Robotrading: CT + Trending Strategy on folios of futures

Ah, and this is the global result chart, covered by the folio performance windows (on the left there is the list of instruments).

<img src="http://www.elitetrader.com/vb/attachment.php?s=&postid=3187636" width="1200" />
 

Attachments

Unco,

I don't believe this will balance the folio as it related to instant price movements. You end up with a common denominator packet size per instrument based on historical ADR. ie. 1 si = 10 ES = 500 M6E etc.

The issues as I see it are correlation: ES, NQ, YM all generally follow the same path resulting in all of their positions opening long or short causing either near instant folio close taking profits or 3 x the draw down.

There is a bit of a leap frog game going on where one instrument may lead but until another follows the direction is uncertain. The volume, volatility and time of day come into play.

The folio opens new positions based on instruments price moving to set position on their individual grids. (calculation maintained by G-bot) Say ES opened long 1300 in the previous trade set but the folio TP forced a close with ES at 1299 or 1301. The 1300 ES entry is blocked out for the next folio set. ES will sit out until 1305 or 1295 is hit.

One idea is hold and monitor the 1st instrument scheduled to open (typically the fastest moving instrument), let the next two instruments open per G-BOT rules and calculated grid. If you get an exit great... If not play the faster moving instrument your held back to counter the draw down.









Quote from unco:

Hi Tom
Not sure this has been mentioned here before... What about calculating the average daily range (and/or average true range) in $ for each product and then use a "clip size" in terms of number of contracts that would allow each product to contribute the same way.

Ex:
Silver has a daily range of 2 points, that is 10,000 USD for each contract.
ES has a daily range of 20 points, which is 1,000 USD for each contract.
Etc... with other products
So the "clip size" for ES should be 10 contracts.
This way the folio PL should depend less on 1/2 products only.
More...
 
Thank you pocketchange and unco for the interesting discussion.
I think that this discussion is actually useful, and what emerges, from a broader perspective, i think, is a <b>discretionary</b> use of the robot.

I am infact convinced that while the robot certainly provides useful automatism for most cases, the user can manage to stay out main troubles with some discretionary intervention.

I have written a short <a href="http://www.datatime.eu/public/gbot/Strats%20G-BOT/default.htm" target="blank"> page</a> on this, and i am pasting some headings here:

<b>
1. Folio selection
2. Discretionary bounds
3. Double packet size
4. Manual STOP
5. Clone instruments
6. Forced first entry
</b>

These are just a few ideas, and i believe that many other will emerge from our discussion and use. Please, suggest more!

Let's see an example of <b>discretionary stop</b>.
Let's take SI for instance. Last few hours it has been relatively "sleeping". But we know this isn't the sleeping beauty. Isn't it ? ;-)) )

Besides, i note that the bot is <b>short</b> and, say (just an hypothesis for the sake of example), i am afraid a long run up is starting up. So what i can do - if i feel unconfortable with my current capital (assume i am short of cash) - is to use a "discretionary stop".

This means going manual and "neutralizing" the current position with a manual order. This is done in 1 second: just click on "manual" and buy 1, in this case. This allows me to <b>"take a break"</b> (temporarily "suspend" the strategy) and watch what happens (without actually quitting on the automatic game). In case a runaway of the price actually starts in the UP direction i can possibly <b>resume the autotrading</b> strategy later when i feel being near to the end of the move (and in that case i might even decide to double the packet size :-)) ).

<img src="http://www.elitetrader.com/vb/attachment.php?s=&postid=3187885" width="1200" />

Clearly another alternative, could be going manual and close the position. But this would sound more like an ordinary stop (and actually doing that systematically may result in systematic unprofitability).

I think that best results can obtained by putting together the discretionary and the full automatic world. There are infact some aspects that only the human mind can process effectively (impact of some news, perception of ranges, etc) and it would be probably stupid throw away all this information.

Clearly, a discretionary use is more meaningful when there is a sensible and intelligent trader or fund manager. One image i use sometimes when skyping with my friends (traders/fund managers) is that the robot is like your divisions but you are the "general" and you are in charge. In other words, your soldiers may know how to fight, but higher level decisions may often be more important than the battlefield ... ;-)

Tom
 

Attachments

By applying this concept to the idea of pocketchange to fight correlations, one might for instance "suspend" the most expensive instruments and let run the cheapest of a group of correlated instruments.

Take for instance (SI, PL, PA) or (CL, QM, HO, RB) or (ES, YM, NQ) etc. They are probably strongly correlated, so in some way in case of large move we can let one lead and the other ones be "suspended", waiting to approach a reversal point. Or some of them can use a doubled packet when near to a retracement...

Clearly, by practice, one would develop more ideas and best plans of execution.


---------

Another item which could be added to the discretionary list, is:

<b> 7. Use of options </b>

There are infact traders that anyway play option strategies, and for instance a <b>strangle</b> formed by 5-10 options could perhaps represent a (good ??) form of protection against any from of catastrophe for certain risky instruments.

Clearly, options are very expensive in terms of time decay and spread, and on this topic pocketchage can give the real best suggestions ;-)

Tom
 
Quote from fullautotrading:


You will certainly come up with new ideas about interesting metrics to compare instrs.
More...

