2 replies to this topic

[Islander]

Russel 2000. It is coming up on resistence, its up trend line, and there is a gap just above that needs to be filled at 818 or there abouts. If it breaks up thru both I think there is a new up trend. For now it is too early to tell. Tomorrow might be the day. The uptick in ST bonds at the close is not helpful, I think the new house sale figures are being considered. My NNets say there is a good chance of a breakout tomorrow. Best, Islander.

Edited by Islander, 26 March 2007 - 08:49 PM.

[LeroyB3]

If you don't mind, what are your neural nets based on? Best, LB

[Islander]

If you don't mind, what are your neural nets based on?

Best,

LB

Please forgive the delay. The nets are trained based on a wide spectrum of correlations of market variables we run. The best correlated, independent, variables are then used to teach the network now to forecast a change in the dependent variable (usually an index). We then run the networks and take the output as trending value which suggests where the markets we study (SP5, R2000) are going over forecast horizon (week to ten days at most). What we do not do well is deal with news driven markets, such as we have had recently. NNets are best in reasonably stable market setting that has a trend. They are good at deteching early changes in trend. Recent volitility has been rasing hell with our forecasts which we sell to clients. Hope this is clear, but I am trading and I am pressed. I wrote a book on this, but it is not popular reading. Islander.