You know roughly what the future looks like, but it only takes a tiny (sub-millisecond) amount of error (due to user movement, clock drift, wireless lag etc etc

You know roughly what the future looks like, but it only takes a tiny (sub-millisecond) amount of error (due to user movement, clock drift, wireless lag etc etc) to throw the phasing off and ruin the effect. So you still need some combination of precise modeling/listener positional data and/or a sophisticated algorithm that uses an input signal to detect and correct the error and keep you “out of phase”