Wins, losses, and why they don't matter

As you probably heard if you follow the sports world, Greg Maddux won his 300th game over the weekend, becoming the 22nd such pitcher in major league history. As baseball analysts have been tripping over themselves to say, he might very well be the last one to achieve the feat (with the possible exception of Maddux's former teammate Tom Glavine).

Along with Roger Clemens, Maddux is one of the top two players of the past 20 years. It's difficult to compare their careers. Clemens has more wins, but has also played for a few more years. Maddux has a lower career ERA, but he's spent his entire career in the DH-less National League. At their absolute best, Clemens is probably the better pitcher, but Maddux has been the more consistent of the two, considering his ERA hasn't been above 4.00 since 1987.

To be sure, 300 wins is quite an accomplishment. You don't win 300 without being one hell of a pitcher. But I'd argue that career win totals aren't a very good indicator of how good a pitcher is. The reason is quite simple.

Wins and losses do not reflect how good a pitcher is.

Let's review the definition of wins and losses for pitchers to see why this is the case. From MLB.com:

Wins
The starting pitcher is credited with a win if he has pitched at least five complete innings and his team not only is in the lead when he is replaced but remains in the lead the remainder of the game. When the starting pitcher is not credited with the win, the win shall be awarded to the relief pitcher judged by the scorer to have been the most effective, or who is the pitcher of record when his team assumes the lead and maintains it to the finish of the game.

Losses
The total number of losses charged to the pitcher. A pitcher is charged with the loss of the game if he is replaced when his team is behind in the score, or falls behind because of runs charged to him after he is replaced, and his team thereafter fails either to tie the score or gain the lead, regardless of how many innings he has pitched.

The short version: if you're pitching when lead in the game changes for the last time, you get a decision. If that change puts your team in the lead, you get the win. If the other team takes the lead, you get the loss.

This is pretty intuitive stuff.

The problem is that it depends on a whole lot more than how you pitch. In particular, wins and losses reflect how your team does around you. If your team scores, say, 10 runs, you've got a pretty good shot of winning. If, however, your team doesn't score at all, there's no chance you can get the win.

Imagine pitching shutout ball all season long, never giving up a single run. If your team also fails to score, your record will be 0-0. Your record reveals none of your pitching dominance. Yes, this example is extreme and absurd. But it reveals a fundamental flaw in the win-loss criteria. Without adequate run support, pitchers cannot win games.

The question of run support doesn't even touch upon the whole "pitcher of record" business.

A recent Phillies game provides the perfect example on why wins and losses aren't accurate reflections of this problem. On July 25, Phillies starter Eric Milton pitched a no-hitter through eight innings. Up 'til the ninth inning, Milton had been nearly perfect, only walking one. Thanks to some shoddy defense, Milton gave up two runs in the ninth, losing the no-hitter and allowing the game to be tied. Reliever Ryan Madson came in and got the final out of the top of the ninth inning. In the bottom of the ninth, the Phils scored a run and took the win.

Ryan Madson got the win.

Let's review. Eric Milton pitched 8 2/3 superb innings, giving up only three hits and two runs. Ryan Madson pitched to a single batter, getting Sammy Sosa to ground out. Yet, since Ryan Madson was pitching for the Phillies when they retook the lead, he got the win. As for Eric Milton's win-loss record; it was exactly the same as at the beginning of the game, in spite of the fact that he pitched his best game of the season.

To sum up, win-loss records simply do not reflect how a pitcher's performance. Instead, they reflect how a team does when a given player pitches. You could argue that, hey, wins and losses are what really matters, so this is a useful statistic. But you'd be holding a pitcher accountable for something that he has almost no control over.* You could also argue that certain pitchers bring out the best in their teammates, somehow inspiring them to score lots of runs. But it's been shown that, year-to-year, this doesn't happen. In the long run, run support is randomly distributed.

So where does this leave us with Greg Maddux? Well, I'd like to point out that he spent over half his career with the Atlanta Braves who, it just so happens, won their division in each of the years Maddux played with them. Now, a big part of the Braves' success can be attributed to their top-notch pitching staff. But you don't win that many division titles without scoring some runs. In other words, you can't look at Maddux's win-loss record in his Atlanta years without also considering the fact that the Braves scored a fair amount of runs around him. Would Maddux have won that many games if he were pitching for, say, the Phillies? In short, no.

This doesn't diminish the tremendous career that Maddux has had. Rather, it suggests that we should look at other statistics to see his greatness. Like, for example, his 2.93 career ERA.

I've rambled long enough. If nothing else, just take pitchers' W-L records with a giant grain of salt. Better yet, don't look at them at all.
*In the National League, unlike the American League, pitchers bat. Pitchers, to put it lightly, aren't renowned for their offensive prowess. So while NL pitchers do have the ability to "help themselves" offensively, it doesn't happen much.