Sunday, July 28, 2013

#killthewin

I love researching questions but don't have the most imaginative mind in the world. That's why when I see intriguing subjects arise on Twitter, I'll take that kernel and flesh it out. It started with this tweet from MLB Network's Brian Kenny (@MrBrianKenny):


I was intrigued but didn't really understand the point and started figuring it out over the next couple of days. I've never been a huge fans of wins as a pitching statistic, especially in the modern era with teams routinely using four or more pitchers in a game, since it was established in a different time when it was much more likely the pitcher who started the game finished it. Kenny's point became much clearer once he tweeted out that pitchers with Game Scores of 60 were racking up no-decisions or losses in droves. It's bad enough to pin a loss on a pitcher that performs well, or in other words, held up HIS end of the bargain, but it's downright injustice to reward an effective performance with nothing. This became my research target--how pitchers with a Game Score of 60 or higher perform over time.
To begin, I'll use a screen grab from baseball-reference.com to define Game Score:
It's relatively straightforward and helps differentiate between the pitcher who won a 14-12 pitching duel and someone who was effective in keeping runners off the bases. Since Kenny set 60 as his threshold, I retrieved all games from 1950 through Friday, July 26th in which the starting pitcher (and he's the only one that gets a Game Score, by the way) had a Game Score of 60 or greater. There were a total of 80,363 games, a pretty robust sample size. To give a sense of how good a Game Score of 60 is, here's the Game Score breakdown for ALL games from 1950-2013:
Not too often you see a distribution THAT bell-shaped, and Games Scores of 60 or more are 32.6% of games pitched. This is the set-up--here's where I begin to prove Kenny's point.













This graph shows the winning percent of both the pitcher and the team in those games with Game Scores greater than 60:

There's a pretty serious divergence between team winning percent and pitcher winning percent, almost 100 points. I don't trust that gap around 1950-1952 since Game Score calculations require play-by-play data and only about 80% of those games have complete play-by-play vs. box scores, but as the data became more complete in the 1960s that constant gap continues. Pitchers pitched well enough that their teams won three out of four of these games but the pitcher was only credited with the win around  65%--and that number is trending down.




There are perfectly good reasons for this:
1. Pitchers simply don't pitch as many innings as they used to--everyone knows this, but why it matters is that modern pitchers aren't given the opportunity to let the offense help them out as often as in the past. In the modern era, if a pitcher is losing 1-0, and especially in the NL, coming to bat in the bottom of the 7th, he's done and won't be around to potentially benefit from a score in the 8th.
2. Teams lose games 2-1, 1-0 and the like all the time. That's Kenny's ultimate point--it's disingenuous to look at a pitcher's win-loss record and make snap judgments, like saying "Well, he was only 13-12, how good could he have been?" In Felix Hernandez's case in 2010, good enough to win a Cy Young. He had an OUTSTANDING year, one that was totally and completely not reflected in his win total. I don't claim to be an expert, but I was stunned when he won the award, and it gives me hope that going forward the full picture of a pitcher is reviewed without giving undue weight to one outmoded stat.

This chart shows the incidence of 60+ Game Score games over time:

The 1960s were a mini-Dead Ball Era, and it's up to people smarter than me to determine if it was truly ascendant pitching, a temporary decline in hitting or other factors (probably all three) that caused that to occur, but starting around 1970 a long-term decline in these types of games began, from around 35% of games down to the nadir of just over 25% around 2000. And look what happened around 2007--again, it's up to people smarter than me to explain, but pitching is back and back with a vengeance. As I listen to Chicago's 670 The Score I've been hearing quite a bit lately from their baseball experts (folks like Cubs broadcaster Len Kasper, Cubs beat reporter for the Sun-Times Gordon Wittenmyer and Ken Rosenthal of Fox Sports) that good hitting is becoming much more difficult to find. While I don't doubt that, it's also very possible that good pitching is overwhelming the hitters.

I don't know what the replacement to the win is--it made sense back in the days when men were men and Old Hoss Radbourn went 59-12 for the Providence Grays, but that was in 1884 and nobody cares today. One of the blessings of baseball is its rich statistical heritage in that we can go back to 1871 and see how well players hit, pitched and fielded. The game has changed tactically for numerous reasons, but the underpinnings remain the same--hit the ball to drive in runners. Unfortunately, one of the shortcomings of that history is a fascination with numbers that no longer are relevant. Baseball stats have been invented almost since it's inception, and even in modern times they continue to be--the save by Jerome Holtzman, the blink-and-you-missed it Game Winning RBI in the 1980s, the hold and something I just saw this morning on FanGraphs--the shutdown and meltdown. The newer professional sports don't have these issues because they were organized 50 to 70 years after baseball and don't have the baggage of useless numbers that don't accurately reflect ability or accomplishment.

