Trading FUTURES with IB

Quote from mickson:

Ok so I am a bit of a dummy when it comes to posting pics in a nice format. :confused:

The truth be told I have purposefully tried to "load" the overall portfolio to try and stress test the bot. So yes it can be argued that these returns are random.

Having said that I got off to an immediate drawdown and was very interested to see how long it would take to comeback. What becomes apparent while giving high commands to the bot, is that it tends to behave in a relatively mechanical way, which is of course what you would expect as it has been hard coded. That is not to confuse it with saying that it is predicting the market.

I don't think that I could necessary do justice to a descriptive narration of the PNL behaviour as I clearly dont have those facts, this is a black-box. However, I will say that from an intuitive point of view one can anticipate certain aspects of the model and how it will react.

Tom is clearly holding some cards back which I compliment him for doing as the process of actually trading the bot is where the learning takes place, and I am not sure there is a simple explanation for its behaviour.

In conclusion, I have set myself a new goal, and that is to firstly maintain my current profits, and secondly to grow them but at a more steady and modest pace. So I guess one can expect a reduction in the leverage, through the next drawdown phase. We will need to wait and see.
More...

Compliments, my general ! ;-))

try also to use the discretionary bounds weapon, when appropriate. I think those can sometimes save from major drawdowns (when price approach low/bottom of range).

I have created another variant of ultraflat, more frequent, (say "flat") which i am now testing

I am now also adding some new rules at folio level (open positions, correlations, etc.), so that small capital investor can reduce the global exposition.


PS.
[ i use sometimes the width="1200" attribute in the img tag ]
 
Thanks for the formating tip.

Yes I intend to start using my combat skills as I grow in confidence.

I have tried to leave it to do things as programmed by you to first observe and learn. Also I have overwhelmed myself with instruments. All 5 folios have more than about 150 instruments combined, so there have been too many moving parts to focus on.

I am now in the consolidation phase, I have closed my most successful folio 1, where 90% of the profits were made and thus eliminated 66 instruments.

As mentioned earlier the next phase will be focused on protection and stable growth. Having said that I am a looking forward to the next DD period so I can learn more about the bot.
 
Just for further clarity, you have addressed this before but just to be sure.

Do the folio's behave as completely seperate machines. When looking at open positions as you describe is there ever leakage across the different folios so that the bot makes decisions with all the folios in mind??
 
Quote from mickson:

Thanks for the formating tip.

Yes I intend to start using my combat skills as I grow in confidence.

I have tried to leave it to do things as programmed by you to first observe and learn. Also I have overwhelmed myself with instruments. All 5 folios have more than about 150 instruments combined, so there have been too many moving parts to focus on.

I am now in the consolidation phase, I have closed my most successful folio 1, where 90% of the profits were made and thus eliminated 66 instruments.

As mentioned earlier the next phase will be focused on protection and stable growth. Having said that I am a looking forward to the next DD period so I can learn more about the bot.
More...

A crucial idea in this construction is also the (delayed) "instance overlay".

In this approach, each single "trading session" is programmed to strive toward profitability and the keep it.

Infact, you see that as time passes by, it engage less and less, and on more favorable positions.

This sort of "convergence" to profit (which clearly cant be simply guaranteed to happen any time) is to be used in a context of instance overlay, as to also benefit of a (self-)hedging effect of the various instances.

In other words, we want to spread and distribute the risk as much as possible, all over the range ;-)

Tom
 
Quote from mickson:

Just for further clarity, you have addressed this before but just to be sure.

Do the folio's behave as completely seperate machines. When looking at open positions as you describe is there ever leakage across the different folios so that the bot makes decisions with all the folios in mind??
More...

Yes each instance is self contained.

The PNL shown by each instance may have nothing to do with the real account.

I am now adding interaction within the folio (inside a given instance).

At the moment there is no (algorithmic) interaction among the instances, but it's clearly the "general" who can create it ... ;-))

Tom
 
Ok

That is the way I understood it from a previous explanation, I just wanted to make sure when you described your work on the open positions it was purely at the folio level.

If I could make one further observation which I hope I am correct in, is that the DD is in fact the lifeblood of the default strategies.

The very reason for the exceptional profits is the way the bot embraces DD. I will not use that word that can easily be mistaken for the approach.

So something the General needs to understand that you will need to lose a few soldiers to win the war. Alternatively you need to play with the flatter versions, whereby the focus is more on the scalping.
 
Quote from mickson:

Ok

That is the way I understood it from a previous explanation, I just wanted to make sure when you described your work on the open positions it was purely at the folio level.

If I could make one further observation which I hope I am correct in, is that the DD is in fact the lifeblood of the default strategies.

The very reason for the exceptional profits is the way the bot embraces DD. I will not use that word that can easily be mistaken for the approach.

So something the General needs to understand that you will need to lose a few soldiers to win the war. Alternatively you need to play with the flatter versions, whereby the focus is more on the scalping.
More...

Yes, right. DD is actually the "source" of profit. It's not by chance that we call it "investment". One could have occasional profit without DD, but that can be seen as exception within the whole picture.

DD has also a cathartic value which makes what we gain, "deserved" ;-)) Money, seen as reward for risk taking, joint with know how and hard work.

Systematic money making without DD would rather look like a form of stealing, and the society would not tolerated it.

So DD is actually necessary. All we want, algorithmically, is to make it as bearable as possible, while we continue profiting and spreading the risk ...

Tom
 
Quote from fullautotrading:

Yes, right. DD is actually the "source" of profit. It's not by chance that we call it "investment". One could have occasional profit without DD, but that can be seen as exception within the whole picture.

DD has also a cathartic value which makes what we gain, "deserved" ;-)) Money, seen as reward for risk taking, joint with know how and hard work.

Systematic money making without DD would rather look like a form of stealing, and the society would not tolerated it.

So DD is actually necessary. All we want, algorithmically, is to make it as bearable as possible, while we continue profiting and spreading the risk ...

Tom
More...

Whoever is observing these comments, TAKE NOTE, the post from Tom is part of the "secret sauce".

:D
 
Hi there,

I just redistributed the application to all people kindly testing and to traders.
If anyone has not received the update, please let me know.

I have been testing the "ultraflat" strategy with real money (and done pretty good).

What i think is better is to use (with this special strat) very liquid instruments, rich of small fluctuations (es, cl, cad, eur, aud, etc. ...). In fact, the basic idea is to quickly scalp as long as possible, to be able to slide over the price curve, to find better and more favorable spots from where to open the new trades.

I have been trading it live (real money) for 2 weeks, and the global folio position is still 0. Some trades just lasted few seconds.

It's a pretty good approach, i think, for small capital. Intruments need to selected carefully for this game. Later i will also publish the result of simulations with random data, but they are pretty good. [The ultraflat exceeds systematically a Sharpe of 4 even on large sets of randomly generated price curves (GBM with mean reversion). ]

In the new version i have also added <b>folio rules</b>. For instance now the entries will take into account the correlations among instruments.

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