Weight assignment to indicators

Discussions about the testing and simulation of mechanical trading systems using historical data and other methods. Trading Blox Customers should post Trading Blox specific questions in the Customer Support forum.
Post Reply
Chessman
Full Member
Full Member
Posts: 10
Joined: Fri Aug 01, 2003 11:59 am
Location: New York City
Contact:

Weight assignment to indicators

Post by Chessman »

Here is a Hedge Fund that day trades, has consistent track record of monthly gains w/ very low drawdowns < 4%.

http://www.ironwoodassetmgt.com/

I am trying to get a better handle on their technique used. Just wondering if anyone has experience with weight assingment to indicators?

I have always believed that simplicity works in a trading system. However Rao's returns are very impressive and trading system complex.


Ironwoods Investment Strategy:

Each trading signal that is generated is given a weight determined by the number of models that are simultaneously giving buy or sell signals. Only trading signals whose weight is higher than a predefined parameter are executed. Those trading signals in which the signals of the individual component models are conflicting are ignored. The Advisor believes this synthesized approach produces trading signals that are stronger and more reliable than those originating from single-factor systems. Once a buy or sell signal is given, the Advisor will then determine an appropriate entry point in the stock index future (Nasdaq-100 or S&P 500) that is giving the stronger signal.
edward kim
Roundtable Knight
Roundtable Knight
Posts: 344
Joined: Sun Apr 20, 2003 2:42 pm
Location: Silicon Valley / San Jose, CA USA
Contact:

Post by edward kim »

I think a system's performance is robust if you can vary the parameters across a broad range (e.g. 40 works really well, but so do 20, 30, 50, and 60.) In a similar way, if an indicator is good enough to be used partially, it is good enough to be included wholly and not just in part.

Think about the Turtle System and what would happen if you partially weighted the "Trade if last is Winner" indicator. To me, and indicator is an indicator, so how would I weight this at only 50%? It really doesn't make sense.

I keep all indicators with a binary constant (0 or 1). If I start weghting the indicators in my system, it doesn't hold up very well if I vary the input parameters. My system is long-term trend-following in nature, so other people might have differing experiences.

Edward
Mark Johnson
Roundtable Knight
Roundtable Knight
Posts: 122
Joined: Thu Apr 17, 2003 9:49 am

Post by Mark Johnson »

Pick an odd number, say, five. Imagine five variants of the Original Turtle System "OTS" which operate over different timeframes. For example: OTS(10,5) and OTS(20,5) and OTS(25,10) and OTS(40,15) and OTS(55,20). Treat these five different turtle systems as indicators which put out the number +1 if that variant of the Original Turtle System is theoretically long, zero if they are theoretically flat, and -1 if they are theoretically short.

Now each day you run all five systems (um, I mean, indicators) and collect the five output values. Add them up. If the sum is greater than zero, go long. If the sum is exactly zero, go flat. If the sum is less than zero, go short.

Thought problem: why an odd number?
Sir G
Moderator
Moderator
Posts: 243
Joined: Wed Apr 16, 2003 12:21 am
Location: Salt Lake City, Utah

Post by Sir G »

Thanks Mark-

That seems like a nifty idea..

I've put that idea on my to-do list.
Thought problem: why an odd number?
I'll take a stab at it...

Because you need a tie breaker?? Two indicators short =-2 and two indicators long=+2 has you flat... but throw in a fifth and the scale will tilt most of the time.

Gordon
Phil Hees
Contributor
Contributor
Posts: 4
Joined: Mon Aug 11, 2003 4:09 pm
Location: New York City

Post by Phil Hees »

Edward,

William Eckhardt took the same position in his Market Wizards interview: if an indicator is good enough to be used at all, weight it equally with the other ones.

If I understand the Ironwood strategy correctly, it adheres to this approach in that each model generates a buy/sell indicator that is then weighted equally with the ones generated by the other models; the fund goes long or short if a sufficient number of models vote in the same direction simultaneously.

In other words, their voting model corresponds to the Senate rather than the House of Representatives. Virginia's vote counts the same as Rhode Island's, and Ironwood's counter-trend model carries the same weight as their sentiment indicator.

Phil
Phil Hees
Contributor
Contributor
Posts: 4
Joined: Mon Aug 11, 2003 4:09 pm
Location: New York City

Post by Phil Hees »

Mark,

What you are describing is a fund with N portfolio managers, each of whom makes a decision to go long, short or flat each day. To avoid paying commissions and incurring slippage they swap positions amongst themselves whenever possible, and only place orders with their broker if their aggregate position changes. To complicate matters somewhat, there is a corporate risk manager who limits the fund's overall position to +/-1. In this example, I don't see an obvious advantage for N being an odd number. Am I missing something?

Anyway, this is not what Ironwood is doing, if I've understood their prospectus correctly. They are using a voting model where a position (+/-1) is only taken if some minimum number T of these portfolio managers all vote in the same direction. Depending on the values of N and T, this model may behave much differently than your example. For instance: N=5, T=4; yesterday the portfolio was flat; today three portfolio managers want to go long and the other two want to be flat. In your example the portfolio would go long (+1), whereas Ironwood's voting threshold of T=4 means the portfolio will remain flat (0).

Phil
Chessman
Full Member
Full Member
Posts: 10
Joined: Fri Aug 01, 2003 11:59 am
Location: New York City
Contact:

Post by Chessman »

Thanks for all the replies.

Interestingly Ironwood suffered their biggest one month loss last month. Down almost 5%. I am no longer able to view their performance, they recently changed it to a password protected page. But IASG has it.

http://www.iasg.com/intro.htm

I find quite a few similarites between Ironwood and Mark Weinstein, featured in the original mkt wizards book. Both synthesize quite a few indicators and have a large discretionary element to their trading.
jimsta
Full Member
Full Member
Posts: 13
Joined: Sun Jun 15, 2003 2:45 am

Post by jimsta »

I have seen an example where many different moving averages are used at once, When they all have the same value it is taken as a signal. It looked impressive.
Anything that distracts one from the price change looks risky to me. While busy watching this "indicator" the eyes are not on what is happening to the price, price change equals account change.
The simpler the entry signal the better. Providing the exit with correct risk management is how profits are protected and losses reduced. This is hard enough without worrying about why you want to take a trade.
Post Reply