What if I told you... you don't need an edge to make money?
Yet another thread on Shannon's Demon
Yet another thread on Shannon's Demon
Continuing my ramblings from this thread
Since the stocks move in random walks, by definition you cannot have an edge. Amazingly, it's even possible to extract a profit from stocks that both lose money over time.
I'm still not 100% sure this is real. It looks like alchemy to me. Like one of those trick algebra problems that reduce to 1=0 because you divided by zero somewhere. I'm not a quant, and I'm still struggling to wrap my head around it, so I apologize if any of this is wrong/dumb.
What are the requirements for Shannon's Demon? You need the stocks to be volatile and uncorrelated. Going down this rabbit hole led me to some excellent blogs by @KrisAbdelmessih. Not actually about SD, but really important concepts for every portfolio. moontowermeta.com
Here's another post to delve deeper. My main takeaway is that the value of a security is highly dependent on what else is in the portfolio. A portfolio is greater than the sum of its parts. SD is a simple example of how that value can materialize.
moontowermeta.com
moontowermeta.com
IMO the importance of correlation is poorly understood, and while everyone knows diversification is a free lunch, it's even more powerful than people realize. Most large cap stocks are highly correlated and so the diversification benefits are muted.
What if you could take diversification to the extreme? Combine many uncorrelated things? Not just stocks, but also options, spreads, baskets, and entire strategies. Could you then apply Shannon's Demon to the whole portfolio?
An example of this was LTCM, who traded many strategies with low correlation. Each of the strategies had relatively unimpressive looking returns on their own, but combined into a portfolio smoothed out the volatility. Then with leverage added, produced spectacular returns.
But they were undone by a spike in volatility and correlations. They were a victim of their own success since the fact these were known LTCM trades arguably caused them to become correlated.
LTCM volatility, correlations, and losses. (From the book The Crisis of Crowding)
LTCM volatility, correlations, and losses. (From the book The Crisis of Crowding)
LTCM's Sharpe ratio reached 2-4+ before they blew up. Many uncorrelated strategies each with mediocre Sharpe ratios combined into a portfolio can have a surprisingly high Sharpe. And that sounds much easier than finding one high Sharpe strategy.
And I guess that's what some quant funds have done or tried to do since. Find relatively marginal anomalies but get a lot of them and then combine them into a portfolio. The diversification benefits make the drawdowns so small that they can (hopefully) safely leverage up.
But is this relevant for peasants like myself just messing around in the PA? I think there are still things to learn. My opinion: edge isn't everything. Concentration is overrated. Most of us aren't the next Buffett and the high conviction style is glorified by survivorship bias.
There is lower hanging fruit in achieving excellent diversification, especially for most PAs, which are way too concentrated. Not just in terms of which stocks they hold, but also in terms of sectors, strategies, factors like value/momo/short interest, etc.
Diversification reduces risk, but can also itself create returns, as SD demonstrates. It's unintuitive, and I think related to another unintuitive concept, that according to the Kelly theorem, reducing position size can increase returns, even for very high expectancy trades.
Claude Shannon remarked that his Shannon's Demon concept was just theoretical because in reality the tx costs would eat up all the profits. But with fees falling to zero and the proliferation of securities like options, maybe some doors have just recently opened. End thread.
Further reading: thread by @Valuegroupie
Loading suggestions...