but she also has a really interesting story – she is a Harvard-educated psychologist, who takes up poker, and her background in psychology is very helpful in dealing with human ‘tells’, the misogyny that can go on in poker rooms and so on but it is really a story about decision-making. And she is very good at telling the story. It is great, of course, because it is about poker and money but she didn’t know if she would have the capital allocation gene or the risk tolerance gene. It is just a really fun book – something I have read within the last year and really enjoyed.

AE: Yes, it is a great book – and Juan, who usually hosts these podcasts, will be very happy you picked The Biggest Bluff because he has given it as reading material to our newest joiner and they are having a ‘mini book club’ on it. Thank you very much for coming on the show, Joe. And thank you, Robert, also for sharing your insights. It has been a great discussion. We really appreciate you coming on. Thanks.

JP: Thank you, Andy and Robert – I love talking about all this. Thank you very much for having me.