I don't evaluate earthquake predictions in any kind of useful way, but if I did, I'd be a lot less demanding. If someone were to make a prediction of a M8 quake on a subduction zone, I would give them a LOT of leeway on the location, and would not care about depth hardly ar all. That is because such a quake ruptures such a large area, and ruptures a whole range of depths...often from the sea floor to maybe 30 or 40 km depth for a typical interplate quake. That is why I was a bit "bent" about press coverage of the Dec 24 M9.3 ...because epicenter meant almost nothing to where the tsunami did damage or where the quake itself did damage. As for deep slab quakes (200-700 km, for example), sure, depth matters, but the epicenter in many areas gives the depth, because it the dip of the slab is generally known.
I suppose one could say that the epicenter does matter if the predictor is using triggering, or assuming that the pre-earthquake signal comes from where it will start, but then the predictor would not be able to tell if it will be a M3 or a M9. Yes, I understand that probably no one is successfully predicting such quakes except by chance.
For California strike-slip quakes, if you are predicting M6 or larger, depth is not generally going to matter either, in the same way as I discuss above.
Chris