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

Here is a quick folio update. Nothing remarkable happening today, market slipped a bit, we are quietly scalping away some profits with small packets.
The <b>statistical drift</b> we have created has pulled up the PNL around 20K (out of 34K of "net gain" we have accumulated) and the margin usage at the moment is just around 1/3 of the available capital.

The return on avg margins used appear pretty high at the moment, but soon I will start a new phase of "investment" which should cause a new wave of DD. In the meantime, I am letting some remaining "red" layers scalp away their way "above the water". FAS is still "locked" with options and we just watch it for the moment. GASX doesn't seem to have any clue about where it wants to go, I will continue hammering it up and down until even the last layer, which I used for some "forced hedging", becomes green :-)

It's apparent that some instrument do scalp better. And as soon as their players get all closed, I will remove the least performing and slow instruments.

<img src="http://www.elitetrader.com/vb/attachment.php?s=&postid=3923493" />
 

Attachments

This is an example of scalping/hedging action (<b>order cloud</b>), as a result of the example "game" called "CT_T". The first picture shows <b>all the orders issued</b> on this layer, while the second picture only the currently open players (all the closed players were closed in profit, of course.)
[I almost always stay in "player-view" mode when managing, because I don't need to care about the closed players.]

In particular, a layer of <b>DGAZ</b> (VELOCITYSHARES 3X INVERSE NA). In this particular case, the game has a short-position constraint.

Notice that "position constraint" and "forced entry direction" are 2 different modes (they can be set in the instrument settings dialog).

While the "Forced entry direction" is very rarely used (by me), the "Position constraint" allows for <b> both short and long players</b>: what is being constrained is simply the <b>overall position</b>. So this game, even if constrained, retains excellent "<b>self-hedging</b>" capabilities.

<img src="http://www.elitetrader.com/vb/attachment.php?s=&postid=3923590" />

(Note how UGAZ is signaled as inversely correlated instrument.)
 

Attachments

<b>About frequency</b>

Addressing the issue of "frequency" and commissions, let's see briefly some points. The algorithmic engine I have developed overlays BUY and SELL players following "rules" that the fund manager can modify at will. Therefore, all aspects relative to the scalping/hedging game are essentially a choice of the fund manager. The frequency and "style" of the entries can be set at will, and one could make thousands orders per day, or even one order per day, just as he wishes. Experience, however indicates that to achieve a <b>"smoother" equity curve</b> one needs to play rather "tight" games, which are able to promptly correct possible drawdowns which may start developing. Usually, the more spaced are your "decisions", the larger is the maximum drawdown you suffer. That is intuitive and, of course, it depends on the specific rules of the game, but, very generally speaking, a more tight game allows (in the sense of being a "necessary", but not sufficient condition) to build a "smoother" equity curve and more control on the DD. Well, of course, this has a cost in terms of commissions. But experience suggests that usually it "feels" better to trade more frequently, even spending more in commissions, than trading more rarely, and staring at larger DDs. Well, of course, it is also a matter of personal preference, and anyone can just calibrate the scalping/hedging game the way he likes.

See, for instance, the order cloud we have just shown above (DGAZ). It has made several entries both short (directional) and long (hedging). Now, while, in this specific case, the price has been sliding down, how do you know when it is going to make a big surge up, and when the surge begins? You don't. There is no such a thing as a "strategic entry" (= prediction) because it's arguable that <b>nobody can "predict" direction up to a given take profit</b> in a statistically significant way, and also because, in any case, <b>we do not use prediction</b> (this is a <b>non predictive</b> approach). So, very simply, if the scalping/hedging game sees the price go up, it does not bother to try to "predict" if this is some "trend" (provided the word means anything) upward that is beginning or not. It just hedges it, and that is it. Then, if the price goes down again, it simply re-establish the short position. Anyway all those <b>hedging players</b> are also closed in profit whenever possible, therefore they will be transformed into gains too, thus feeding the "drift". And that is what essentially causes the <b>"statistical edge"</b> we exploit here.

<b>About "timeframe"</b>

In forum discussions you often hear people asking about the "timeframe" of a certain strategy. Note that in my approach, just as the concept of "trade" is meaningless, also the concept of "timeframe" does <b>not</b> actually apply, as the "clock" that dictates the trading decisions is actually the "price" itself, and "time" does not play an important role in the decision process (except, of course, when it influences the hedging mechanisms through options and "time decay"). Another way to look at it, in practice, is that the scalping engine will work on all possible "timeframes". In the sense that, if, to close profitably a player, it takes half second, then let it be. And if it takes 1 year, also that is the way it is. It simply does not care. It will only care to close it in profit. [Well unless it's closed ("forced close") by the fund manager, but that is another story].
 
A quick folio update. Today started pretty quiet. Then we had some good moves on energy related instruments. The application has been loading up in gas related instruments. We are still waiting for the metals and indexes.

I have been adding a few instruments (layers) to watch and the margins have been growing along with some drawdown, due to the new investments.

Net gain has raised to about 49K.

<img src="http://www.elitetrader.com/vb/attachment.php?s=&postid=3925265" />
 

Attachments

End of the week.
Today has been an active day with a good number of orders. The rise of metals has helped hedging the gas related instruments, which have been "loading up".

