From a post to DAVIDWEBER.NET forums on 8/16/2011

Predictability of a ship's position

    The predictability of the ship's position within the wedge is somewhat greater than some people seem to be assuming, but unless you can target the throat or the kilt, "snap shots" (even with SDMs) have always been hard to come by. (Go back and look at what happens while Fearless II closes with Thunder of God in HotQ. Or, for that matter, at my comments about SDM BCs closing with an SDM SD while covering behind their wedges.) This is one reason that missile combat was regarded as indecisive prior to the Havenite Wars. The Manties' (and Peeps', although to a lesser degree) improvements in laser heads and the ability to share information between missiles were two of the main reasons missile combat started moving back to the fore.

    One of the reasons for the telemetry control links has always been to keep missiles informed of things their own sensors simply cannot see. The tactical sections of warships are responsible for modeling the entire combat environment (including prediction of the target's position within its wedge) and feeding that to the missiles which have not yet attacked. Mutual wedge interference cuts big holes in the ability of individual missiles to talk to each other, but all the missiles report back what they "see" to the ship which launched them. That data is usually out of date for light-speed telemetry by the time the ship receives it, but it uses it to build a predictive model for the target's position, which is then passed on to the missiles still in the firing queue. The missile's onboard sensors have to be able to "see" the target ship (or at least its wedge) well enough to adjust its position within the "dated" information provided to it, but the data itself is absolutely critical if it's going to have a hope in hell of scoring a hit anywhere except down-the-throat or up-the-kilt. Even with the best models, however, fire control can only predict a basic "zone" in which the target is going to lie, so part of the trick has been to saturate that zone with laser heads. If your missiles get an opportunity to establish positive lock, great! In those cases, you are virtually guaranteed a hit unless the missile is picked off before it can fire or the sidewall (if any) bends the beam into a miss. Far more often, though, each missile is firing at a target it never actually sees because its telemtry links told it "Look for the target somewhere around… here. Oh, and your buddies are going to be firing at other points in the same zone. Good luck, guys!"

    One of the huge advantages of Apollo is that for the first time, the telemetry data can be real-timed both ways, allowing for much more effective "steering" of the missiles. This is also one of the reasons the combination of Ghost Rider platforms and the Mark 16 are so effective; the delay on corrective data from the missiles that "miss" because of faulty target prediction can reach the firing ship in half the time. What we are going to see the Manties' SD(P)s doing here shortly is hanging Ghost Rider platforms close to their targets whenever possible and using them to provide a continuous stream of updated data better even than Apollo can provide without such assistance. On the other hand, you've already seen an example of the way in which Apollo can be used in this same role from Mike Henke's sim in Storm from the Shadows.