tag:blogger.com,1999:blog-261139923818144971.post4780569798384209365..comments2024-03-27T07:58:49.946+00:00Comments on This Blog is Systematic: Investment and trading performance - year threeRob Carverhttp://www.blogger.com/profile/10175885372013572770noreply@blogger.comBlogger23125tag:blogger.com,1999:blog-261139923818144971.post-25207922910063189332017-10-27T15:50:27.928+01:002017-10-27T15:50:27.928+01:00Thanks always nice to be appreciated.Thanks always nice to be appreciated.Rob Carverhttps://www.blogger.com/profile/10175885372013572770noreply@blogger.comtag:blogger.com,1999:blog-261139923818144971.post-18854313636166986562017-10-24T04:09:40.632+01:002017-10-24T04:09:40.632+01:00Just a quick ty:
I'm reading through your tut...Just a quick ty:<br /><br />I'm reading through your tutorials on the IB API. Since I'm just starting out in Python, too, the level of detail is exactly what I need.<br /><br />And now, while clicking through your blog, I'm seeing your humor established in a level-headed view of quant, "this stuff will never be as interesting as a decent book or a good bottle of red wine." <br /><br />In addition to wanting to be free from some kind of wage-slavery through automated capital management, I've always wanted to find ways to free others to purse their dreams instead of only their necessities. My thanks to you for walking through these posts like a sensei. I should follow this lead and convey what I know among those who care to listen.Anonymoushttps://www.blogger.com/profile/02296779215465206349noreply@blogger.comtag:blogger.com,1999:blog-261139923818144971.post-53639215371478837332017-08-14T11:33:56.059+01:002017-08-14T11:33:56.059+01:00I don't trade ETF/stock risk in non base curre...I don't trade ETF/stock risk in non base currencies (this is covered in my new book, out soon). For futures only margin is subject to currency risk. I keep the minimum amount of margin in foreign currencies to reduce this problem. I don't bother hedging this small exposure.Rob Carverhttps://www.blogger.com/profile/10175885372013572770noreply@blogger.comtag:blogger.com,1999:blog-261139923818144971.post-29085489398745600352017-07-30T01:22:58.557+01:002017-07-30T01:22:58.557+01:00Hi Rob
Would be interested in seeing more informa...Hi Rob<br /><br />Would be interested in seeing more information on how you hedge currency risk. Trading stocks or futures in non base currencies. Do you do this systematically? Is it covered in your books?<br />DRhttps://www.blogger.com/profile/05803821217741972701noreply@blogger.comtag:blogger.com,1999:blog-261139923818144971.post-3087039915787990622017-07-19T06:05:12.495+01:002017-07-19T06:05:12.495+01:00I hate leveraged ETFs. I hate ETNs. I hate expensi...I hate leveraged ETFs. I hate ETNs. I hate expensive ETFs. So I wouldn't touch them.Rob Carverhttps://www.blogger.com/profile/10175885372013572770noreply@blogger.comtag:blogger.com,1999:blog-261139923818144971.post-48617463076021009242017-07-19T03:44:23.518+01:002017-07-19T03:44:23.518+01:00Hi Rob,
would you consider holding these 2 ETNs (o...Hi Rob,<br />would you consider holding these 2 ETNs (or similar things) in your long-term portfolio in small quantities: MORL and BDCL? <br />http://www.etf.com/BDCL<br />http://www.etf.com/MORL<br /><br />Both ETNs appear to be highly-speculative but with very attractive yields (>10%) and seem to provide some narrow exposure to specific(risky?) market-segments, they're also 2x leveraged and cray a UBS counter party risk, as I understand. Expense-ratio is also of course higher than simple passive tracking ETFs..Dmitryhttps://www.blogger.com/profile/17093240013111409494noreply@blogger.comtag:blogger.com,1999:blog-261139923818144971.post-2113896884938664632017-06-12T06:15:09.116+01:002017-06-12T06:15:09.116+01:00It's a way to avoid trading too much. In my bo...It's a way to avoid trading too much. In my book I call it position inertia. You don't trade unless the theoretical position is more than 10% away from what you currently have.Rob Carverhttps://www.blogger.com/profile/10175885372013572770noreply@blogger.comtag:blogger.com,1999:blog-261139923818144971.post-4395913413633010712017-06-11T14:40:30.761+01:002017-06-11T14:40:30.761+01:00What is 'buffering' and what is the 10% in...What is 'buffering' and what is the 10% in this context?Anonymoushttps://www.blogger.com/profile/08604280702492556189noreply@blogger.comtag:blogger.com,1999:blog-261139923818144971.post-54782129101714415752017-05-16T18:07:04.948+01:002017-05-16T18:07:04.948+01:00Yes, but it's quite an involved calculation wi...Yes, but it's quite an involved calculation with a lot of different inputs. 10% is a conservative value that will be fine in most cases.Rob Carverhttps://www.blogger.com/profile/10175885372013572770noreply@blogger.comtag:blogger.com,1999:blog-261139923818144971.post-56973602877291133252017-05-16T17:09:29.417+01:002017-05-16T17:09:29.417+01:00Thank you. I apologize I promise this will be my l...Thank you. I apologize I promise this will be my last question!