Results of a blind test of photometric redshift predictions against spectroscopic galaxy redshifts obtained in the Hubble Deep Field with the Keck Telescope are presented. The best photometric redshift schemes predict spectroscopic redshifts with a redshift accuracy of Δ z< 0.1 for more than 68% of sources and with Δ z< 0.3 for 100%, when single-feature spectroscopic redshifts are removed from consideration. This test shows that photometric redshift schemes work well, at least when the photometric data are of high quality and when the sources are at moderate redshifts.