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: Gary E who wrote (687)	10/8/1999 7:19:00 AM
From: LastShadow	Read Replies (1) \| Respond to of 871

There isn't any difference for creating a net for QQQ or SPY than there is for any other stock, with the exception that volume, if you use it in other analysis is even less important. Use of a neural net presupposes that one understands, or at least suspects, what the underlying conditions/relatinships/values are indicative of future price movement. Without an understanding of what forces drive price change, the net is just so much math to process and as a tool is no more useful than any other in a trader's arsenal. So, if from technical analysis it looks like a stochastic, or a trading band crossover, or close/lag relatinship of the OHLC (or some combination of them) appears to precede a price movement in the direction you desire, that is what one will use to forecast probability. you set up a test of the conditions you seek and let the net tell you if, historically across some data it hasn't seen before, does in fact generate valid signals. Then you vary, add or subtract the inputs to tweak the performance such that forward testing gives a result you are comfortable trading with. My personal opinion is that the spyders are more predictable as they are groups of like stocks, such as tech or cyclicals or financials, etc. Using the whole market, or at least a broad based indicator like the SPX, one has to account for many non-market events, and several collateral events (Al's speeches) as well as general sentiment shifts in order to effectively filter your net's tools. The first step is to trendline your financial instrument - whether that is a bond, spyder, stock or index, and see if there is any cyclicity to its movement. From there you look for big-picture movements - moving averages do this rather effectively for longer time frames. Then as you focus on the near term, you look for local influences and the indicators that precede or at least appear to coincide with price movement and attempt to forecast those. The development of a neural net for trading (or anything for that matter) is a science. The effective use and interpretation of it is still an art... lastshadow