RealGM

I was looking at B-R and noticed the Heat are 22nd in SRS. Interestingly enough, they are only 0.25 points ahead of the Mavericks and the Celtics are rated a lot higher despite losing twice to the 21st team (Heat) and 26th ranked team (Knicks).

To be perfectly honest, the numbers I see look strange compared to my picture of how it should work, so I don't have an answer. But I'll say what I can:

This is the method I believe they're using to get SRS:

http://www.pro-football-reference.com/blog/?p=37

Obviously then, margin of victory matters, and more specifically: If someone else destroyed a team you barely beat, that's going to hurt you.

The Heat's interrelationships are complicated enough I don't have the energy to think through it all, but obviously Denver beat Dallas by more than Miami, so that plays into it.

With that said, take a look at Atlanta, New Jersey, and Washington. Jersey played the other two, and the other two have played no one else. By SRS, the sum of the 3 adds up to zero which seems exactly right to me. However, Jersey's at exactly zero, while Atlanta & Washington are opposites of each other, despite the fact that one game was a blowout, and the other game wasn't. To me this only makes sense if you're ignoring margin of victory, which SRS most certainly isn't supposed to do.

My only guess is that what we're seeing is that the thresholding b-r uses to say "good enough" in the iterations of their algorithm is really loose early in the season. Regardless, it's one more reason not to take SRS too seriously early on.

The whole thing is a bit iterative, so an easy example.

Say MIA avg'd 6 pts more than their opponents through three and their opponents averaged 8 fewer than their opponents. The initial rating is 6-8 = -2

The problem is that their opponents may have played teams who averaged fewer than theirs as well, say 5 fewer. The true opponent strength isn't -8 any longer, it's -8 - 5 = -13, so:

Miami's new SRS would be +6 - 13 = -7 . This is somewhat circular and doesn't seem simple at first glance, but it can be solved mathematically in a fairly straightforward manner.

The less simple rating systems that use complex algorhithms actually look at the result vs. each opponent on it's own merits.

SRS is basically a 30x30 matrix, while the complex stuff is basically a 30x30x82 model

The teams they've played's MOV in their non Mia teams vs their game

Dallas -12, Mia wins by 9. Gap: -3
Charlotte +1, Mia wins by 1. Gap: +2
Boston -10.5, Mia wins by 8. Gap: -2.5

So that adds up to scoring 3.5 less than the aggregate MOV of the other teams and thus, the negative score. 3.5/3 is -1.167, the reason Mia's SRS must be -2.12 is adjusting for schedule... I'm guessing since MIA has played a below average schedule so far, they get that number downgraded

Basically SRS is reading DAL and BOS as 7-75 caliber teams, so beating them by 9 and 8 points and then squeaking out a win against an average team is a bad result by the league's average...