I know it probably seems obvious, but maybe some simple explanation would be useful. Over the course of the season, the better teams score more relative to the worse teams. Generally speaking, the teams that score more are therefore likely to score first more, even if scoring first is basically a coin flip. Simply put, if you score a lot, you'll score a lot early. If you score only a little, you'll score only a little early. Therefore, assuming a normal distribution - and the hundreds of games played really does normalize these kinds of distributions - good teams score early more than bad teams. (They also score late more than bad teams. Because they just generally score more than bad teams.) So basing win probabilities on when teams score just isn't a good idea. Too much randomness for a single game, too much dependence on other factors for a whole season.