Interesting end of the week. A new wave of investments is causing a new "load phase" especially on volatility (plunge of mkt) an metals. Our generous investor, who is kindly supporting this test (and research), has thrown in another 170K, which will be helpful to trade more comfortably. Current margin usage is in the area of 150K. So far all is proceeding ok and according to plans, with no major problem or technical issues. We have been continuously growing our "G-L" to the level of about 60K, which provides a nice cushion. The drift ((G-L)/ L) is currently pretty high, due to the sudden mkt move which has been hedged with directional scalps. 4.6K have gone in commissions so far, for over 4,800 fills (we still are in the zone of most expensive comms). Max drawdown has been ok, in the area of 10%-15% the max margin used. Max PNL seen around 14.6K.
Here is the middle week update. Yesterday was a rollercoaster with mkt first continue dropping and then recovering. Along with the current "load" on metals, it caused a significant drawdown (-11K), then recovered at the end of the day (+6K). Most of the investment is still tied with metals and we need some move over there. Pretty disappointing JDST which has been not shortable (along with DGLD) for three days in a row and returning shortable only today about 12:30, which has made impossible both to properly hedge or scalp it. DUST has instead been working fine all the time. I am planning to remove ZSL and DGLD from the folio (when most players have been recovered). About JDST we will see, if it insists with behaviors like that (not shortable), I will also remove it from the folio.
The new week after the holidays restarted with relatively scarce activity. Mkt advanced both yesterday and today. Remarkable move of JDST just at the very last minutes of the session. Being not shortable I was not able to "overload" it as I wished with "manual injections" (lately I have automated this aspect: will show that later). We have still a good "investment" in metals and energy. Drawdown has been very contained and the drift advancing well. All is fine so far: no technical problems.
End of the week. Finally seeing some "green". Ending up again in the 14K zone with about 6K commissions. Drift ((G-L)/L) is ok so far, around 15%. The G-L figure about 80K, providing a nice cushion for further investments. Margins use has rised to around 150K lately, after our investor injected the additional 170K. Drawdown so far has been very comfortable, mostly in the 10-15% zone (with respect to margins used). Wednesday the "miners" (DUST, JDST) moved quite a bit and the following days (Thursday and today) also the metals moved a bit. Most of the $$$ is still tied with energy (ERX) and metals. Once they close most of the players I will remove ZSL and DGLD (ZLS had a large spread lately, and often not shortable).
The weekend pause is perhaps a good time to reflect and remind us our "trading principles", and the nature of the statistical edge we exploit. I would also like to slightly improve some piece of terminology, as I think I used the expression conservation of trading information (pag.29). I am now thinking that "preservation of past trading information" can be a more appropriate wording. The point is that we are preserving "past trading information", that we ourselves generate through the trading process. Clearly, it is by no means information "intrinsically" contained in the past tickprice/volume realization. It's merely the information embedded in the "order cloud" that we ourselves generate in response to the tickdata received, and according to the scalping hedging game rules. As to possible information in the tickdata past realization, we do not use that at algorithmic level. Clearly, this is a reflection of the conceptual stand that past tickdata does not carry information useful to be extracted to the purpose of a statistical edge. (That is why we need to avoid in our vocabulary, words such as data based "prediction" and so on.) [Clearly, such information (or "illusion of information") or other personal beliefs can still enter the system through the manager intervention, who has full control on every aspect of it.] We do, instead, fully consider "structural characteristics" of the instruments, which are public knowledge, and that might, for instance, impose "drifts" (of purely mathematical nature), or change the volatility or the leverage level, which is very useful to consider in the settings of the scalping/hedging games. So our weekend (updated from pag.29) "flash card" might be: Some notes - Mere "preservation" without "use" would be the same as the ordinary and common "Stop & forget" approach (which is always largely dominated by the above approach). - Scalping hedging games and architecture must ensure that past trading information is used. This is obtained by the player superposition (architectural feature) and the game settings so that order sizing and entry styles are affected