tag:blogger.com,1999:blog-261139923818144971.post3564099080554650750..comments2024-05-18T05:54:54.563+01:00Comments on This Blog is Systematic: Random data: Random wanderings in portfolio optimisationRob Carverhttp://www.blogger.com/profile/10175885372013572770noreply@blogger.comBlogger5125tag:blogger.com,1999:blog-261139923818144971.post-7945445491570756882023-07-04T14:25:47.449+01:002023-07-04T14:25:47.449+01:00Thanks vm, very usefulThanks vm, very usefulUnknownhttps://www.blogger.com/profile/06398352602826967887noreply@blogger.comtag:blogger.com,1999:blog-261139923818144971.post-1412726590624003632023-07-04T08:25:22.642+01:002023-07-04T08:25:22.642+01:00That's exactly what I would expect - in theory...That's exactly what I would expect - in theory you get more independent bets the faster the trade and therefore according to the law of active management (grinold and kahn) your SR goes up.<br /><br />"My intuition, based on maths was that the longer the horizon, the higher the Sharpe should be (as StDev only increases with the SQRT of time)." yes but that only implies that for a given underlying SR, the SR calculated over longer periods will increase eg a daily SR of 0.10 translates to an annualised SR of approx 1.60Rob Carverhttps://www.blogger.com/profile/10175885372013572770noreply@blogger.comtag:blogger.com,1999:blog-261139923818144971.post-34638831769790294712023-07-04T08:22:28.110+01:002023-07-04T08:22:28.110+01:00Hi Robert, love what you do!
I've tried to re...Hi Robert, love what you do!<br /><br />I've tried to replicate-ish your idea, to check, for a given system, which trading horizon would produce the highest Sharpe ratios. So I backtested the system across many different instruments, for 4 different trading speed each time (not accounting for costs): 1min, 5min, 10min and 30min. My intuition, based on maths was that the longer the horizon, the higher the Sharpe should be (as StDev only increases with the SQRT of time).<br /><br />But what I observe is the opposite: the shorter the horizon, the higher the SR.<br /><br />I am a little bit surprised and this sparked a debate in my team. What would you say?Unknownhttps://www.blogger.com/profile/06398352602826967887noreply@blogger.comtag:blogger.com,1999:blog-261139923818144971.post-63263013008774078332021-11-01T08:53:55.589+00:002021-11-01T08:53:55.589+00:00I haven't read the paper, but it makes sense. ...I haven't read the paper, but it makes sense. Shrinking correlations makes you less sensitive to small differences in means and standard deviations - not an empirical result, just a property of the maths.Rob Carverhttps://www.blogger.com/profile/10175885372013572770noreply@blogger.comtag:blogger.com,1999:blog-261139923818144971.post-91351562917322804442021-10-25T18:22:35.998+01:002021-10-25T18:22:35.998+01:00Have you read Pedersen's "Enhanced Portfo...Have you read Pedersen's "Enhanced Portfolio Optimization" paper? He finds that correlation shrinkage fixes errors both in the covariance matrix and in the expected returns. I'd like to know what you think, since his results look quite different from yours (obviously the results depend on the assumptions, but still...). Thanksyounggottihttps://www.blogger.com/profile/09011475435226933830noreply@blogger.com