Suppose that you have a league of 30 teams, each of which plays 5 games against each opponent. That’s 145 games for each team. You have also been told that every team is identical in talent. That is, our expectation is that every team is a .500 team. If you insist for something stricter: each team has the identical players, who are somehow able to co-exist at the same time. I will ask you two questions:
1. What is the over/under forecast for each team (or similarly, what is the average forecast for each team)? The answer is going to be 72.5 wins (or .500), for each of the thirty teams.
2. How many wins will the league-leader have after 145 games? The answer is not so obvious, but what is absolutely true is that it will be at least 73 wins. And, it’s going to be alot more than that. I ran a quick simulation of 100 seasons, and the answer was just over 85.5 wins (.591).
That is, if every team was equal, we’d STILL expect some team to win, on average, about 13 more games than their true talent would suggest. And this is COMPLETELY explained by random variation.
Why do I bring this up? Well, when someone, be it Vegas, or Clay Davenport, suggests that the most talented team is going to have an over/under of 91 or 92 wins, this does NOT mean that they expect the league-leader to have 91 or 92 wins. It’s two different questions, exactly the same two questions that I have.
Whereas you can estimate that the most talented team will have an over/under of 91 or 92 wins, you should therefore estimate that the league leading team (whoever that happens to be) will likely average some 10 wins more than that.