Expected Win / Loss Calculation Errors

Maybe it's simple physics but ... one train from New York to Chicago and the other from Chicago to New York leaving at the same (locally adjusted) time and at the same speed will reach their destination near the same time.

So, how can two teams with:

Nearly identical W/L records (46-26 / 45-27)
Identical run differential (+107)

Have two completely different Exp. W/L results?

Team 1: 46-26 (Exp W/L 43-29)
Team 2: 45-27 (Exp W/L 46-26)

Let's take two examples:

1/ the team has won 72 straight shutouts, outscoring their opponents 107-0. The expected W-L record is 72-0.

2/ the team has outscored their opponents 1070-963 (differential of 107) in 72 games. The expected W-L record is 39-33.

It's not just differential, but how you get there.

Bill

Team 1:
W/L: 46-26
RS:580 (RS per game 7.1)
RA:473 (RA per game 5.8)
Diff: +107
Exp W/L: 43-29

Team 2:
W/L: 45-27
RS:449 (RS per game 5.5)
RA:342 (RA per game 4.2)
Diff: +107
Exp W/L: 46-26

Team 1 has outscored opponent by 23%
Team 2 has outscored opponent by 31%

Thus I would expect team 2 to have a better record.

Also, remember reading an article by Bill James where he talked about the expected wins formula tending to be less accurate as scoring increased. He spent pages explaining why but concluded that increased scoring during the steroid era required the exponent be adjusted for accuracy. The original formula simply squared the values while the new formula uses an exponent of 1.83

Think of it intuitively. To get from the runs scored by team 2 to the runs scored by team 1 you have to add 131 to runs scored and runs allowed. That 131 to 131 ratio would be a .500 record. That cannot do anything other than lower the expected winning percentage as runs increase.

Expected wins like many saber stats can be an interesting tools. Just don't get it confused with reality.