There is no direct link, but sure it's likely if we get a +ve forecast in NKY then we're more likely to go long VIX/V2X Rob
I think with handcrafting the top level weight allocation shouldn't really be 1\n, exactly because some groups have too few instruments in them (or we should try to come up with some even higher-level groups e.g. 'financials' for all stocks, bonds and volatility).. And on the other hand, the groups with many diversifying instruments should probably receive higher allocation.. E.g. My agriculture top-level group has 19 instruments (and it contains weird\diversifying things like Coffee, Milk and Orange Juice) and my InterestRate group only 6, so it sort of makes sense to give more weight to agriculture.. This logic makes things kind of arbitrary I think, because we could end up forcing the group weight that we (almost subjectively) like.. Here's my current top-level weights, and I can't really explain how I came up with them, they started as the result of pysystemtrade optimization, but I then adjusted them manually..
Totally agree. I want to do it manually because the implementation of all the estimators is a little bit too much work for me right now - adjusting my execution engine will have priority after I settled on a bunch of strategies. Nevertheless the volatility group (or RiskIndex, as you name it) seems like an odd one. There are just not that many instruments there which will result in overweighting of this group. I ended up moving the vol group to equities - what could possible go wrong? For the other groups I tried to design them in a way so that the number of instruments within them does not vary too wildly. On a related note: What kind of strategy groups do you guys use? The dichotomy of "convergent" and "divergent" looks a little bit too coarse-grained at first glance. Do you also adjust weights manually there (e.g. momentum 50%, carry 35%, skew 15%)?
I have another practical question regarding data providers, and thank you so much for all the insightful answers. Do providers like Barchart or Interactive Brokers face data quality issues? I currently use Norgate for futures data, which offers a selection of around 100 futures. However, it lacks coverage in areas such as emerging market forex and European government bonds. While their stock data quality is exceptional, I wonder if futures data is so straightforward that any provider could offer reliable information. Could you shed some light on this?
IB has data quality issues (-1, 0, incorrect prices, it's been mentioned here before), not too often, and I think mostly temporary, i.e. IB will eventually correct a wrong price in the EOD history.. Actually I often see lots of price warnings from my system on Monday after holidays\long weekends, but I think I usually redownload a couple of recent prices every day to allow for corrections.. These strategies don't need exceptional accuracy, so it's good enough..
I use forecast grouping from the Rob's last book, 60\40 divergent\convergent, then mostly equal weights within them, although I think I gave skew a little less weight than carry..
I use IB as data provider. It has already been noted that sometimes they have mistakes in EOD prices. However, I repeatedly overwrite the last 10 days (i.e. two trading weeks) in the historical price data that I use for the calculations. This ensures that eventually the incorrect EOD prices get overwritten/corrected by IB. In most cases are incorrect prices overwritten within the following 1~2 days.
After having implemented some rudimentary strategies and a corresponding order generator, I will now start to adapt my execution engine for futures. As you may have guessed, I never traded futures systematically yet (apart from a few currency hedges). During the last few days some random questions popped into my mind: 1. From the market wizards books I recall stories of problems with limit ups and limit downs. I guess these are called circuit breakers nowadays. Is this an issue I should be aware of both in live trading and in backtesting? If yes, what is best practice? 2. When building the backadjusted prices I shamelessly used the rollcalendars from pysystemtrade. I did the roll/carry within +/-5 days of expiry + rollOffset and searched for a date that all the contracts have prices. Should I refine the search to find overlaps where there are not only prices but the volume is "maximized" across the contracts? My intuition is, that this kind of optimization does not have a big impact on the system performance and therefore can wait. 3. For time reasons I tend to not implement Rob's execution algo but use IB's adaptive algo. Has anybody experience with this? Is it any good? 4. At the moment I am a systematic stock trader (exclusively mean reversion) and want to diversify into trend following (the meltup Nov to Feb did cost me dearly). From your perspective, is it ok to do everything in one IB account? Or is this something I will regret and should create a sub account or some other structure. 5. I read in the blog and this thread that rolling via a spread trade is preferable if the spread market for that instrument is liquid. How can I determine via IB API if a market is tradable via a spread trade? Does IB have special symbols for calendar spreads? I have never seen that before. Thank you so much for your help!
I use IB's Adaptive Algo with Normal speed (after testing the various speed options). I've paid 73.5% of the bid-ask spread over the last year, so it's definitely better than using market orders. When I last asked Rob, his and my spreads paid were about the same (our numbers have probably changed since I last asked, so they many no longer be comparable). So I'd recommend it. It also helps solve the problem of how to trade instruments without realtime quotes. I think others may use a SNAP algo for that, but I just use the Adaptive Algo for everything.