Traders' Roundtable

I was wondering if anybody had strong views on a strong and robust process on how to select which instruments to include in a portfolio to trade.

Although the system development process of deciding the : WHAT, WHEN, WHETHER, HOW MUCH and OUT is iterative and somewhat chicken and egg, if the objective is the development of a robust system that works in many markets and many market environments then the WHAT to trade can be separated once the parameters and systems have been chosen and somewhat optimized ?

Thoughts

Tomaso

As a rule of thumb you want as many markets as you can handle
Also assuming the systems you vack test have some form of heat control then why not test on all available markets / 100+ markets - then you will left in no dounbt if the system is robust or not
In terms of chosing which markets to actually trade - again you should chose as many as possible and dont look to optimise base on individual profitability of markets
Try to have 7 or 8 "groups/types" of instruments perhaps represented, on average by 5-10 markets in each group
Problem I found is in stock indices where there are so many - I looked at correlatrions and piked a handful from around the World that had as low a correltaion as possible and represented different countries / time frames / continents
Hope that helps
C

Chris

Thanks for taking the time to answer. Your point is very clear. It makes sense as you say to expand the universe to test robustness of the system and allowing for more signals.

There are markets that don't trend well I presume you eliminate those ?

At the moment I am trading 18 markets in 6 groups. I can hence take all the signals because I can all fit them in my potential portfolio if necessary. I don't need to select which signals to take.

The subsequent step for me is to expand the universe but will have to start choosing which signals to take and which not. I trade a type of Donchian brakeout system with MT trend filter.

Would you have any wise suggestion or where to read up on sensible and robust signal filters or risk managers that suited for Donchian so as to decide which signals to take.

Many Thanks

while there is a lot of discussion on this...I will put this question to you Tomaso.
Which is a better choice to make....choose to eliminate certain markets, or choose to incorporate a lot of markets and then choose which signals which to not take.
I think the results will be vastly different, a lot depends on what you wish to achieve. For a really robust more automated system I would imagine you would want to maximise the amount of liquid instruments - I thing Chris makes the point of correlation which is a different kettle of fish again.
If every market was highly correlated over the long run, then supposedly you would only really need to trade one out of that group....
food for thought.
(Personally while I have what might be considered a robust system mentally I have trouble trading it without discretion.....its what you can live with)

"There are markets that don't trend well I presume you eliminate those ? "

Tomasso - No I dont eliminate those - because how do I know those wont be the markets that trend well in the future ?
I took Sugar 11 out of my porfolio in 2005 because it had nver shown my system a profit - it was silly error - but taught me lot - that year Sugar 11 was the biggest winning trade of teh year had I left it in the portfolio
That is why I suggest you add in a minimum of 50 markets and possibly more and take all the signals your system tells you to take
You can easily trade a lot of markets with a small amount of Capital and there are some excellent discussions on this forum about how to do that - use teh search function and you will find them

I feel sort of optimised only trading 100 instruments at present - There are probably 160 I could trade

Thank you.
I will search for the threads.

Read this thread from the early days of the board before Trading Blox existed and bump it with a reply/question perhaps?

viewtopic.php?t=24