DISQUS

TPC · 5 months ago

Cool stuff Jay. I'll be interested to see how it performs.

Gerd · 5 months ago

Jay-

How confident can you be regarding a portfolio composed on a retrospective basis? I'm sure you wouldn't have included any managers that had a tough time last year, but how would you have known not to include Tontine, Glenview or Sowood had Alphaclone's product come out a year earlier? I'd be interested in your take on this.

marketfolly · 5 months ago

hey Gerd, you bring up a very good point there that has been addressed a few times. Survivorship bias is an interesting topic, isn't it? While there is no way for me to obviously prove that I would have selected the same managers, I would like to think avoiding a few of the mishaps would be somewhat easy and I'll tell you why.

Firstly, I like to focus on funds that do the bulk of their transactions in equities. These are the easiest funds to track and in turn, clone. We focus on a lot of the Tiger Cub and value/activist player portfolios as they often have concentrated equity portfolios with a lot of transparency in very liquid positions with a long-term investing time frame in mind. Now, take Tontine on the other hand. They have a more macro theme strategy approach and then Gendell tries to place bets based on these themes. There is more risk in cloning a portfolio like that because he could very well be right on a certain secular theme, but be completely off in his selection of the investment vehicle. Secondly, he often takes illiquid positions and we are not a fan of that at all for cloning purposes. We want liquid stocks that are for the most part known or well-known.

Another example is trading funds like SAC or Rentec, or then global macro firms like Tudor or Clarium. We can't accurately track trading firms since the 13F's are on a lagging basis, so that immediately tosses them out of the picture. Macro funds have too many holdings in other markets outside of equities and so you can't purely replicate them either.

There is obviously always a risk you will run into a fun that blows up. And, while I would like to think the funds we've selected wouldn't blow up... you never know they still could. I am confident I would have selected Baupost still because, let's face it, Klarman's a value legend and is the perfect subject for cloning. Additionally, in my years of tracking the Tiger Cubs, I've noticed a trend where spin-offs or "youngsters" starting their own funds tend to outperform, especially in their early years. As such, when Shumway left Tiger, I would have selected him... (after seeing success from Griffin when he left Tiger or Mandel when he left Tiger).

To back that statement up, I'm coming up with another portfolio that includes David Stemerman's Conatus Capital (an ex-Lone Pine manager). That's a sneak peak at what I'm working on, but it focuses on managers who strike out on their own. His fund just started and I will have invested in that clone, similar to as if I had picked Shumway right from the start.

In short, there's no real easy way to completely rule out survivorship bias when selecting these funds based on backtesting, you've hit the nail on the head there. However, our familiarity with various hedge funds and our specific guidelines for which types of portfolios we would clone immediately eliminate many potential funds from our portfolio. As such, we have a more limited selection to choose from.

I guess another fact supporting all of this would be that I cloned numerous other portfolios that actually had more success in terms of performance numbers on an annualized basis, etc. However, I did not select them as the MF clone because I could never see myself actually investing in the mix of fund managers those particular clones generated. So, there's no way to avoid it but I tried to take a realistic approach by applying the criteria and principles set out for cloning our portfolio and tried to focus on those parameters rather than just on performance. Because, as we all know, past performance is by no means an indication for future performance. We didn't try to focus on performance. Sure, we wanted good performance, but we started our portfolio with the set of criteria in mind and it was essential that we met those before even beginning to toy with the clone to generate performance metrics that were satisfying as well.

Great topic for discussion and I'd love to hear your thoughts on the matter as well.