Silicon Investor (SI) -- The First Internet Community

STOCKTALK

We've detected that you're using an ad content blocking browser plug-in or feature. Ads provide a critical source of revenue to the continued operation of Silicon Investor. We ask that you disable ad blocking while on Silicon Investor in the best interests of our community. If you are not using an ad blocker but are still receiving this message, make sure your browser's tracking protection is set to the 'standard' level.

Strategies & Market Trends : Neural Nets - A tool for the 90's -- Ignore unavailable to you. Want to Upgrade?

To: F Robert Simms who wrote (636)	8/25/1999 12:44:00 PM
From: Larry Livingston	Read Replies (1) \| Respond to of 871

I'm curious to know what levels of correlation people are getting between actual and network output. Naturally it's a moving target and a high correlation doesn't guarantee a good trading system, but I've read that when you're trying to predict percentage moves anywhere from one day out to two weeks or the slope of a projected moving average, a correlation of .2 that holds up consistently over several data sets unseen by the network and in large enough samples "can be very useful for trading". I'm wondering if anyone has achieved this magic .2 correlation or better and whether it's true, as even at that degree you are still getting a fairly high number of false predictions. I suppose you're looking at an odds strategy, and a small edge over time if you have large amounts of capital like a casino will pay off. What kind of capitalization do you need to play a casino strategy? As for me, I occasionally get .2 correlations or better but they tend to be statistical flukes. Generally my results tend to be in the .5 to .12 region, but then I'm trying to predict the next day move which is notoriously difficult to crack if not impossible. Surely some of us are getting good results. I would also be curious to know if any of you have done any work with predicting the slope of a projected moving average?