Dynamic Portfolio Selection

Discussions about Money Management and Risk Control.
Dean Hoffman
Roundtable Fellow
Roundtable Fellow
Posts: 87
Joined: Wed Jul 02, 2003 3:04 pm

Dynamic Portfolio Selection

Post by Dean Hoffman » Wed Oct 05, 2005 7:36 pm

I’ve done quite a bit of research recently on the concept of dynamic portfolio selection. Typically, a portfolio is decided ahead of time and is therefore static. Signals are generated for those various markets and there is either a trade or there isn’t. There are lots of potential pitfalls with static portfolio selection. The most obvious problem is hindsight biased curve fitting (picking only those markets that did the best in the past). We know this will likely lead to failure because almost certainly, yesterdays best markets will not be tomorrows.

The research I have done is best described using a parking lot analogy. Assume that your portfolio is a series of cars and parking spots. If you have 20 markets in your portfolio then you have 20 cars and 20 parking spots. Sometimes you have on a trade and this represents a car parked in one of the parking spots etc. However, what if you went to a portfolio model of 100 cars but still only maintained the 20 parking spaces? This would allow you the minimum account size control afforded by the predetermined 20 parking spots; however, it gives you the flexibility of 100 different cars to potentially park.

Some people have done this with concepts as simple as “first trade in the sectorâ€
Attachments
Rotational Example.zip
(252.47 KiB) Downloaded 1672 times

AFJ Garner
Roundtable Knight
Roundtable Knight
Posts: 2040
Joined: Fri Apr 25, 2003 3:33 pm
Location: London
Contact:

Mechanica

Post by AFJ Garner » Thu Oct 06, 2005 4:11 am

Dean,
I assume you own TBB as well as Mechanica. If so, is it possible in simple terms and without too much effort on your part to explain what Mechanica can do which TBB can not do with regards to dynamic portfolio selection.

nickmar
Roundtable Knight
Roundtable Knight
Posts: 191
Joined: Tue Oct 26, 2004 12:38 pm
Contact:

Post by nickmar » Thu Oct 06, 2005 7:27 am

I suspect that such a system can be implemented in the current iteration of TBB. Each sector would probably need to be represented as a seperate portfolio. The same system would then need to be applied to each sector portfolio and then combined into one meta system. The instrument ranking functions accessed through the Portfolio Manager Blox should come in handy here.
Sounds like an interesting concept to test - similar to a mutual fund switching strategy except that in this case the sectors are truly uncorrelated to one another (for the most part). Another addition to my "to test" list - thanks Dean.

Tim Arnold
Site Admin
Site Admin
Posts: 8537
Joined: Tue Apr 06, 2004 1:41 pm
Location: Boston, MA
Contact:

Post by Tim Arnold » Thu Oct 06, 2005 9:04 am

Yes, in fact the Trading Blox Portfolio Manager Blox is a dynamic portfolio manager. So every day you can apply rules, filters, and ranking to help determine which instruments to be trading, or swapping in or out.

I wrote a system recently that filters by liquidity, and then by strength, and then keeps the top x long and short (equal number) by strength, and gets into those positions. This algorythm runs evey day and gets out of positions that fall off the 'list' and swaps in positions that are on the 'list.' It also keeps the number of long and short positions equal.

I then increases the size of the positions that are in the top x%, and decrease the size of the positions that are in the bottom y%.

This portfolio filter can be used with any entry or exit signals, whether price or time based. You can also set the initial position with any strategy desired.

Tim

damian
Roundtable Knight
Roundtable Knight
Posts: 815
Joined: Tue Apr 15, 2003 8:43 pm
Location: dusseldorf

Post by damian » Thu Oct 06, 2005 9:09 am

Within TB I would like to be able to step between predefined (in Dean's terms, static) portfolios, where the definition of each portfolio could be anything. In this case the definition would be market sector. Then you would be able to take the next step as suggested by nickmar, that being using the instrument ranking functions accessed through the Portfolio Manager Blox to select candidates from each portfolio.

This would allow other simple powers, such as running a simultaneous test of n systems (possible in TB) with each system referencing their own predefined portfolio (not yet possible in TB).

I have not yet achieved anything of note in the field of dynamic portfolio selection research, both for lack of a sound method and also tools (and imagination).

Dean, for a guy with a trading account that started at only $25,000, you sure own some expensive testing software ;) (and one would hope so, given your status as earnest system developer).