No other sport (except hockey, I think) measures wins for individual players--even in football, while pro-football-reference.com does have a quarterback's win-loss record it's rarely mentioned--broadcasters and sportswriters focus on QB rating, completion percent and TDs, because those are in the quarterback's control. A pitcher has no control over two-thirds of the outcome, his team's offense and their fielding, and while efforts to measure that using Fielding Independent Pitching are made they're not widely discussed outside of sabermetric circles. 

This chart will have to become better known as the game moves forward--it's the career wins for pitchers whose careers began in 1990 or later:
  
When Early Wynn won his 300th (and last) game in 1963, he stated he was going to be the last pitcher to win 300 games--other than the 10 that followed him, he was essentially correct, and he said that 50 YEARS AGO. Forget 300 wins as some kind of Hall of Fame benchmark (and there's 38 with less than that, not counting closers), getting to 200 is going to be tough enough. Mike Mussina is up for consideration this year and there will be no shortage of BBWAA voters who will come up with all sorts of bizarre reasons (didn't reach 300, only won 20 games once, etc.) to not elect him, and he's quite likely to be the leader of this generation of  pitchers in wins by a healthy margin. How can any sane voter keep Pedro Martinez out? The only two current pitchers that can even think about reaching 250 wins are CC Sabathia and Mark Buehrle, and Buehrle is a stretch at best.

This table shows the pitchers with the most games with Game Scores of 60 or more:

The proper way to read this table is to see that Nolan Ryan had a total of 428 games with Game Scores of 60 or more, and in 154 of those, he either had a loss or no decision. When people look at Nolan Ryan's record, they see a winning percentage of .526 (324-292) and just sniff "Overrated." It's more like "played on bad teams"--he did what he was supposed to do to help his team win but anemic offenses let him down. This type of analysis also places Don Sutton, Phil Niekro, Bert Blyleven, Tommy John and Jim Kaat in the proper context--other than Sutton, the others pitched for underwhelming teams and have the win-loss records to reflect it. For example, any objective look at Blyleven's stats OTHER than his wins depicts one of the most dominant pitchers in the past 50 years or so, but since he had the bad fortune to pitch for a lot of bad teams, it took him 14 tries to make it into the Hall of Fame.

Take a moment to consider those Greg Maddux numbers--as it was, he had a career record of 355-227, and accomplished that in a hitter's era and pitched well enough to win an additional 105 times and didn't.

I wish I had a solution, but I don't. None of the traditional pitching counting stats resonate (strikeouts, ERA), WAR is still new enough to generate blank stares, but tweets stating that Topps is including it in baseball cards gives me hope. WAR is how I evaluate pitchers because it's the most objective number out there that ties everything in a pitcher's control together. Different computational methods are constantly changing, but the number is far less important than the context--for example, since 1950, the highest WAR for pitchers (per baseball-reference):
4. Randy Johnson, 104.1
3. Greg Maddux, 104.8
2. Tom Seaver, 106.1
1. Roger Clemens, 139.2
I'll save my reasons why I have no problem with Clemens (or Barry Bonds) in the Hall of Fame for another day, but that's a healthy difference between #2 and #1, about 33%. That's a number that helps me understand not only whether a player was good, but by what magnitude--it's a keeper. In other words, I don't care that Clemens had a career WAR of 139.2, only that it was significantly better than the next-best player. It helps me understand how dominant he was without having to rely on an old anachronism that if all baseball statistics were thrown out and started anew wouldn't make the first cut.

Brian Kenny, for this alone you're my hero, but I fully support #killthewin--wins are as much of a dinosaur as a $20 million DH. It's definitely tilting at windmills and will probably never go away, but the focus can change toward measures that more effectively evaluate pitcher performance. Baseball CAN change--games are played at night, teams are integrated, artificial turf came and went, it just takes its sweet time. Kill the win (and the loss, and I'm not so sure about the save either) and our ability to accurately analyze pitching effectiveness will increase dramatically as a result. Hats off to you, Brian Kenny for getting this started--now all you need to do is enlist Ken Harrelson to help.

2 comments :

  1. I also believe in saber metrics, but the level that you have presented them has really opened my eyes more to it. I also have a blog on saber metrics but it only has about five posts. If you want a link you can just say so. But yes Brian Kenny and Harold Reynolds on MLB Now is interesting.

    Amazing work!
    ~Carter

    ReplyDelete
  2. Nice job. You may be interested to know that Bill James, who invented Game Score, recently posted an article that showed the historic winning percentage for pitchers based on their Game Score. Unfortunately, it's subscription only at Bill James Online.

    ReplyDelete