tag:blogger.com,1999:blog-261139923818144971.post3546105431182257030..comments2024-03-29T07:20:07.753+00:00Comments on This Blog is Systematic: CTA allocations, QE, meta-prediction, and conditional return distributionsRob Carverhttp://www.blogger.com/profile/10175885372013572770noreply@blogger.comBlogger11125tag:blogger.com,1999:blog-261139923818144971.post-37707547272124288972019-07-31T16:56:04.260+01:002019-07-31T16:56:04.260+01:00Thanks so much, love your simple explantations.Thanks so much, love your simple explantations.douggiehttps://www.blogger.com/profile/13670295647801338012noreply@blogger.comtag:blogger.com,1999:blog-261139923818144971.post-58448976899229740642019-07-31T15:51:23.468+01:002019-07-31T15:51:23.468+01:00The raw carry signal is *very* noisy, and at any f...The raw carry signal is *very* noisy, and at any frequency of less than a day it's likely you are just seeing pure noise since the non traded contract in many markets may not even be trading, or you may be picking up prices that aren't exactly aligned. I would advise using daily carry, and then taking a smooth of that. I use EWMA smooths on the raw carry signal from ~10 days halflife up to 6 months halflife.Rob Carverhttps://www.blogger.com/profile/10175885372013572770noreply@blogger.comtag:blogger.com,1999:blog-261139923818144971.post-83865384272737843072019-07-31T09:57:40.200+01:002019-07-31T09:57:40.200+01:00Hi Rob, great post as always, looking at the carry...Hi Rob, great post as always, looking at the carry and EWMA rules, the EWMA tend to have multiple timeframes/lookbacks, however for Carry it is just daily. I did a quick test checking price differentials every 3,6,12 or 24 hrs for the carry rule and it does seems to produce different correlations between timeframes, however a given time frame is correlated about the same for across a set ewma lookbacks as you would expect. I was just wondering intuitively why the carry forecast is just computed daily, it just because the price differentially is not effect by timeframe in the same way ewma is affected by the lookback period, i.e. the price difference would not vary widely if I was to check it every 12 or 24 hours.douggiehttps://www.blogger.com/profile/13670295647801338012noreply@blogger.comtag:blogger.com,1999:blog-261139923818144971.post-21546434116450255602018-04-21T11:03:24.800+01:002018-04-21T11:03:24.800+01:00You're probably right about Natgas which is wh...You're probably right about Natgas which is why for Oil I track a consistent winter contract, obviously the liquidity for Natgas isn't as good. The results are interesting, but I don't know how good if they were done properly: on an out of sample basis fit the sign and magnitude of the filter to apply to each trading rule, and do this on a market by market (or at least asset class specific) basis - that's a *lot* of extra parameters to fit.Rob Carverhttps://www.blogger.com/profile/10175885372013572770noreply@blogger.comtag:blogger.com,1999:blog-261139923818144971.post-40312123518097395922018-04-21T09:34:19.833+01:002018-04-21T09:34:19.833+01:00Firstly, congratulations on the Top 60 award! It&#...Firstly, congratulations on the Top 60 award! It's well deserved. On meta-prediction it was a bit disappointing to hear that you might not do anything with these results. There seem to be some rich possibilities here. My guess is that volatility could be another good conditioning variable. Trend following has not worked well for Natural Gas probably because it jumps from high volatility periods in winter to lower volatility in the summer and so a generic model fails. With interest rate environment I think you're onto something that could systematically improve performance.Steve Bellhttps://www.blogger.com/profile/17708635422483072210noreply@blogger.comtag:blogger.com,1999:blog-261139923818144971.post-80137423424782565982018-04-07T11:53:27.627+01:002018-04-07T11:53:27.627+01:00Got itGot itjamesethaanhttps://www.blogger.com/profile/17209422228302823058noreply@blogger.comtag:blogger.com,1999:blog-261139923818144971.post-15031390273375518612018-03-21T08:27:26.549+00:002018-03-21T08:27:26.549+00:00When rates are falling modestly the rolldown and y...When rates are falling modestly the rolldown and yield change tend to cancel out resulting in a total return series that goes sideways.<br /><br />In bonds carry is a very high beta strategy, very close to long only with some timing properties. So in an environment with modestly rising interest rates carry will tend to be flat.<br /><br />The sideways move of total return is particularly bad for momentum as it makes prices choppier.Rob Carverhttps://www.blogger.com/profile/10175885372013572770noreply@blogger.comtag:blogger.com,1999:blog-261139923818144971.post-23631838867825286802018-03-20T23:32:42.395+00:002018-03-20T23:32:42.395+00:00What do you think would be the drivers for these k...What do you think would be the drivers for these kinds of results? It seems intuitive that a long only bonds would do poorly ina rising rate environment (falling prices). But what do you think would explain the cta and momentum and carry having noticeably worse sharps in a rising rate environment? E.g. do prices “trend less” rendering the signals less effective? Interested in your thoughts.JMW100https://www.blogger.com/profile/18171394406805453830noreply@blogger.comtag:blogger.com,1999:blog-261139923818144971.post-37427399451211396022018-03-17T09:48:29.560+00:002018-03-17T09:48:29.560+00:00Got it, thanks. I guess there is always some desig...Got it, thanks. I guess there is always some design discretion. I suppose the trick is to try to choose one without peeking at the results first.Patrickhttps://www.blogger.com/profile/01034323689613977187noreply@blogger.comtag:blogger.com,1999:blog-261139923818144971.post-11426921843539578962018-03-15T09:47:43.081+00:002018-03-15T09:47:43.081+00:00Finding the right window size is a matter of balan...Finding the right window size is a matter of balancing two conflicting factors: the better statistical significance of longer windows versus the more up to data of shorter windows. For relatively slow trading rules of the kind I use we normally can't find enough data even without restricting ourselves to artifically shorter periods.Rob Carverhttps://www.blogger.com/profile/10175885372013572770noreply@blogger.comtag:blogger.com,1999:blog-261139923818144971.post-36216310588999736962018-03-15T09:30:55.175+00:002018-03-15T09:30:55.175+00:00Hi Rob,
This is a truly excellent and original po...Hi Rob,<br /><br />This is a truly excellent and original post, which is hard to find elsewhere, so like others, I am sure, I remain truly grateful for your continued sharing. One assumption in the post which got me thinking is that of the 10 year cycle of (normalised) interest rates. If this is correct, might it be preferable to use 10 year rolling windows, as opposed to expanding windows, for weight optimisations across assets? I currently have about 50 years of price data (albeit with a sparse set of instruments for the first 25 years), so rolling windows are appealing. Although arguably exponential smoothing of weights downweights earlier periods in either case, rolling does it more explicitly.Patrickhttps://www.blogger.com/profile/01034323689613977187noreply@blogger.com