Yep, understood. However, my point was that (I would guess) your current bottleneck is the overlap tests. The pseudocode I posted is essentially doing a ‘Point inside cone’ test, which is a great deal more simple than a cone/sphere overlap test (especially if it’s a mesh approximation to a cone!). So what I was suggesting is that you try just removing the collision shape from your missile, and for now just use a TActorIterator on the world to give you your candidate set (ie. everything). Sure it’s super suboptimal, but you may find that doing a much simpler test on a larger candidate set works out to be a lot more efficient than expensively maintaining a small candidate set. Add to that some of the more simple optimizations suggested, and you may not even need to bother with spatial subdivision (though if you eventually want thousands of both missiles and targets, you probably will).