Trading algorithmically a folio without stops (with IB), real $$$

[fullautotrading]

... I am not disappointed. I accept that science or logic has often a difficult stand especially against bizarre theories and misconceptions circulating in the trading scene. So, I wish you success with your miracle system - after all, your chance to end in profit is almost 50% .

No, without, a significant positive drift (stronger than the massive trading expenses), it would not. The trading expenses, spread and interests, with this kind of frequency would drag you down pretty quickly, with no coming back, ever.

Btw, I never presented anything as "miracle", I am just a scientist doing a journey. A fully transparent one, where every trade is under the light of sun and any concept can be openly challenged. But I guess that misrepresentation is sometimes needed to diminish the work of others when one feels uncomfortable. So no worries

I wait for your "journal" on ET to show your concepts and facts at work and propose some constructive criticism

 

[jcl366]

I wait for your "journal" on ET to show your concepts and facts at work and propose some constructive criticism

Fair enough. In fact I'd invite your criticism, but you had to come to my website for this. If I posted here my journal and trade tool the way you're doing it, I would be banned by ET due to unfair competition to their sponsors. This is no joke - it happened before.

It was not my intention to diminish your work, I was only curious because it's an unusual approach and I didn't understand the method at first.

 

[jb514]

Not sure it's really possible to profit off of a random walk as it is technically both random and mean reverting. Free of friction I believe the pnl would be exactly zero

 

[fullautotrading]

Not sure it's really possible to profit off of a random walk as it is technically both random and mean reverting. Free of friction I believe the pnl would be exactly zero

That's right. In the case of "Open/TakeProfit/MemorylessStop" strategy types ("single-threaded"), the asymptotic avg of the PNL is obviously zero, if assuming no spread and no trading costs, and enough capital to sustain the trading indefinitely (unlimited). No matter what is used to "open", and regardless of take profit and stop sizes.
Trading costs, spread, and other fees add a negative drift (increasing with traded volume).
(With limited capital, it can and will obviously blow up the account before showing the asymptotic behavior.)

 

[fullautotrading]

In other words, with commissions, spread, etc., the PNL is going to -Infinity, or -AvailableCapital, if the capital is any positive amount (enriching the broker ) ).

In my personal perspective, this is to be interpreted as the inability of a strategy (trading methodology) to use the trading information (in fact, the "single-threaded" class of strategies described above does not use any "trading information".)

This is also a clear indication that the tickdata "alone", contain no useful information, if not used jointly with trading information (a trivial observation carrying however deep consequences for any approach based directly or indirectly on tickdata-based "prediction").

This has also parallels with the concept of consistent estimator in statistics. Similarly, an inconsistent estimator is, intuitively speaking, one which is not able to use additional sampling data to improve its performances. And consistency is obviously the very basic prerequisite for an estimator to be acceptable at all. (Similarly the concept could be formalized in terms of conditional probability, if one likes.)

Conceptually, as I see it, the crucial point is the preservation and use of the trading information. It's essentially the continuous "waste" (or no efficient use) of information (or the missed containment of entropy rise, if you will) in the system which provides the conditions for the trading expenses to surely prevail (even with unlimited capital).

 

[fullautotrading]

End of the week. We had some large moves on the ETFs related with miners and metals which have absorbed a relatively large investment, with consequent nasty drawdown. Still stuck the situation with energy stuff, which requires continuous hedging.

In the meantime I had the chance to shut down ZLS in slight profit, so I did. Same for one of the SLV layers. ZSL had been sporting lately a large spread and often non shortability. So it's gone. SLV is fine as ETF, but has too little action compared to the other ETFs and tends to keep tied capital with relatively smaller scalping results (pretty much like stocks). So one layer is enough for the moment. I am waiting for the first chance to happily throw away DGLD which has been non shortable for several weeks now. Same thing will probably happen to DSLV and JDST.
Candidate "substitutes" for these ETFs had already been loaded and they are actually those absorbing most of the current "investment" (GDXJ, NUGT, USLV, ...).

No technical problems. (Next week will be trading with the new game versions.)

 

[fullautotrading]

One question I get practically everyday by new enthusiasts is how much $$$ is necessary to trade within this framework.

Clearly, the answer depends crucially on how much money it is allocated to a single initial entry (this is called a "packet").
Currently, I am defaulting, for ETFs, to a "packet" of about 1K, adjusted according to instrument volatility.

