Predictions: Game vs. Final Standings

Why don't the individual game predictions match with the predicted end of season standings?

We get this question from time to time because it seems intuitive that if my team is the predicted winner in X number of individual games for the season then the end of season prediction should be that my team has X wins. But this is not always the case. The reason for this is that when we predict a winner of a game we saying that the chances of the team winning are greater than the chances of the team losing. However we can't discount that there is a chance that the team will lose. And over multiple games that chance of losing adds up.

Some examples may help illustrate this idea:

If the best team in the league played a 100 game season with the worst team in the league then the predicted winner of every individual game would be the best team, but the prediction for the end of the season records would probably be 90-10 for the best team and 10-90 for the worst team.

And the situation is even more exaggerated when teams are more closely matched. For example suppose we are predicting a team as the winner of all 16 of their individual games but we are only predicting that with 51% probability for each game. In this case although the chances are that the team would win any given game, the predicted final record for the team would be around 8-8.

So based on the discussion above, while the end-of-season prediction and game predictions seem inconsistent, they actually are not in conflict when considering the randomness of outcomes