There are a number of improvements that could / should be made to the tracker pattern recognition, which we currently lack the man power to implement:

1. When performing the last squares fit, all possible spacepoint combinations should be tried and the one with the best chi squared selected, not just the first that passes the cut

2. After a track is selected a search should be made of any remaining unused spacepoints to see if they lie within a road of the track. It so they should be added to it.

3. Possibly the cuts might benefit from further optimisation, but this is debatable.

4. A second method of pattern recognition would be interesting e.g. a Hough transform