Tim Arnold
Site Admin
Site Admin
Posts: 8537
Joined: Tue Apr 06, 2004 1:41 pm
Location: Boston, MA
Contact:

Post by Tim Arnold » Thu Oct 06, 2005 9:29 am

This would allow other simple powers, such as running a simultaneous test of n systems (possible in TB) with each system referencing their own predefined portfolio (not yet possible in TB).
Each system can reference their own predefined portfolio or dynamic portfolio in Trading Blox today. Did you mean each instrument can reference their own sector/group? That is available in 2.0.7.

With Trading Blox Groups, you can set each instrument to a group, and access many built in group functions, like group risk and group margin, as well as create your own group functions like a group ranking function that ranks instruments within a group for dynamic portfolio management.

damian
Roundtable Knight
Roundtable Knight
Posts: 815
Joined: Tue Apr 15, 2003 8:43 pm
Location: dusseldorf

Post by damian » Thu Oct 06, 2005 9:31 am

Yes, my mistake.

Slave2MYT$
Roundtable Fellow
Roundtable Fellow
Posts: 51
Joined: Wed Jul 21, 2004 4:45 pm
Location: Toronto, Ontario, Canada

Post by Slave2MYT$ » Thu Oct 06, 2005 10:48 am

With all of these new features Tim, would you kindly post sample code so we can see how this would be utilized, not asking for your proprietary systems but samples like your filtering by liquidity then by strenght etc..

Sample code helps the learning curve at least for me...much appreciated.

Slave
Last edited by Slave2MYT$ on Fri Oct 07, 2005 6:40 am, edited 1 time in total.

nickmar
Roundtable Knight
Roundtable Knight
Posts: 191
Joined: Tue Oct 26, 2004 12:38 pm
Contact:

Post by nickmar » Thu Oct 06, 2005 11:06 am

Tim - in order to implement the type of system that Dean described without using distinct sector portfolios, I suspect that additional functionality will be needed in the area of group management. For instance, there is currently no way to use the Ranking function to rank the instruments of a specific sector.

BARLI
Roundtable Knight
Roundtable Knight
Posts: 650
Joined: Sat Jan 17, 2004 6:01 pm
Location: USA

Post by BARLI » Thu Oct 06, 2005 6:11 pm

Slave2MYT$ wrote:Sample code helps the learning code at least for me...much appreciated.


that'd help me as well! :D

Tim Arnold
Site Admin
Site Admin
Posts: 8537
Joined: Tue Apr 06, 2004 1:41 pm
Location: Boston, MA
Contact:

Post by Tim Arnold » Thu Oct 06, 2005 7:53 pm

Absolutely. I'll post a bunch of examples on the Customer Support Forum next week.

Slave2MYT$
Roundtable Fellow
Roundtable Fellow
Posts: 51
Joined: Wed Jul 21, 2004 4:45 pm
Location: Toronto, Ontario, Canada

Post by Slave2MYT$ » Thu Oct 06, 2005 8:03 pm

Excellent...!!!

Thanks a bunch Tim.

Slave

Slave2MYT$
Roundtable Fellow
Roundtable Fellow
Posts: 51
Joined: Wed Jul 21, 2004 4:45 pm
Location: Toronto, Ontario, Canada

Post by Slave2MYT$ » Fri Oct 21, 2005 8:23 am

Hi Tim, just wondering if you had a chance to add sample code for this topic...

Thanks,
Slave

Tim Arnold
Site Admin
Site Admin
Posts: 8537
Joined: Tue Apr 06, 2004 1:41 pm
Location: Boston, MA
Contact:

Post by Tim Arnold » Fri Oct 21, 2005 8:41 am

Yes, thanks for the reminder. Got sidetracked on other things. I have a couple of good samples I can post today.

Slave2MYT$
Roundtable Fellow
Roundtable Fellow
Posts: 51
Joined: Wed Jul 21, 2004 4:45 pm
Location: Toronto, Ontario, Canada

Post by Slave2MYT$ » Fri Oct 21, 2005 8:43 am

Great...thanks. A gentle reminder once in a while is a good thing :wink:

Thanks again Tim...

Slave

AFJ Garner
Roundtable Knight
Roundtable Knight
Posts: 2040
Joined: Fri Apr 25, 2003 3:33 pm
Location: London
Contact:

Post by AFJ Garner » Sat Oct 21, 2006 3:53 am

I was recently re-reading Eric Crittenden's paper on long term trend following applied to stocks, which can be downloaded here: http://www.blackstarfunds.com/.

Eric quite understandably does not detail his money management strategy but dynamic portfolio selection seems crucial.

