Saturday, August 3, 2013

#KillTheWin, Part 2

I wrote a post in support of MLB Network's Brian Kenny's (@MrBrianKenny) #KillTheWin movement, his effort to place wins in their proper perspective, i.e., no longer discussed as a serious baseball statistic. I strongly suggest reading that post before continuing with this one since it sets out my definitions and rationale. This post is predicated on this tweet:

Now I know how Moses felt at the burning bush--that's baseball royalty asking ME a question. My knees have just about stopped shaking since that tweet was sent 3 days ago.

I did the research, looking for games in which the pitcher had a game ERA of 6.00 or higher and got a win--the exact opposite of what I studied in the first post, pitchers with Game Scores of 60 or above who got either a loss or no decision. It took me about half an hour using the Baseball-Reference Play Index feature ($36--find it, pay it, learn) to do this and find out that the type of game that Mr. Tango describes in this tweet have occurred...around 10,000 times since 1916. There have been around 165,000 major league games since 1916, so cheap wins (one term is "vulture wins," and thanks to Cubs TV play-by-play announcer @LenKasper for setting me straight on that) occur around .6% of the time. Seems unfair, no? This chart breaks it down by innings pitched:

We can categorically state that relievers had wins in situations like this 3,240 times (anything less than 5 IP) and I'd be speculating after that for reasons that will bore almost anyone. I'll break this data down by starters and relievers.

Starters in situations like this are ones who used crafty stuff and guile to notch a 14-12 victory. The games of Friday, August 2nd included three like this--the Diamondbacks Randall Delgado got the win over Boston by only giving up 4 runs in 6.0 IP, Kris Medlen had the same numbers as the Braves beat the Phillies and Chris Tillman gave up 6 runs in 5.1 IP for the Orioles for a game ERA of 10.12--and the win. This chart breaks down the runs allowed for starters credited with wins:

Almost 1200 games where the pitcher gave up 7 or more runs and they got the win--contrast that with the number of times starters have given up 2 or fewer runs and got the loss:
Doesn't seem fair, does it? The pitchers in these situations did everything asked of them and their teams didn't follow through either offensively or defensively, and yet it's the pitcher's fault.

This chart shows starting pitchers with the most cheap wins since 1950, as well as the number of unfair losses or no decisions:

I tweeted out a version of this yesterday and had an epiphany today. To illustrate, Tim Wakefield had 25 wins in which he had a game ERA of 6.00 or higher, as well as 19 losses and 24 no-decisions in games in which he had a Game Score of 60 or higher. In other words, he had 25 cheap wins and was robbed 43 times. The last two columns demonstrate that the win (AND the loss) did an ineffective job of demonstrating Wakefield's success--he received the decision in 380 games (out of 627 total appearances) and in 68 of them (17.9%) was tagged with a win, loss or no-decision that in no way reflected his actual pitching performance. That percentage holds fairly steady between 15-20%, suggesting that any measurement device that is inaccurate 1 in 5 times can be described as a MIS-measurement.

Relievers are viewed differently. You hear it all the time:
"Yeah, well, you know, wins and ERA don't matter as much for closers"
Yeah, well, you know, if Mariano Rivera enters a game with the Yankees leading 4-2 and gives up a run, he walks out with a Save--and a game ERA of 9.00. Did he accomplish what he was supposed to do--of course, he ended the game, but that Save will look exactly the same as one in which he comes in, throws three cutters for three ground outs and hits the showers. This chart shows the most wins in which a reliever had an ERA of 6.00 or higher:

I'll used the aforementioned Mariano Rivera to illustrate--he had 16 wins in 1,084 relief appearances (he had 10 starts in 1995--starts are NOT included in this number) in which he had an game ERA of 6.00 or higher, out of 158 decisions in his career. 10.1% of his decisions were cheap wins.

A solid argument can be made that ANY decision by a closer is bad, since that suggests the lead disappeared while they were on the mound. As such, let's shift to the measure by which closers are evaluated and see how many SAVES pitchers got with a game ERA of 6.00 or higher? 

3,819--out of 63,422 saves in baseball history--6%. Shoot, let's get rid of that too. Here are the leaders in that category:

Trevor Hoffman had 47 saves in which he had a game ERA of 6.00 or higher, 7.8% of his 601 career saves. This is the end result of statistics that purport to measure pitching effectiveness but instead reflect GAME result. To view all wins (and losses and saves as well) as equal doesn't just blur the line between useful information and disinformation, it obliterates it.


Call Chicago's 670 The Score (312-644-6767) between 1-6 pm and chances are producer Matt Abbatacola (@mattabattacola) will answer the phone. In fact, call on Mondays after Bears games and ask why they didn't line up in the shotgun more often for a thoughtful and erudite discussion. On Thursday #KillTheWin came up briefly on the show, and Matt correctly pointed out it will be impossible to truly eliminate it, so why not focus instead on pointing out its shortcomings and fleshing out pitching performance with other more descriptive measures. We don't live in a binary world--if X is a problem, very rarely is Y the sole answer. I'm more amused than anything else with the #KillTheWin movement, but behind it is a very important issue--a number that people use as a significant reference point and that agents use to justify contracts IS NOT an accurate reflection of game performance around 15-20% of the time. 

The obvious retort is that it IS accurate around 80-85%, which isn't bad--not much in the world with any complexity can be described with one number with that degree of accuracy. But it can be improved and augmented using the wealth of data that's out there on Baseball-Reference, FanGraphs and Baseball Prospectus and the advanced thinking from icons like Bill James, Voros McCracken, the Tom Tango collective and others far too numerous for me to mention. Leisure suits, "According to Jim" and the Game-Winning RBI all received their well-deserved demise. That won't be the case with the win--but we can improve the discussion. 

I'm in.


  1. Awesome post. I would like to point out that its either one of the fox sports channels or the CBS sports network channel, or maybe both for that matter, have been using GW RBI on their scroll this year. Really annoying.

  2. Seriously? I remember seeing it on my Topps baseball cards from the late '80s and thought it went away. Makes me sad in an era of excellent data availability to rely on something so trite.