by MARCPELLETIER » Wed Aug 23, 2006 4:10 pm
dave nailed it right on the nose.
4.5^1.82 divided by [(4.5)^1.82 + (4)^1.82] = 0.553
is greater than
5.5^1.82 divided by [(5.5)^1.82 + (5)^1.82] = 0.543
(the difference here is at most 2 wins, which may seem to be not so great, but 2 wins here and there, and it makes the difference between a championship and a playoff missing).
That said, this formula uses the same constant (1.82) in both environments, and I am far from certain this should be the case. If the constant is rather 2 in low scoring environments, and 1.75 in high scoring environments, then the difference is even more bigger---almost 3 wins, in the example selected).
Furthermore, there is something that these formulas don't capture, but that I have experience many times in my Petco teams. Because I rely on high-priced pitchers, I tend to have very cheap long relievers, and have them set at "mop-up". They rarely come to pitch, and when they come, they get badly beaten up. The result of all this is that I will win many games by low score margins, and these loose one big game, like 12-4, or something. As a consequence, I end up with runs for and against nearly equal, but still end up with 90 wins. In Coors team, it's the opposite: I always outscore my opponents, even though I might struggle to get 85 wins.