The portfolio is not curve fit, not selective, not narrowed in any way other than by excluding penny stocks and low volume counters. No exclusions for dull utilities, no concentration on high beta. Given the huge universe of stocks, even with a very large capital base, the available car parking spaces get rapidly filled. To use Dean's analogy.

It would seem advantageous to test many different methods of freeing up spaces when new signals are given. Selling loss making positions, scaling down proportionally overall, selling the most profitable positions, selling random positions. Many possibilities need exploring.

LeapFrog
Roundtable Knight
Roundtable Knight
Posts: 695
Joined: Mon May 17, 2004 4:18 pm
Location: Boston, MA

Filling and shuffling parking lots.

Post by LeapFrog » Tue Nov 14, 2006 3:30 pm

Thank you Dean for initiating this thread and posting some example results. I've been recently examining this area as a further way of justifying purchasing TB and wonder how far users have gotten with actually doing Dynamic Portfolio Selection (after Tim posted some examples).

As Dean's examples show, there is a VERY SUBSTANTIAL benefit to be gained by using this approach. Either by reducing drawdown or boosting ROI or both (MAR).

I'm surprised really that more postings about this are not showing up - is it because it can't be done in TB or because nobody is really doing it?

Anyway, I think Dean has pointed out something that could be of great benefit to us all. Just my 2 cents.

Forum Mgmnt
Roundtable Knight
Roundtable Knight
Posts: 1842
Joined: Tue Apr 15, 2003 11:02 am
Contact:

Post by Forum Mgmnt » Tue Nov 14, 2006 4:35 pm

The Turtle System always had this concept. Rich and Bill called them correlation limits. We could trade all the markets up to 4 units each and up to 6 units in each group of correlated markets.

This effectively limited exposure and drawdowns. We'd generally only be long two or three at most of the 6 currencies we followed for example. My testing indicates that this improves performance substantially for the Turtle System.

This is one of the reasons that people have generally not reported the true results of the system, they look at the results for all the signals across all the markets and we'd have generally only taken a portion of them and then only in the strongest or weakest markets.

Trading Blox comes with a Unit Limiter Block that you can use to apply this logic to other systems.

- Forum Mgmnt

Dean Hoffman
Roundtable Fellow
Roundtable Fellow
Posts: 87
Joined: Wed Jul 02, 2003 3:04 pm

Post by Dean Hoffman » Tue Nov 14, 2006 11:30 pm

I wrote an article that talks about the concept (as I use it) in a bit more detail here:

http://www.traderstech.net/The%20Small% ... undrum.pdf

Regards,
Dean Hoffman

sluggo
Roundtable Knight
Roundtable Knight
Posts: 2986
Joined: Fri Jun 11, 2004 2:50 pm

Post by sluggo » Wed Nov 15, 2006 8:07 am

It's interesting that the money management firm Transtrend claims to trade 184 differrent futures markets ( http://www.transtrend.com/DISCLOSURE_DO ... 06-V19.PDF ) in their $10 million managed accounts for US clients (30 more mkts for non-US). I suspect this is just a list of the markets that they "monitor"; some of the markets on Transtrend's list have very low volume and open interest. Still, it's nice to know what the big boys are watching.

After reading the article Dean mentioned above, I felt there was an inconsistency among these four statements
There are over 100 tradable commodity markets worldwide
Small accounts are able to efficiently trade markets that would be far too illiquid for large accounts
(Dean Hoffman offers) a trading program that monitors and trades over 70 diversified commodity markets
If an exceptional opportunity develops in a market outside of your predetermined portfolio wouldn't you want to take advantage of it? By trading with our strategies you don't arbitrarily rule out any market that may perform well in the future AND you have eliminated the tendency to pick a portfolio based merely on past performance (curve fit) considerations.
See what I mean? Statement 4 says "don't arbitrarily rule out any market", but statements 1 and 3 say that Dean has arbitrarily ruled out more than 30 markets; he doesn't even monitor them.

Why? It's not because of illiquidity; statement 2 takes care of that. It's not because those other 30+ markets are "too big" or "too foreign" or "too closely correlated" or "historically too unprofitable" or "just too darn many"; those are arbitrary excuses for ruling out markets, which statement 4 forbids. I haven't yet figured out a self-consistent reason why dynamic portfolio selection should be applied only to a subset of less than 70% of the tradable commodity markets worldwide.

Sounds to me like Dean has simply got himself a "predetermined portfolio" of 70 markets and some rules for skipping lots of trade signals.

Post Reply