The problem with W/L record is that is so widely used yet can so poorly describe performance/worth. Take these two this year:
|
IP |
W |
L |
WHIP |
ERA |
K/BB |
OPS |
RS |
Cliff Lee |
211.0 |
6 |
9 |
1.11 |
3.16 |
7.39 |
0.690 |
3.60 |
Lance Lynn |
176.0 |
18 |
7 |
1.32 |
3.78 |
2.81 |
0.728 |
5.90 |
(RS - Run Support) Anything that can be that wrong shouldn't be relied upon. There are so many stats that do a better job, W/L assigned to a pitcher should be unnecessary. It's just convention. There were 88 qualified starters in 2012, if you group them into percentiles by Run Support, the groups look like this:
PLAYER |
IP |
W |
L |
WHIP |
ERA |
K/BB |
BABIP |
OPS |
RS |
Top 25% |
189.3 |
15 |
8 |
1.26 |
3.78 |
2.73 |
0.280 |
0.702 |
5.15 |
25% - 50% |
197.9 |
14 |
10 |
1.24 |
3.67 |
2.80 |
0.278 |
0.708 |
4.57 |
50%-75% |
190.6 |
12 |
11 |
1.29 |
4.21 |
2.61 |
0.280 |
0.738 |
4.15 |
Bottom 25% |
196.1 |
11 |
12 |
1.25 |
3.69 |
2.79 |
0.281 |
0.699 |
3.65 |
There's very little to explain the W/L difference between those groups, other than run support. That 50%-75% group would have to be considered the worst but the .5 higher run support is all they needed to out-win the Bottom 25% group, which has a case for the "best" group. Those pitchers in the bottom group didn't fail as much as their offenses did. Kershaw dropped from 21-5 last year to 14-9 this year. I don't think he pitched meaningfully worse or differently - he just lost about .7 runs of support per game.
Good pitching should lead to more wins, even at the individual starter level, but we all know more wins don't mean best pitching. Assigning an entire win or loss to one player per game in a team sport just can't end up being descriptive in a meaningful way often enough to be really useful, can it? Even if you like W/L record, how many other individual stats would you say do a better job of describing pitcher performance by themselves? ERA, ERA+ or ERA-, etc? WHIP? OPS against? K/BB? FIP, xFIP, etc? WAR? Quality Start %? If you like/know a couple of those, W/L probably doesn't add anything except some evidence for a good guess about the pitcher's run support.