For a subset of black hole collisions, where the black holes collide head-on, there exist as of today (Feb 1999) reasonably reliable full numerical simulations of the collision. Because of the lack of an inspiral phase, the waveform profiles of these kind of collisions are completely dominated by the ringdown of the final black hole [for a recent reference see Anninos, Brandt, and Walker  (ABW). Even for this simple case, there are some discrepancies between various numerical codes.
A separate approach to black hole collisions has been the close-limit approximation (see Khanna et al. gr-qc/9905081 for a description of the close-limit approximation applied to inspiralling black holes), which describes the merger of two black holes as a perturbation of a single black hole; the perturbation is based on a small parameter measuring the separation of the two black holes. This approximation has had an uncanny degree of success in replicating (at least for the head-on case) numerical estimates of the merger waveform, and it provides ``a little-bit more'' of the merger waveform than the quasinormal ringdown. It is useful to examine how good ringdown filters will work in detecting the more realistic waveforms produced by the close-limit approximation. In this addendum we will describe the use of both close limit and full nuemerical simulations in the GRASP package.
There are two data sets containing full numerical waveforms for head on collisions of two black holes released initially from rest. These correspond to the two codes described in ABW  for a moderate separation of the holes ( in the ABW  notation, about 10 in terms of single black hole radii). These waveforms are shown in figure .
These two waveforms correspond to the two codes described in ABW , based on two different set of coordinate systems. To a certain extent, they exhibit the limitations of the state of the art in numerical relativity: both waveforms represent the best effort by correct, ``convergent'' codes, and yet there is some disagreement among them. This disagreement can be settled by comparing with the close approximation (see ABW ). But it is also instructive to compare how bad the disagreement is from the point of view of a data analyst.
The fitting factor is defined by
From the point of view of source modelling, the difference in head-on collisions between the full numerical waveforms and the close limit ones is not substantial.