The simplest algorithm to find the vertex in a symmetric collision would be to find the center of gravity (CG) of all hits. It is not clear how well this should work in the asymmetric case of p+Au. However, by taking the difference between the CG in the inner and outer barrels, if could be possible to project towards the vertex in this case. Information from parallel strips only was used to find the vertex from the CG, using two iterations. First, the whole hits distribution (see the left sides of figures 4-5) was used. Then an equal number of channels above and below this initial CG were used to improve the vertex position measurement.
There is a statistical limit to this method. The minimum
uncertainty on the vertex
position is 
(rms width)/
0.4cm for central Au+Au
collisions, which is below the vertex resolution required at the trigger level
( 1cm), but insufficient for the offline analysis ( 1mm). The
actual vertex resolution found using the CG method was 
=2.1cm for
Au+Au. For p+p and p+Au, the number of hits on the vertex detector is much
smaller and the statistics do not allow a sufficiently accurate estimate of the
vertex position using the CG only, especially when random hits due to noise are
included.