Empirically, human-flipped coins have about a 1% bias toward the same side they started on: https://www.ams.org/publicoutreach/math-history/hap7-fifty-o...