STREET DOGS

From the Nomad Investment Partnership Interim Letter, 2004:

I f we could find 50 ideas at equal discounts to value, with equal probability (conviction) of value being realised, then they could all be equally weighted in the Partnership. We could then look forward to a nice smooth rise in the value of our shares, free from the swings a more concentrated portfolio might create.

But life is not like that. In reality, opportunities in which we are comfortable to deploy capital are rare, and the highest conviction ideas the rarest of them all. The issue then is how much to invest in each idea.

Bill Miller, who has run the Legg Mason Value Trust so brilliantly for many years, suggests the use of a system devised in 1956 by

JL Kelly. A simplified version of the Kelly criterion is that investors should bet a proportion of the portfolio equal to 2.1 x p - 1.1, where p is the probability of being right. The common-sense outcome of this equation is that if one is certain of being right, one should invest the entire portfolio in that idea. Even if one is, say, 75% certain of being right, the correct weighting remains high at 47.5% ((2.1 x 0.75) – 1.1).

But does anyone do that? As far as we are aware, only the early Buffett Partnership portfolios had anywhere near this level of concentration … But if you know you are right, why would you not bet a high proportion of the portfolio in that idea?

Apply the Kelly criterion, and the average fund manager would appear to have almost no clue as to the likely success of any one idea. In our opinion, the massive overdiversification that is commonplace in the industry has more to do with marketing, making the clients feel comfortable, and the smoothing of results than it does with investment excellence.