We compare alternative computerized model-selection strategies in the context of the vector autoregressive (VAR) modeling framework focussing on subset modeling strategies and the general-to-specific reduction approach automated by PcGets. We compare the chances of finding the true DGP, the accuracy of the implied impulse-responses, and the forecast performance. In Monte Carlo experiments, all procedures recover the DGP specification from large VARs with anticipated size and power close to commencing from the DGP itself. We find that subset strategies and PcGets are close competitors in many respects. Only the forecast comparison indicates a clear advantage of the PcGets algorithm.