<br /><br />For the buffering you use 10%, how did you derive this? You said it is theoretically the most efficient way to do it so I assume there is a way to derive this number?Cal2223https://www.blogger.com/profile/09457839737383784545noreply@blogger.comtag:blogger.com,1999:blog-261139923818144971.post-9942183756339857452017-05-16T15:58:03.697+01:002017-05-16T15:58:03.697+01:00Great. So I suppose to the CS carry equity strateg...Great. So I suppose to the CS carry equity strategy above it would be best just to rebalance when "optimal" and using the buffering method?Cal2223https://www.blogger.com/profile/09457839737383784545noreply@blogger.comtag:blogger.com,1999:blog-261139923818144971.post-39589576466099239972017-05-16T15:49:21.468+01:002017-05-16T15:49:21.468+01:00Yes path dependence is a pain, but the fact remain...Yes path dependence is a pain, but the fact remains that buffering is theoretically the most efficient thing to to do.Rob Carverhttps://www.blogger.com/profile/10175885372013572770noreply@blogger.comtag:blogger.com,1999:blog-261139923818144971.post-73682185548656610762017-05-16T15:37:15.468+01:002017-05-16T15:37:15.468+01:00Thank you. My initial thought was that it would be...Thank you. My initial thought was that it would be better to do the 21 day moving average because a.) it would eliminate path dependency based on what day you decide to rebalance(for example, doing cross sectional carry for equities using raw carry(not averaged over a year) has a decent amount of path dependency because it appears to capture the dividend run up effect so rebalancing at month end is "optimal" and a decent amount of variation by changing what day of the month to rebalance) and b.) for less liquid markets it seems to be more effective moving 1/21 of the portfolio each day vs potentially trying to move the whole position in one day.<br /><br />Any thoughts on this??Cal2223https://www.blogger.com/profile/09457839737383784545noreply@blogger.comtag:blogger.com,1999:blog-261139923818144971.post-68294060788221762922017-05-16T09:21:32.825+01:002017-05-16T09:21:32.825+01:00No, it would be better to rebalance monthly using ...No, it would be better to rebalance monthly using a buffer.Rob Carverhttps://www.blogger.com/profile/10175885372013572770noreply@blogger.comtag:blogger.com,1999:blog-261139923818144971.post-1409856503915526302017-05-15T21:27:39.810+01:002017-05-15T21:27:39.810+01:00Thank you for the response!
One last question if ...Thank you for the response!<br /><br />One last question if you don't mind. In general, would you recommend the 21 day moving average('tranching') approach vs just rebalancing once a month and holding for a month?Cal2223https://www.blogger.com/profile/09457839737383784545noreply@blogger.comtag:blogger.com,1999:blog-261139923818144971.post-50048700357446514582017-05-04T09:31:10.818+01:002017-05-04T09:31:10.818+01:00Swell81:
1) You could use such a signal but it wo...Swell81: <br />1) You could use such a signal but it would very difficult to work out if the alpha was statistically significant, unless you're using a lot of data eg cross sectional equity data (think back to DeBondt and Thaler showing that 3-5 year returns predict mean reversion).<br /><br />2) This is a reasonable thing to do, a technique I've used myself, and will in fact reduce your costs. It's going to be virtually identical to applying a moving average to the underlying signal.<br /><br />However there is no free lunch. Like using a buffer, or a smooth, or any way of slowing down your system you will lose pre cost return from having the 'wrong' signal. In this case your signal will be delayed by an average of two weeks. <br /><br />So I'd avoid using this method AND a buffer; and depending on the character of your signal the best cost reduction strategies are:<br /><br />a) buffering<br />b) exponential smooth of signal<br />c) moving average or 'tranching'Rob Carverhttps://www.blogger.com/profile/10175885372013572770noreply@blogger.comtag:blogger.com,1999:blog-261139923818144971.post-90289213399127567692017-05-03T21:39:45.059+01:002017-05-03T21:39:45.059+01:00Also assuming there is an appropriate buffer in pl...Also assuming there is an appropriate buffer in place.Cal2223https://www.blogger.com/profile/09457839737383784545noreply@blogger.comtag:blogger.com,1999:blog-261139923818144971.post-22079570890362790952017-05-03T21:35:09.732+01:002017-05-03T21:35:09.732+01:00Hi Rob,
