Hi Rob, I have a couple of questions: - if I undertand correctly, your EWMAC trading rule (and I guess also the other ones you use) scale "raw signal" by short term volatility. Doesn't this create an implicit long bias? A +5% gap between fast and slow EWMA will probably result in a larger absolute signal than a -5% one, because uptrends are usually associated with lower volatility than downtrends. I guess you have many more +20 signals than -20s (obviously assets like vol futures and maybe bonds would work the other way around). My personal choice is to use long term vol (10y) to scale raw signals in order to have more "neutral" final signals. Is your choice intentional? I guess one reason could be that betting too heavily against risk premia is probably not a good idea, unless the strategy acts as an overlay on a traditional long-only portfolio. - I read somewhere that you get your data from IBKR. Can I ask you how much you pay per year? Have you considered web scraping as an alternative? I wrote some code to scrape data from ba3chart website, but I recently found an alternative way which requires very few lines of code (happy to share with private messages if someone is interested). Also, this way I have access to market data which is really expensive with IBKR (ICE futures). Obviously I'm talking about EOD data (ba3chart doesn't have bid-ask spread info for quite a few contracts) but for intraday data (which I only use to have an idea of how liquid the contract is) I found that IBKR lets you poll delayed data for free for all the futures I tries. Thanks
Hi Rob, Would it be possible for you to share the dataset which was used to generate the daily return chart?
I wish... but I'm skating a bit on thin ice with the data I supply in pysystemtrade as it is. Although it's 'user generated' (not raw contracts prices, only backadjusted) and I gain no commercial benefit, ultimately it's probably a data licensing grey area. I think if I put up the adjusted prices of 140++ futures markets, I'd be taking the piss. GAT
A very subtle question, I like it! Well, there's an implicit long bias anyway, because most markets have gone up over time I will always have more +20 and -20. And the effect you describe is probably strongest in clear risky assets like equities (or risk off assets like vol). There is some analysis that goes part of the way to answering your question in these two posts: https://qoppac.blogspot.com/2020/07/do-non-binary-forecasts-work.html https://qoppac.blogspot.com/2020/09/forecast-linearity-and-forecasting-mean.html So for example, the plots of forecast vs realised return show that if anything the forecasts aren't long biased enough (a zero forecast associated with a positive ex post return). And the second post talks about making the vol measure more long term in nature - something I've implemented - but not for the reasons you state. ... but I haven't actually explored this specific topic. My gut feeling is it will make a very small difference, and since it only works on a subset of assets I'd be reluctant to implement. Still, as they all say, "Further research is needed". This question might even make it into my next book (I will credit you) In fact I do web scrape from the website you name, but only for historic data. (I paid them for a premium subscription as well - did some multi year deal which finishes in 2023. I won't be renewing, as I will probably have added all the markets I need by then) I've considered web scraping, but it's extremely hacky to get right, and you're at the mercy of the provider if they decide to change their formatting or website in one tiny way, or worst still drop the data entirely (like yahoo did). That's fine for historic data where the process of backfilling is quite manual anyway, but not for production data. We've had a debate on this thread before about whether it would be worth getting daily data for an ICE market, and then trading 'blind' without live market data. And with my new dynamic optimisation I would also have the option of bringing in data, calculating an optimal position, and then not trading the market (but that market would give me useful information for a market I could trade). But incorporating another data provider into my production pipeline, especially with the potential it could go wrong, is a big faff and probably not worth the very small marginal benefit (although it would be nice in particular to have short sterling and euribor in my portfolio, for sentimental reasons). GAT
Oh no, I didn't mean that! What I meant is just the daily returns that were used to generate the chart.
I can give you an impression about this this. At IBKR you subscribe to those markets that you are interested in. So you won't pay for markets you don't need. In case of futures I am subscribed to: EUREX: about 12 USD/month (it could be that I'm having too many subscriptions here) Korea: 2 USD/month Singapore: 1.5 USD/month US "futures bundle": 10 USD/month (however, this is waived if your account size is sufficiently large) I am not subscribed to Japanese futures (I use the Nikkei future on SGX instead) and London futures as that is extremely expensive. I haven't found other markets that I'm interested in.
Seems the google drive file was not shared (it says "You need access Ask for access, or switch to an account with access.")