The problem with answering that question, Petra, is that one has to define what a
"success" is. Without a satisfactory definition of that, anything said about the
"number" or "percentage" of successes is not very useful. It seems to me that
the criteria for success could be determined by a discussion among members of
this board.
Once it is has been determined what constitutes a successful forecast/prediction,
then it might be possible to answer your question.
It is also then necessary to consider the number of "successful hits", the number
of misses and the number of false alarms - i.e. predictions for which no event
coincided. This is another topic for conversation for this board. What is an appropriate
percentage of forecasts which must be "successful" before the method is considered
useful or successful. Furthermore, what is an inappropriate number of false alarms
and misses to cause the method to be dismissed as unreliable?

On a personal note, I think that Don has beat the odds on too many forecasts with
too few misses and false alarms to have his method dismissed. It certainly seems
to me that the statistics are in his favor at this time. His method does seem to consistently
be able to pick the approximate time, magnitude range and location of earthquake
in California and other areas of the western U.S., and that is a remarkable thing, given
the abyssmal record of many previous attempts by well-educated people to do that. He
is not always perfect, some events he misses and he has had false alarms, but these
are much fewer than the forecasts that appear to describe real events - however
success may be described.
I would encourage you, Cathryn and Don (and others who are working with Don
to fine tune the forecasts - like 2-cents, Chris, and EQF among others) to continue
along the same general path. It looks like you may be getting somewhere. At least
I haven't seen a long string of negative comments about earthquake prediction in the
past several months on this board.