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 : Systems, Strategies and Resources for Trading Futures -- Ignore unavailable to you. Want to Upgrade?

To: Tom Trader who wrote (7586)	10/31/1998 8:46:00 AM
From: Patrick Slevin	Read Replies (1) \| Respond to of 44573

Looks as though I'd be better off trading stocks Message 6216471 The guy in the lead is a dog, by the way. (Seriously) Life is stranger than fiction.

To: Tom Trader who wrote (7586)	10/31/1998 7:32:00 PM
From: Patrick Slevin	Read Replies (2) \| Respond to of 44573

Tom, I lifted this from a discussion about training systems. I though you might be marginally interested in the comments. The topic is concerned with the length of time one should train a system. The following is attributed to a gentleman named Howard Bandy. ~~~~~~~~ You are asking about one of the key problems encountered when trying to model non-stationary time series -- that is, time series which have characteristics which are different in one period of time than another. If the market is characterized by a stationary time series, then using as long a training period as possible will result in a lot of trades, a high level of confidence, and likelihood that the model will perform well in the future. If the market is characterized by a non-stationary time series, the time covered by the training, testing, validation, and trading periods must have sufficiently similar characteristics that using the model in the trading portion will be profitable. That is, that the market is near enough to being stationary over that time. Since the financial markets change characteristics quite rapidly, using a shorter amount of time helps keep the model in phase with the market. The short answer is: use as much data as possible to get as many trades as possible to build as much confidence as possible, but don't use so much data that the market behavour is different at the trading end than it was at the training end.