Discussion about this post

User's avatar
Levi Ramsey's avatar

The generalization of Kelly is (for the quants) essentially to maximize the expected logarithmic returns over the subperiods. One implication of this approach is that it seeks to avoid large drawdowns (as larger and larger drawdowns have unbounded negative logarithms).

So even a full-Kelly approach will tend to keep a substantial cash allocation or perhaps allocate some to puts (about the only strictly defensible 100% equity full-Kelly portfolio would treat conservative estimates of the price that would ensure a PE bid as cash equivalent: something like positive net-net (more current assets than total liabilities) might approximate that). Kelly bettors in games where there are definable odds generally practice half-Kelly which is exactly equivalent to 50% in cash and 50% in a full-Kelly portfolio (which itself will have some cash allocation!), if nothing else to help account for uncertainty around whether you've correctly figured out the payouts and probabilities.

The key point is not only to think about the odds and payout of success, but to consider the odds and payout in the worst-case scenario (and consider those at least as much as the success case). This is pretty easy to analogize to the Graham & Dodd "margin of safety" (and there's at least some research indicating that Buffett sizes bets in a Kelly-esque manner: he also endorsed Ed Thorp as a good manager for his limited partners who didn't follow him into Berkshire (someone who took their money from Buffett and put it into Thorp's fund and then bought BRK when Thorp stopped managing other people's money may actually have done better than someone who stayed with Buffett)).

I've toyed with the idea of an equal-weight portfolio that sells flowers and waters weeds but when the sales (plus dividends, covered call premiums, etc.) of a flower exceed what you put in by enough to have gotten an acceptable annual return, you exempt it from equal-weighting and only sell enough to keep that realized return profile (until you decide it doesn't deserve any weight).

Expand full comment
Emerging Value's avatar

There is a time element to it. I started my portfolio in 2012 more or less, I bought some stocks, and I still find new ideas in 2023, so I could not have started my portfolio with 15 equal weights Ideas. I did not even have 15 ideas when I started! So new ideas get added progressively, and their weight depends not only on confidence, but also on available funds and saving rates, and things to sell.

On top of that I am willing to add less weight on a new idea if I just discovered the company. I like to have followed it for two years to really feel confident to concentrate more.

Very complicated.

Expand full comment
14 more comments...

No posts