% Accuracy

G

GIG

Hi All,

I'm a Computer Science major, and have been focusing my research efforts on linear time-series analysis to try and make short term (3-5 day) predictions in the stock market.

I've implemented four (4) different algorithms, appropriated from technical analysis techniques.

I'll just give a quick overview, but if you want more information please let me know.

I've randomly selected 500 companies listed on NASDAQ. For each company, I download all available historical data. Then, I let the algorithms loose on the data.

When an algorithm finds a potential entry point, it logs relevant information into a database table. To validate the prediction, the system looks 5 days 'into the future' to see the stock price.

3 out of the 4 algorithms, over a long term period, don't do all that well (~47% accuracy). One algorithm with a given parameter set does well (~52% accuracy) over a long term period.

Since I've been able to show that there are 'greater than chance' probability 'pockets' in the stock market, the academic side of the project has been quite sucessful.

I would like to get an idea of how accurate experienced traders are as a metric to the potential benefit of the system.

Obviously there are many more factors to consider other than accuracy (but I'm not worried about them with this post), I'll probably leave those for other posts.

So, if anyone would like to share their experiences, it would be most appreciated.

Regards,

Brandon
 
>>3 out of the 4 algorithms, over a long term period, don't do all that well (~47% accuracy). One algorithm with a given parameter set does well (~52% accuracy) over a long term period.<<

They can all be very profitable with good money management techniques.

Suggest you read up on money management and if possible backtest the system.
 
"Suggest you read up on money management and if possible backtest the system."

That's actually going to be my main focus in my thesis in the Fall :)

Brandon
 
Quote from GIG:

That's actually going to be my main focus in my thesis in the Fall :)

Brandon
More...

My understanding is in academia circle you'd better compare your results (% accuracy) against other research articles/thesis.

Finding a benchmark through a (formal or informal) survey (such as this thread) could cause you another thesis.

You must be very much focussed within your topic. Talk to your project supervisor first.

:confused:
 
I hope you enjoy what you are doing.

If you ever do another project where making money is the focus, there are several aspects you need to consider that are not on the table.

I am guessing your courses are oriented to gaining facility in using a computer somehow. dopes your school encourage you to check into academic areas where computers are applied as support tools?
 
You're right.

Traditionally, you would have a control group and an experimental group.

Unfortunatley, it has been painfully difficult to find research papers with the results I'm looking for.

Mind you, in the research context of my project I'm not interested in whether algorithm XYZ is better than ABC. The main focus was to determine whether any linear time-series analysis techique could provide greater-than-chance probability in determining stock market movement. So, in my case it's more of a "yes or no" kind of paper.

The %accuracy of other individuals/methods is supplemental to the project, but it would be nice to be able to put my results in some kind of context.

Brandon
 
bubba7,

Yes, doing this stuff is really fun.

What I'm presenting in this post is just the 'tip of the iceburg' so to speak. I've taken into consideration other factors as well, such as

-percent return per transaction
-percent return (only when correct)
-percent return (only when incorrect)
-smart percent return (drops the stock if it dips below 1.5%)
-intraday volatility
-yearly results for the above
-the effect of the "next day's" opening and closing price
-price movement between the 'buy' and 'sell' day.


There's probably a few more that I haven't added. Anyway, as you could imagine if I wanted information about all of the above I would have a post 10 pages long :)

Looking ahead for the fall, I'm planning to focus on
Money management
Slippage Effects
Commision Effects

If there are crucial metrics I've left out, please let me know.


All of this research has been implemented as one big software system (I've been developing the application for quite some time now).

There's quite a few electronic resource available, including a supercomputing facility (which can speed up analysis quite a bit), access to electronic publications, other universities, etc...

Brandon
 
Quote from GIG:

The main focus was to determine whether any linear time-series analysis techique could provide greater-than-chance probability in determining stock market movement.
More...

I would still think the above area is propably too broad. It seems I read a paper in the past that is roughly related to your topic. I believe your very first priority now, rather than to do a survey here, is to find out the best known scholar and his/her papers that are directly/closely related to your current research (in order to get a minimum pass grade).

You have to do a proper search on the WWW to proceed.

Good luck!

:eek:
 
Quote from GIG:

You're right.

Traditionally, you would have a control group and an experimental group.

Unfortunatley, it has been painfully difficult to find research papers with the results I'm looking for.

Mind you, in the research context of my project I'm not interested in whether algorithm XYZ is better than ABC. The main focus was to determine whether any linear time-series analysis techique could provide greater-than-chance probability in determining stock market movement. So, in my case it's more of a "yes or no" kind of paper.

The %accuracy of other individuals/methods is supplemental to the project, but it would be nice to be able to put my results in some kind of context.

Brandon
More...

How did you get past the data problems? Linear Time series analysis usually apply to continuous functions. And markets are anything but linear except for applying different kinds of series and synthesizing (or analyzing).

What range of data types and kinds are you allowowing yourself to use?
 
You must log in or register to reply here.
Back
Top