Hi Tom

- ratio volatility in USD / PL (or something like that)

Can give an idea of how much a product has been able to profit from a range/trend

Another suggestion would be to have 3d optim charts for backtest, where we could see result in PL depending of the choice of instrument/packet/scalp size etc etc

Something like that
http://www.amibroker.com/guide/gifs/h_optimize.gif
Not sure how difficult it is to code.

Hi PocketChange

Thanks for your advice, we haven't had time to backtest my idea to see if it's worth it, but will let you know if anything relevant. ;)
 
Quote from unco:

Hi Tom

- ratio volatility in USD / PL (or something like that)

Can give an idea of how much a product has been able to profit from a range/trend

Another suggestion would be to have 3d optim charts for backtest, where we could see result in PL depending of the choice of instrument/packet/scalp size etc etc

Something like that
http://www.amibroker.com/guide/gifs/h_optimize.gif
Not sure how difficult it is to code.
More...
hi Unco,

About the first one, it's an interesting idea.
"profit from a range/trend" is actually dependent on the "pattern" the instrument price follows while covering the range.
For instance, for a "straight line" pattern there would be DD only ("investment"). For a pattern with significant retracements, there will be large profits, clearly proportional to the amount of retracements.
I am going to add it (along with the range too, at this point!) :-)

About the 3D chart is no problem to make it. The point is to make it in a way which makes sense. For instance, i think that it's useless to correlate instruments, scalps and packet size for the reason that they are conceptually independent. For instance, by doubling the packet, every relative performance remain the same including the reward /risk ratio: we only have double profit and double drawdown. Similarly changing the scalp, unless we choose unreasonably small or large values, does not impact on the strategy performances, due to a principle of "scale invariance" (random fractal): so any discrepancy which one would see could be imputed to chance. A smaller scalp will just mean in general larger profit and larger dd.

The concept of studying visually possible <b>parameter interactions</b> is valuable, and has to be applied to form of correlation (or more broadly of dependence) which do not fall in the categories so-called of "spurious" or "nonsense" correlation (dependence) and, of course, in case of independence.

http://en.wikipedia.org/wiki/Spurious_relationship
http://en.wikipedia.org/wiki/Udny_Yule
http://ideas.repec.org/p/ags/uguewp/34151.html
http://www.triple.net/fractalmarkets.php

Tom
 
Ok here is how i have "reorganized" the folio performance panel.

I have added the metric smartly suggested by unco.
[and also some variant of it like Range in ticks, and using the Realized].

Finally opted for the treeview (on the left) which is handy for the trader and easy to maintain codewise.

Probably other interesting metrics will come to mind soon :-)

To measure the "speed" (wondering if there is a better word) of the instrument i have introduced a specific indicator based on the number of scalp lines which are crossed per hour (or, more precisely, the inverse of the average time between 2 hits).

<img src="http://www.elitetrader.com/vb/attachment.php?s=&postid=3188342" width="1200" />

Tom
 

Attachments

Here is another trading idea for people who like a "touch of discretion" or more than that ;-)) An approach for traders which dont have good capitals could be as follow.

It's pretty clear that we make rapidly good money when capturing "noise", which means taking advantage of waves that are "narrow" and not very "deep". Rapid fluctuations.
"Wider" (time) or "deeper" (price) waves clearly makes us more like "investors" than "speculators".

So we might consider anyway a relatively large folio [clearly compatibe with the capital], to maximize the chance of finding intruments which are currently in a more sideways mode. We would minitor the DD and when some instrument begins to "absorb" comparatively way too much in the folio investment, we simply "suspend" it. This would allow in the meantime the realized component to grow, due to to other instruments which are scalping, and "rebalance" the situation.

When the folio in again in "balance" we might resume again the running instruments an let them play having a "realized" cushion.
(If the suspended instruments in the meantime have continued their race, we will find ourseld in a much better position to continue the investment on them.)

Just an idea. It should be seen how it actually also "feels" in practice.

Tom
 
Here is an example of that.

Watching the new folio performance monitor ;-) , I immediately noticed that SI and CT were "absorbing" too much money, unbalancing my folio.

So i have "suspended" them (you can see the realized drop were i have been experimenting and playing with "discretionary stops").

This allows me to let the sideways component catch up. Also CL is absorbing money. But here i decided to let it go.

Why ? Well here is the discretionary component ;-)) (**)


Tom

<img src="http://www.elitetrader.com/vb/attachment.php?s=&postid=3189160" width="1200" />


PS
(**) i guess that i feared a long and slow upward "wave" for SI. And about Cotton, being this the first time i follow it and having no absolutely clue about its "personality", i just decided that, as first encounter, this guy was being too intrusive for my taste ;-))
CL is a more familiar guy to me, and i know he needs to breath with deep waves which almost regularly retrace. ;-)
 

Attachments

Once one feels the folio is rebalanced, in case he might decide to "resume" the suspended instruments to autotrading.

Or he might wait to accumulate more realized....

Unco pointed out in a communication also the importance of correlations.

[For instance, we might suspend some instruments in a group of correlated ones. Clearly we are talking about "realtime correlation", not historical. ]

Therefore i had the idea to add a <b>chart of folio correlations</b> too. I think it can be very useful to the wise fund manager.

Let's see if i can get it done before this evening ! ;-)

Tom

<img src="http://www.elitetrader.com/vb/attachment.php?s=&postid=3189376" width="1200" />
 

Attachments

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