by the whole (past) order cloud. This property (that is the capacity to exploit the past trading information for some form of asymptotic convergence) might be called "self-consistency" of an order cloud (let's skip any attempt of formalization at this point and keep all fluid, conceptual, and thus more powerful). - Architecturally, the most powerful math device for past information preservation and hedging is the "player superposition" concept, which allows to decompose the entire PNL into even the most minuscule contributions of each and every order, and also make visually apparent such contributions. - For any strategy S, a new strategy S* can be created which "dominates" (in the sense of >= performances) it by using the above "preservation principle". (We might call this, "Theorem of strategic dominance", which could take different incarnations, depending of the formalization one introduces, but let's keep it purely conceptual for the time being, so that we leave open a wide range of possible approaches such as, probabilistic, fuzzy, chaotic, etc., suitable to researchers coming from diverse perspectives and backgrounds). As to a "formal proof" of this statement, it is clearly dependent on the formalization one introduces; anyway, the conceptual idea is the one I have sketched in the previous posts. That is, we transform a reduced increase of entropy of the system, into an actual imbalance of the Gain/Loss components, which in the long term generated a "positive drift", which is hopefully able to stand against all the various growing trading expenses (commissions, spread, interests, etc.). This understandably may work in any practical meaningful settings, and especially well in purely theoretical settings (and unfortunately unrealistic) such as the "random walk hypothesis". In fact all we need for it to work is that we are in the zone of 50%/50% to close with a profit/loss any position we open, and clearly, no overwhelming trading expenses or other structural "drifts" strongly working against us. The loss recycling feature (player superposition) will add, in the relatively long term, the desired statistical edge to it, in the apparent form of a statistical drift.
A quick folio update. Scalping quietly. Most active instruments have been "small caps" and "treasuries" (TZA, TNA, TBT, TMV, ...). Most of the investment is still tied with metals and energy. (In the meantime I have been working on some program refinements, which I am planning to deliver soon.)
Hi QuantWizard, Bottom chart shows the current (realtime) individual PNL, instrument by instrument (46 layers). Top chart represents history of aggregate (whole folio) PNL (cyan solid line), Unrealized (orange dotted), and "G-L" (green dotted). (cf., initial posts: more details on result display were explained and discussed at the beginning of the thread.)
I feel like you should be trading less correlated assets. There's no reason to be trading both TNA, TZA or VXX, UVXY. Looking through most of the symbols on the bottom it appears as if most of what you are trading moves in lockstep with one another, most of which can be accomplished by just trading SPY. Why don't you start slowly adding equities with low market correlation?
Hi jb514, Well, yes. In "theory" it's like you say. However, in real world application there are some good reasons which become evident with daily practice. For instance TZA was not shortable most of the time these days, and TNA served well the purpose (also the dynamics are not exactly the same, as the inverse ones tend to have stronger decay and to "close" more easily the scalps). In general, it can be useful to have correlated items for the purpose of diversifying the hedging opportunities. In general, it is often handy to have both the bear and bull version of an instrument (this has to be expanded with some warnings in future posts, about leveraged bulls). After all you can always control the order size in such a way that you regard the aggregate of "correlated" (negatively or positively) instruments, just as "one" investment opportunity. About UVXY, yes, I have disabled it some time ago it (that was noted a few posts ago, because having VXX and many others, it was not adding anything useful.) Anyway when starting, it's always useful to test the instruments one by one, and gradually discard those which are no more needed or useful. Later, in case I will make a post about my current personal preferences on instruments (I have tested most of them), and of course if you have candidates to propose, suggestions are welcome. (Talking in terms of "mkt components", after all, the useful and tradable occasions of really diversified investments are actually relatively few, so in any case one has to reason in terms of aggregates.) [to SPY, the leveraged versions (fas, faz, upro, ...) are usually preferable in this framework, because they they often scalp "faster" and better, due to the "amplification", up and down, of the price wave]