Wonderful talk at Quantcon and excellent ...Hi Rob,<br /><br />Wonderful talk at Quantcon and excellent book!<br /><br />I had two questions I hope you don't mind sharing your thoughts on.<br /><br />One is somewhat in regards to the above comment. Let's say you have a variable x to predict y. Would too long of a signal be the 5-year change in x to predict y? for example if x is an economic variable.<br /><br />Second:<br />For strategies with a one month holding period, what do you think of the idea of "tranching" for lack of a better term. What I mean is, to take an extreme case, if you rebalanced 1/21 of the portfolio each day based on the new signal. Or essentially, on day 1 you calculate the signal and rebalance 1/21 of the portfolio, then on day 2 you calculate the new signal and rebalance 1/21..etc. <br /><br />What are your thoughts on this? or the potential cost implications of doing so?<br /><br /><br />Thank you very much for your insights.<br />Cal2223https://www.blogger.com/profile/09457839737383784545noreply@blogger.comtag:blogger.com,1999:blog-261139923818144971.post-2931772777057841052017-04-13T09:53:39.668+01:002017-04-13T09:53:39.668+01:00Thank you. I agree that having only two speed vari...Thank you. I agree that having only two speed variations is also not a desirable situation.<br />I notice that it is difficult to suppress the urge to tweak a running system. Part of the learning curve, I guess.Jeroenhttps://www.blogger.com/profile/02819426964787793954noreply@blogger.comtag:blogger.com,1999:blog-261139923818144971.post-39127328909057303772017-04-13T09:36:11.705+01:002017-04-13T09:36:11.705+01:00Very slow rules tend to perform badly and are stat...Very slow rules tend to perform badly and are statistically indistinguishable from noise (and in the case of markets with secular up trends, from long only). How slow you should stop at is partly a matter of choice / judgement (not overfitting if you do it without looking at real data). Bear in mind that very expensive instruments can only trade 16,64; 32,128; 64, 256 - I would be wary of dropping the slowest of these and leaving you with only two speed variations.Rob Carverhttps://www.blogger.com/profile/10175885372013572770noreply@blogger.comtag:blogger.com,1999:blog-261139923818144971.post-28129523046316361992017-04-13T03:01:06.182+01:002017-04-13T03:01:06.182+01:00I agree about the risk of overfitting. But at the ...I agree about the risk of overfitting. But at the same time wonder what trading rules should be taken into account and what not. For a trend following rule (e.g. EWMAC) should a slow lookback period of 1 year be used? 5 years be used? Or only 6 months be used? To the purist even this could be considered overfitting?<br />I am including EWMAC(256,64) in my system as slowest but wonder whether it is dampening/suppressing the response to faster moves. My system is running less than one year, so it is indeed too short to be conclusive.Jeroenhttps://www.blogger.com/profile/02819426964787793954noreply@blogger.comtag:blogger.com,1999:blog-261139923818144971.post-6384281955131356162017-04-12T16:52:56.397+01:002017-04-12T16:52:56.397+01:00The correct weights depend on (a) costs (b) correl...The correct weights depend on (a) costs (b) correlations and (c) pre-cost returns. A proper optimisation would produce weights that factor these in, but also account for the uncertainty in estimating them: costs [almost none], correlations [some], pre-cost returns [loads and loads].<br /><br />The costs of JPY are fairly average; about 0.16% in SR units a year compared to about 0.04% for the cheapest markets. <br /><br />For correlations we'd want to avoid the middle variations since more 'spaced out' speeds would be better diversified. <br /><br />Although looking at a single example it might seem that the pre-cost returns would be higher for faster trading rules, this isn't generally true; it's almost impossible to distinguish between them statistically and inside a broad range of speeds the returns are very similar (although very fast, and very slow, don't do that well).<br /><br />Generally glancing at one year of performance and drawing these kinds of conclusions is a rocky road at the end of which is someone overfitting their system... Rob Carverhttps://www.blogger.com/profile/10175885372013572770noreply@blogger.comtag:blogger.com,1999:blog-261139923818144971.post-81164057069948862762017-04-12T16:12:21.179+01:002017-04-12T16:12:21.179+01:00Dear Rob,
thank you for providing so much insight...Dear Rob,<br /><br />thank you for providing so much insight in the structure and performance of your total portfolio.<br />I have a question about the JPY chart and your comments about it. You show the clear trend, starting in Nov. 2016 and ending in Jan. 2017, and the response of your trading system. You notice the lagging system response at the begin and end. Would the system perform better if less weight would be placed on trading rules using longer lookback periods and more weight on trading rules with shorter lookback periods? This would (I guess) lead to more frequent trading and trading costs, but would capture the trend better.<br />In your book you give (more or less) equal weights to faster/slower EWMAC trading rules, other than taking care of correlations, but I wonder whether that gives optimum results?Jeroenhttps://www.blogger.com/profile/02819426964787793954noreply@blogger.com