With such initial allocation and as very rough estimates (to be possibly revised in the future), I'd say:

- For RegT (not advised) and capital > 80K-100K , say 20K/25K per instrument
- For Folio margin, say 10K/15K, per instrument

Clearly, in the first case, the folio possible is pretty small (max 4/5 instruments), and therefore it's wise to choose instruments in such a way to maximize the folio hedging effects.

For futures, it must be kept into account that the base packet is significantly bigger (due to the fact that you can't trade less than 1 contract and the multiplier can be pretty large). So the amount necessary per instrument should be scaled up accordingly (proportionally to the packet value).

[Clearly, in the case of futures, the ratio profit/margin, will be larger, however, in my opinion, capital must still be sized to "packet value", for the reason that, imho, people really perceive the "danger" of drawdown relative to their assets, rather than margin usage. And it is of course crucial that the drawdown remains relatively "bearable", in order to avoid violating the basic principles of the approach]

Unfortunately, most of the people making such inquiries do not have "sufficient" capital to trade. In such a case, I usually suggest a couple of options:

- Paper trade, create enough experience and build a track record and enough self confidence to be able to talk an intelligent larger investor into the trading venture

- Join forces with other investors interested in trading activities to put together a decent capital and share possible profits/losses (well, paper trade first, in any case)

What I systematically strongly discourage is taking a "gambling" approach, where you hope to rise "relatively" large money from small money. That would lead to almost sure loss of the available sum. Needless to say that most of people hate to hear that, and in most cases they just get very upset.

 

[fullautotrading]

Mid week update. More drawdown on metals and energy. Pretty large load and we are back "underwater" (-10.5K).

Today we had both the market and miners/metals going up, with S&P 500 ending at record.
The move on the miners ETFs has been quite massive and absorbed significant resources. Also we are still stuck with energy.
I also intervened manually to control and empathize the hedging action, and at one point I even "inverted" the position on GDXJ. However, within a couple minutes after the end of session (when orders could still be executed), I removed almost all the "hedging" buys, because of the fear that tomorrow some of these instrument might show up as "not shortable", which then would be a situation pretty difficult to deal with.

This already happened with JDST (see picture below), and has caused practically losing all the move down (several Ks).

Same thing, and even worse with DGLD. Clearly, if an instrument goes up significantly while you are building a short position and you need to hedge with buys and then, when the instrument reverses, you are not able to close the buys because the instrument is flagged as not shortable, it's quite a problem, as it breaks any algorithm. So these instruments have fully "deserved" to be placed in the list of those to be "thrown away" at the first good occasion.

In the meantime I have introduced some computations to express the exposure of an instrument relative to the correlated instruments (on other layers), and today this has proved to be invaluable to help hedging across layers (different instruments), as you can probably tell from the "steepness" of the last part of the "G-L" curve.
Margin usage had a peak at about 280K, while currently about 220K.

 

[jcl366]

Just out of curiosity: Your system seems to often require manual intervention for limiting losses. Wouldn't it make sense to automatize this, so that it won't need permanent supervision?

Also, do you compare live performance with simulated performance, so that you can pull the plug when the total loss exceeds a previously calculated limit?

 

[fullautotrading]

Just out of curiosity: Your system seems to often require manual intervention for limiting losses. Wouldn't it make sense to automatize this, so that it won't need permanent supervision?

Also, do you compare live performance with simulated performance, so that you can pull the plug when the total loss exceeds a previously calculated limit?

Right jcl366. Full automation is the limit where we continuously tend.

To arrive at that ideal point it's a slow and painful process where you gradually and carefully incorporate in the trading engine all the mechanisms and actions you would like to carry. The feeling that you must have is that the application at any time is anticipating your "intention" (where clearly the "intention" must be "educated" through experience and a lot of simulation).

So as we proceed the engine becomes more and more capable to fully substitute and anticipate the intention of the fund manager. It's however a process, a journey, and it's very useful to interact with it, because from interaction stem the ideas for new automated actions and mechanisms. For instance, the lately introduced mechanism to grab quick moves, derived from such interaction (and the desire to automate some actions which where previously carried out manually).

Remember that "automated" does not mean that you leave it running and go to the beach It simply means it allows you to deal with a mass of work with precision and reaction speed which would be impossible without it.

When you have money on the line it's always wise to supervise operations. Actually the larger is the capital the less you can care of small optimizations and strict supervision. But especially when capital is relatively scarce, you better watch it as a hawk

I don't usually compare with simulations. The simulator to me is simply an environment which helps the fund manager educate himself and to make up his mind about the hedging/scalping rules he wants to enforce, as he can see them in action in the wildest possible ("realistic") scenarios that can be conceived.