Meet director says 25% of the people set their PBs at CIM:
http://coachjoeenglish.wordpress.com/2007/11/27/interview-california-international-marathon-race-director-jim-mansoor/
This is actually quite meaningless without knowing how many prior marathons the typical participant has run.
For instance, if the typical participant has run one prior marathon, and we further assume that the meet director really means "conditioned on having run a marathon before (thus excluding first-timers) 25% of participants set a PR" - 25% would be much *lower* than expected if Cal was a marathon of average difficulty. In fact, we would expect that a full 50% of participants with one prior marathon experience should PR on that course.
Conversely, if the typical participant has run 10 prior marathons, 25% would be much *higher* than the fraction expected if Cal was a marathon of average difficulty. In this case, we would simply expect that 10% of participants would have their PR on that course if Cal was a marathon of average difficulty.
Of course, this discounts the fact that marathoners generally tend to improve their performances over time (which would thus "boost" the PR rate of any marathon over the random expectation).
The statistic is not even very comparable across events. For instance, the Boston marathon has been extensively discussed in this thread as a marathon with a large net downhill. Thus, one might like to compare the PR rate at Cal to the PR rate at Boston. Of course, this would inappropriate since the average experience level of Boston marathoners is likely to be considerably higher (due to qualifying standards) than is the average experience level at Cal.
The statement that "25% of participants PR on our course" thus strikes me as a seemingly impressive statistic that is being used to promote a race - but is actually totally uninformative in isolation with regard to the difficulty of the course.