Notice from your third chart that 10s vs 2s have an interesting support around 1.44 ish. Look at that spread today.
This position looks an awful lot to me like you basically short US SIFs, since both of your positions are betting against it. What would happen if you added a Long ES and short DAX, in some ratio favoring the ES? Imo, you are not assessing risk correctly by the question you are asking. Here is how I asses risk in complex positions: Lead-lag relationships between the legs Direction index By far the hardest to measure is 1. Direction index is actually pretty simple, as I was able to state it in one sentence above. 1 is where all the money is because it is where the risk is mitigated. Once risk is mitigated correctly, scaling is possible in highly liquid instruments. Correlation analysis is a very dangerous thing to do since it doesn't take changing dynamics into consideration very well, unless your code (I am assuming some sort of ATS. If it is not an ATS you will probably blow out your account one day because the dynamics of complex multi-legged positions need to be adjusted almost continuously, imo) has some intelligence to detect it. Much of institutional trading is trading correlation and implied correlation as an asset class.
I do think at this moment that short US SIF for the next 2 weeks or so is the way to lean due to fed speakers walking back some dovishness and some US macro data (CPI) looking a bit more hawkish. Long outright ZT is in line with this view and is behaving as such. The only way i know what to expect with a small portfolio of postions is to net them going back years to see periods of broken correlatioms, market stress, etc to worst case it. Being able to better measure , analyze risk may allow me to take on more reasonable risk. Currently I err on the conservative side and thats probably not bad. I have never, in my simple methods, looked for lead/lag relationships. I do not run ATS. So... Maybe i will step aside here and listen to professionals discuss risk. Thanks. Oh, and yes I do watch US indexes in relation to Europe and other countries so a trade such as US/germany would be within my scope. No opinion on that at this time. Lastly, the funding yen leg is more of a conceptual thing I do . I am strictly USD denominated but any position I take I ask "which currency do i want to sell here" And spread or net the position with this currency. Ie: owning US Indices in USD in 2014 was a poor trade. Long post as usual but electroms are free, so tHey say.
Well, while I think you are heading in the right direction of playing a multi-dimensional chess game, it has the feel of an outright position. For example, why not just go short ES? If it is because you have more room to wiggle with your position, then how do you plan to take off legs or add more instruments as the markets evolve? What will you do for example if the $/Yen does something totally "unexpected"? (Unexpected as it relates to your model, which here almost certainly means correlation) What do you do if you discover that in fact it is the SPX (ES) that leads your legs, and not the other way around? Or worse, what if the lead/lag relationship is totally dynamic? How are you able to tell one from the other? How does that affect your risk? How does the evolving probability distribution(s) affect your model(s)? Do you add in certain cases, or throw in the towel in others? Notice, if you are trading multiple instruments at once in some symphony, adding and removing units to the legs is probably a must. Another reason retail traders fail. They don't have the capital to fully realize a position because all they have is one or two bullets in the chamber. It is an all or nothing prediction with no room for error correction. If they can even put on all the legs on to begin with. You see, once you enter the multi-dimensional chess game Matrix, complexity explodes exponentially because search space and combinatoric dimensions expand. But so does agility and potential for profit, and perhaps more important, keeping risk quantitatively n-dimensional as well.
Well, looks like I can spend a lot of time pondering the many aspects of your post. Thanks for that. I do believe that you may think you are talking to a more sophisticated trader/investor than I am. I also realize this is a large community and its "not about me" . I have not explored lead/lag EOD relationships at all. I assume obvious/famous ones such VIX leadin ES or, when a short JPY move leads ES cannot persist or morph shortly into random. Did not Marty Schwartz find a Treasuries leading SP500 and milk it big? Im not sure of a simple algo to test for this within Excel VBA . Any thoughts on this? Also, I am not clear how a dymamic lead /lag directly effects risk other than the relationship didnt pan out. Knowing me, I would exit in that case. I toss in the towel rather than adjust. Most times I am looking for trend , a strong direction, And if i dont get it I usually see that fairly clearly. Repairing/adjusting positions (options) is not in my tool bag. I close out for a loss within my risk tolerance. Sizing, in a small account is admittedly difficult. Precision, granulatity errors mist be swamped by correct directional picks. I keep data in a volatility neutral format and initially size accordingly. That neutral relationship( the chart) is what i am trading. The account size vs. position count is more of an issue to me where i get scared of the notional size or leverage of open positions so I dont put on more size. Whats the proper way to quantify the additional size vs reduction in risk? Its important for me to keep it real. Most of the time , under time/pressure constraints i take the single leg ... . Often I have NO positions, never mind a portfolio. I view the ability to Profitably manage a portfolio of asset classes as the realm of the professinal . But, some sample portfolios i have simulated are doable for the amateir retail guy and quite pretty. Any comments welcome. GlobalArbtraders material on this site is helpful as well. Lots to think about in your last post Nitro. Thanks again.
Nitro, how would you approach analyzing lead lag relationships? Are there any statistical tests that you've found useful?
http://arxiv.org/ftp/arxiv/papers/1111/1111.7103.pdf I can get the gist of some of what they're saying.
eurusdzn the reason I am addressing you is that your position shows great promise in that it is aimed in the right conceptual direction of leaving retail trading behind. The beginning of that is to trade one thing against another to reduce risk. But it has to be done right or in fact you will increase risk. What I am trying to propose to you is that you are half way there. The other 50% of the loop I am suggesting you close is the following: IMO, try to be less directional. That is why I suggested some sort of hedge with ES/DAX spread favoring the ES. Once a trader learns to be comfortable reducing the direction of his position, he begins to see markets more like a casino, where this time around he is the house. If you trade multiple instruments, you have to have some way of updating your current view. If you put on a position, and there is no updating of that view, while that may work, I think it is contrary to the spirit of the way you are approaching trading. A huge part of that is being able to tell when the lead-lag relationship of your "Spread" is changing. How one does that is not trivial at all because market time-series are non-stationary. So the tools needed are considerably more complex. A part of the solution is for example Gambit suggestion above to use a cross correlation measure.
Gambit, I had not heard of the "The Hayashi-Yoshida cross-correlation function". That seems like a superior method to measure lead-lag since it is able to deal with asynchronous time series that aren't pretty lined up the way you see in the typical literature about this. My approach is to use time-frequency analysis, e.g., wavelet transforms, short-time Fourier transforms, Wigner-Ville, etc. I prefer the frequency domain to the time domain in my specific case, but this measure seems perfectly fine to me. Thanks for the suggestion.