Variance is the statistical noise between what should happen on average and what actually happens in any finite sample. In betting it is brutal and unavoidable: a genuine 55%-probability bet loses nearly half the time; a portfolio with a real +5% edge can sit in loss after two hundred bets; a coin-flipping tout can post a glorious month.
Respecting variance changes behaviour. It sizes stakes (see bankroll management), it sets evaluation windows (hundreds of bets, not dozens), and it inoculates against both despair and euphoria. It is also why transparent full records matter more than hot streaks — over a long enough sample variance washes out and only edge, positive or negative, remains.
500 independent bets at 50% / odds 2.10 (a real +5% edge): about one such bettor in six is still at a loss after all 500 bets — pure variance, no mistake made.