Strategy Investing: Black Boxing Fundamentals

C

Corey

It seems to me that all discussions around mechanical systems tend to strictly revolve around technical systems. Not only are they strictly technical, but often the parameters used exist only for the shortest of time periods before being invalidated by the next set.

I sometimes wonder if this is due to the software currently available for strategy developers -- certainly getting price and volume information (allowing you to calculate endless technical indicators) is much easier than getting historic fundamental information.

I spent my summer developing a strategy system that invests based on fundamental information. One of the largest issues I faced was that fundamental information is always inherently lagging -- if news breaks between quarters, your system may not recognize the fundamental change until the next 10Q. So I also rolled in a primitive trend system that helps identify when price and fundamentals are disjointed -- normally indicating that the next quarters fundamentals would not be as good.

In back testing, this system did extremely well -- from 1990 to 2007, it returned 1300% on extremely low volatility -- taking very few (<20) trades a year and having over 85% trades successful. However, the majority of these 'trades' were actually held for a year or more, making my system a strategy investment vehicle, instead of a mechanical trader. Before you quack 'curve fitting', I only 'taught' the system on the years 1990-1995. So I know that mechanical investment systems can be written -- and they can be simple. My system has a total of six indicators ... and it does a pretty good job of weeding out the winners and losers.

But as I said ... this is an investment system, not a trading system. I am very interested in the application of fundamentals -- and specifically the divergence of fundamentals and price -- to trading systems.

So I was wondering if anyone had effectively developed a mechanical trading system based on fundamentals ... or even if anyone else had developed a mechanical investment system and wanted to talk about the issues of finding data for back testing and results.

Thanks, and happy developing!

-C

EDIT: I wanted to clarify that my results did NOT include slippage or tax. Slippage was not of much concern due to the fact that the system was for investment, not trading. As for tax -- well, it all depends on who is using the system :D But there are always tricks for minimizing tax, and I didn't want to try to roll those rules in -- so I just assumed the system was trading in a tax-free account, like an IRA.
 
As I said, I do use both. I use fundamentals to identify my targets, then technicals to confirm my analysis. No matter what you think the price should be, price movement will always determine the value of your investment.

I have spoken to representatives for both CompuStat and Factset. They seem to have an oligopoly on the market -- one of them, for 20 years of quarterly fundamental info and daily price info on only 300 companies, quoted me a price of $60,000+ for a years LEASE on the information. That is right ... after a year, I had to give the info back.

The information that they have is great ... but as an individual investor, I can't afford to pay $60,000 for that sort of info. Absolutely absurd. Though, I suppose for big banks, it is pennies...
 
Portfolio123.com

Free backtested fundamentally driven systems.

-IMO, all are likely overoptimized and focused excessively on small porfolio sizes (under 100k) low and liquidity in the positions. Sharpe ratios all look good on the big winners, but I think if you want to make a career from trading on that information, youll need more than 10 illiquid stocks to get started. But check it out.
 
Quite surprised at the cost of Factset. I had it for a while when I was working at an IB here in London but didn't think it was that costly.

I assume you've checked out Bloomberg. Cost is certainly less that you have been quoted for Factset but not certain how far back their fundamental data goes.

BTW - something to be aware of is that fundamental databases tend to get updated over time. Of course this is done to make the data more accurate but the figures you see now for (eg) 2Q 2007 may not be the same as were immediately stated post results. These updates can be made months or even years after the fact which should add a "treat with care" label to results derived from your testing.

All sounds interesting though - good luck
 
RhinoTrader -- the 'updated' databases are actually a big concern for back testing. There are a few places that do offer non-altered historical databases, though -- which is necessary for realistic back testing. I was lucky enough to get my hands on one this summer.

Anyway, thanks for the responses. Anyone else use fundamentals in their mechanical systems, or am I the only nutjob?
 
I've been doing (almost) purely mechanical investing using fundamentals for the past 10 years. A very straightforward strategy that I've adapted several ways is to simply buy the 10 most undervalued stocks at any given time. Sell anything you wouldn't buy.

I like to use the EBO valuation method. This is strictly fundamental no technicals. I've found that adding some momentum parameter usually helps too but not needed.

Anyway, my methodology is to develop a strategy, backtest the strategy to improve it, then implement. Stick to the strategy. I don't change the recommended trades. Every month or two I go back and refine my strategy, re-test, and use that new strategy.

Finding a good data source is difficult. There doesn't seem to be many out there. I use Stockworm.com. They have tools for developing pretty complex strategies (which can include fundamentals and technicals) and then can backtest them using different pricing models.

It was already mentioned but portflio123.com is another site. Their rule setup didn't seem as intuitive to me though.

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