svquant wrote:From an older Winton pitchbook - perhaps this helps? May have an even more detailed presentation from back when they had less than $500MM AUM. Note as they have grown the marketing message and details have changed greatly but I have seen version of this slide in presentations as late as 2009.
While many CTAs do not talk about moving averages as a forecasting methodology, i.e. just look for crossovers, it is a very well developed field. Just look at any time series forecasting text book and you can see how others look at exponential moving averages for example.
So the mechanisms shown by Winton could all be done via single or double moving averages, i.e. well recognized LTTF techniques.
From that picture, it seems he's employing the Markovitz mean-variance optimization. I fail to see how it is possible to forecast the 'expected returns', 'expected correlations' and 'expected volatility' without succumbing to the hindsight bias.
Edit:
Just read some of the comments by Chris67 and I have to agree.
With regards to volatility forecasting, while it may be perceived that it is easier to forecast market volatility opposed to market returns, conventional vol forecasting methods such as GARCH, ARIMA and VARIMA are also extremely misleading.
Edit 2:
If Winton is using those techniques to allocate across strategies, then it would make sense.