In the Critical Chain world, we also like simplicity. We don't need to use Monte Carlo to deal with integration effects of feeding chains since feeding buffers help to isolate the critical chain from variation in the feeding chains. We also eschew earned value measures that, when they're not driving the wrong behaviors, provide meaningless guidance (SPI approaching 1, for example). There is also no basis in reality to say that past performance is a good predictor of future performance.
In a CC environment, project project completion is easy. A simple recalculation of the project buffer for the remaining critical tasks plus remaining critical chain length would give one a reasonable view of project completion. If there was, my IRA portfolio would be in a lot better shape.
By the way, this all assumes that the project has also been isolated from interference with other projects through shared resources, through something like synchronized multi-project management. If you're living in a multi-tasking world, all bets are off on any prediction of completion.
