Idling away I decided to look at the CP neural algorithym. The rules I looked at were:

1) It must be the top CP selection with a minimum of 40 CP points.
2) It must have had at least two career starts.
3) It must have placed at last start.

The results over the past two days are not too bad.

There were 10 selections for 5 winners for a return of $14.50 using Unitab divvies. And 8 placegetters for a return of $13.70.

Backfitting can produce attractive results but there's only one way to see if the method is worthwhile - and that is to record the selections and the results.

I'll list today's selections later, if there are any. I'm more interested in Place betting but I'll keep the each-way results if they both show promise.