Thread by @ReformedTrader

1/ Superstar Investors (Brooks, Tsuji, Villalon)

"Identify structural edges and commit to seeing them through inevitable periods of underperformance. As each of our superstars shows, “merely good” edges over time compound to great long-term performance."

aqr.com/Insights/Resea…

2/ Looking at managers' factor exposures based on their stated investing philosophies may underestimate alphas:

* we have the benefit of hindsight
* factors are gross of costs
* factors may be data-mined

They may also overestimate alphas:

* Only the best managers are analyzed

3/ Based on Berkshire Hathaway's stated investment philosophy, AQR chose to look at BRK's α after controlling for the MKT, HML, BAB, and QMJ.

"Buffett does not talk about momentum... We exclude it in our analysis and thus reduce the likelihood of potentially spurious results."

4/ Berkshire Hathaway's factor exposures are all highly statistically significant, but the regression α is not.

After accessing cheap leverage through the insurance business, BRK was able to provide full beta 1 exposure to the stock market plus an additional 1.3 factor betas.

5/ AQR has a more detailed paper that looks at BRK's factor exposures and also compares the stock's performance to that of a hypothetical factor-based "replica" portfolio.

https://twitter.com/ReformedTrader/status/1128474864770138112

6/ For Bill Gross/PIMCO Total Return Fund:

* Market = US Aggregate Bond Index
* Credit = 5-year US High Yield CDX
* Low-risk = long 2 & 5 year and short 10 & 30 year Treasury futures, duration-neutral
* Short vol = short 1-month, 30-delta strangles on 10-year Treasury futures

7/ As with BRK, the factors were picked based on Gross's descriptions of his own strategies. (Factors are hypothetically systematic ways of accessing the premia.)

Gross's factor exposures are statistically significant, but the alpha after controlling for those factors is not.

8/ Based on the focus on trends and currencies Soros/Quantum Fund are known for, AQR uses

* MKT
* Trend (UMD for stocks and time-series momentum for asset classes)
* Currency momentum+ = "fundamental" momentum based on changes in non-price measures
* Currency value based on PPP

9/ As with Buffett and Gross, Soros's factor exposures are statistically significant, but the alpha after controlling for those exposures is not.

10/ Peter Lynch/Magellan had a more complex set of strategies that was harder for the authors to distill into a set of factors. They ended up picking some of the more commonly used academic factors:

* MKT (equities and bonds)
* SMB
* HML
* UMD
* QMJ
* BAB

11/ For Lynch there is a large, statistically significant alpha left over after controlling for those factors.

This, combined with the fact that Magellan's *CAPM* alpha was no longer stat. significant after Lynch left, suggests an unusual level of skill on Lynch's part.

12/ With respect to market timing:

* BRK - likely none
* PIMCO - none for duration but some for credit
* Quantum - some for equtity beta
* Magellan - likely none

13/ This is consistent with Jack Schawger's books:

* Impressively, "factors" were used before they were factors
* There are *lots* of ways to invest, suggesting that almost no one is fully diversified
* Even super-star managers have a lot of market beta

https://twitter.com/ReformedTrader/status/1217655775650181122