I'm not assuming anything. This is just maths (and it's not my own maths, it's standard statistics). The variance of parameter estimates for correlation and volatility are much lower than they are for means and Sharpe Ratios. In plain english we need less data to be more confident about a correlation estimate than for a Sharpe Ratio estimate. You might find that surprising and inconsistent, many people do, but it is a fact. That doesn't mean that correlations won't change over time, or become temporarily unhinged as in LTCM type crisis, or indeed in 2008 when the returns of CTA's like Cantab or Winton with equities went from mild positive correlation to massively negative correlation (and a good thing too). All it means is that we have a better idea of what past correlations were historically than we do about past Sharpe ratios. It also means we have a better chance of forecasting future correlations than we do of forecasting future Sharpe Ratios. GAT
If you don't believe me then from https://www.quora.com/How-do-you-calculate-the-variance-of-the-sample-correlation-coefficient You can work out the variance given r=-0.1 for yourself. I'll check your answer GAT
Oh yes you are assuming too much. The fact that they are less variable is cold comfort to the rational investor. The point is that 10-20 years of data is nothing, and has been proven to provide a false sense of security. Cantab returning 0% over the last 4 years should therefore cause one to be highly skeptical of claims like yours that it's still a good idea to invest with funds with a profile like that. https://www.jpmorgan.com/cm/BlobServer/AM_Non-normality_of_market_returns.pdf?blobcol=urldata&blobtable=MungoBlobs&blobkey=id&blobwhere=1158543264199&blobheader=application/pd JPM Asset Management admitted as much 10 years ago: "The results are striking. Nearly all the correlation coefficients increased to some degree (during this period of high volatility), with only four showing a decrease. Our main diversifiers did not provide the diversification benefits anticipated" They must have been a bunch of dumb investors too, ignorant of basic statistical truths... You contradict your earlier point by acknowledging this. Sorry, can't have your cake and eat it too. When correlations change, people get hurt in the real world. And the only reason why Cantab and other CTAs went from mild positive to massively negative correlation in 2008 is because there was a sustained uptrend in bonds thanks to QE. Next time it won't be so easy, given recent developments and the Fed's admission that they don't know what they're doing with inflation.
When or if trendies (or value investors, or gold bugs...) will ever get their mojo back is a matter of religious belief. No one will convince anyone else. Who knows what 2018 /19 /20 will bring? There may be trends, or there may not. Expect the unexpected...
I have another misgiving about CTAs (and similar) that I find difficult to let go of, about which I'd be curious to hear peeps' thoughts (if any). Specifically, it appears that these strategies are inherently of the 'greater fool' variety. In other words, the economic significance of the investment decisions made is often not a meaningful consideration. Instead, it's mostly about predicting, with varying accuracy, who is likely to buy/sell the relevant instrument next. This violates one of my most basic risk management principles, so I find it hard to muster the necessary enthusiasm.
I saw the returns of those funds. I can see how allocated with an equity portfolio would be a good diversifier for a pension or an endowment: someone who has a fixed income obligation higher than the long term risk free rate.
I prefer to live in the Stone Age myself than to trust any of these statistics. And before anyone says anything, I hold a degree in a mathematical science from a top 5 WORLD university.
'World' top-5 or top-10 rankings are mostly US. Maybe you mean non-US? Some UK schools plus ETH-Zurich?