The current situation is summarized in the pictures below.
We are currently 19.4K "above the water", 86 days, 5K comms, 43.5K "net gain" (difference between realized gains and temp losses), RD=18%, HI=109%, current margins less than 1/2 of the initial capital.

FAS is still "locked" with the options and the "red" layer of GASX has improved a bit through some scalping.

Some instruments (DRN, IYR, DZZ, ...) have not been scalping as desired, as they appear to be too "slow" and tend to stagnate, so I will remove them as soon as their players find a way to close. I have added some new instruments that for now are just being watched (120 layers), before possible activation or removal.

<img src="http://www.elitetrader.com/vb/attachment.php?s=&postid=3925888" />

PS. If anyone is interested in the details of the executed orders, drop me a msg with your email.
 

Attachments

This weekend I have been adding a new feature to help dealing with the "disruptive"<b>split</b> and <b>reverse split</b> (merge). Later I will explain it in more detail, with an example of usage.
I hope you have also been having good time with your favorite activities and research.

Here is a picture representing the statistical edge we aim to exploit with our methodology.
The upper drawing represents the most common case, where traders build methodologies using the "ordinary" stops.
The drawing below represent our procedure where the stop concept is replaced by the <b>player superposition</b>. This could also be called a form of <b>"memory-full" hedging</b>, in a certain sense, if we agree to call the common approach, <b>"memory-less" stops</b>.

<img src="http://www.elitetrader.com/vb/attachment.php?s=&postid=3927045" />

Now I think it is possible to prove (theoretically) a form of strategic "dominance" of the second approach, in the sense that, given any strategy belonging to the first class, it is always possible to find (and even actually "construct") an element in the second class which (vastly) improves it. In essence, an edge provided by the reduced entropy. This would probably regard at least two separate aspects: one related to the conservation of information ("player"), and one due to the player "superposition" activity, which, clearly, creates strong interdependence among the entries.

This does not mean, of course, that any element taken from the second class is systematically profitable (I am merely making a statement of strategic "dominance"). You can, of course, lose consistently with any framework. In the first case, I think you are almost surely guaranteed to consistently lose, no matter what you use for your entry/exit, as the loss of information caused by the stops, combined with the other expenses, piles up too strongly against you, and it is generally impossible to rebalance it (except, of course, accidentally for short periods). The high variance of the phenomenon may make it less evident, especially in the medium low frequency and with small folios, so that one can be entertained for a whole life happily trading and "fluctuating around zero" or occasional lucky strikes (even sitting on a negative drift).
 

Attachments

Monday was hectic. Just the way I like it :-) A lot of activity, with market strongly pulling back.

The activity in bonds and metals has been suggesting to take seriously the mkt move.

At a certain point, I took the decision to remove all the FAS options. The decision had the following rationale:

1 The folio situation is significantly better than when we placed them in place, so we can afford some extra risk
2 I felt confident I could continue hedging through layers (in case helping it a bit with the "strategic injector" if necessary) now instead of using the options, as we have been doing with GASX
3 The option structures were all in profit and now that the mkt seems ready for a pull back, there seemed to be not much point in leaving there the <b>time decay</b> we have collected by "locking" FAS in these structures

So I closed them all, and their price dynamics are summarized in the pictures below.

<img src="http://www.elitetrader.com/vb/attachment.php?s=&postid=3927349" />

Let me know if you agree with this decision, or whether you would have done it differently. We will have plenty of new hedging situations, where we can apply different hedging actions :-)
 

Attachments

Tuesday (Jan 14). Pretty active day.

The application has placed many orders, probably also because I slightly changed the game rules, making the entries a bit more "sensitive" to direction changes.

The options on FAS are gone and the app is now free to edge it by superposition alone. It will be fine.

We are still "above the water" of about 23K. "Net gain" = 56K (difference between gain and temp "losses"),
RD = 18%, HI = 107%. 90 days, volume traded over 500,000 shares, almost 6K comms.

<img src="http://www.elitetrader.com/vb/attachment.php?s=&postid=3927667" />
 

Attachments

Here is a picture which aims to simply illustrate the nature of the statistical edge.
This is the ordinary picture of the PNL and its 3 components, which by now you should be your friends:

PNL (cyan curve) = Gain (green dashed curve) + Loss (red dotted curve) + Unrealized (cyan dotted curve)

but with and extra Loss curve (red-dotted curve) "mirrored" above the time axis.

In the yellow bracket it is indicated the <b>Gain/Loss difference</b> which is building up due to the player superposition (or "stop recycling" if you prefer).

<img src="http://www.elitetrader.com/vb/attachment.php?s=&postid=3927874" />

That difference is what we have been calling "<b>net gain</b>". That is, of course, actual $$$ that was scalped, which in part shows as PNL, and in part goes into the unrealized component ("investment"). It also contributes to "sustain" our investment with an increasing "cushion", along with the preexisting capital (in fact if we did not have it, the PNL would be lower).

In the meantime, you see the continuous hedging action due to player superposition keeps the unrealized bounded (almost flat), so that the building up of the net gain, impresses a <b>drift</b> to the PNL (which we measure with the <b>RD</b> metric).
 

Attachments

You must log in or register to reply here.
Back
Top