We analyze a repeated first-price auction in which the types of the players are determined before the first round. It is proved that if every player is using either a belief-based learning scheme with bounded recall or a generalized fictitious play learning scheme, then for sufficiently large...