The All-Time Leader in Lost Wins

It’s amazing the number of times I go to the Internet to find the exact phrasing of some quote I want to use, only to discover that I’m completely wrong about who said it. The management guru Peter Drucker is famous for the line: “What gets measured gets managed,” which he apparently never said and which was better stated anyway by a journalist (Simon Caulkin) summarizing a 1956 article by VF Ridgway as:

“What gets measured gets managed — even when it’s pointless to measure and manage it, and even if it harms the purpose of the organisation to do so”.

Those of you who follow my opinions (all none of you) will remember that I think the Win stat is baseball is near-useless.  But as long as we have it, people are going to manage to it, so it should be as good a measure as we can make it.  In this note I propose one change that will improve it in the current climate, is easy to re-calculate retrospectively and makes very few changes in those retrospective adjustments: elimination of the innings requirement for starting pitchers.

The 5 inning requirement is an outlier for a lot of reasons. It is the only stat I can think of (other than perhaps the qualifying minimums for seasonal titles) that it based on an expectation, that starting pitchers ought to go 5 innings before we really think of them as starters. But even if that was the expectation once, it is less so now and has been falling dramatically as we shall see. But if we were to go on expectations, why not a one inning minimum for anyone to get a win? Or at least one pitch? And when the minimum isn’t met, we substitute a purely subjective call for the win. Why not just allow a subjective call for every game? Case 1: starting pitcher goes 4 2/3, giving up no runs, leaving with a 22-0 lead. Every subsequent pitcher gives up 3 runs and the team wins 22-21. Case 2: starting pitcher goes 8 giving up 20 runs, leaving the bases loaded and the team up 20-19 and nobody out. The next guy comes in and strikes out all three batter he faces. You want to give the win to the starter? Both of these cases are highly unrealistic, of course, but I can think of lots of times that it would make more sense for the Official Scorer to assign a win, just like he or she assigns hits and errors. But assuming we’re not going to do that, let’s just eliminate the 5 inning rule for starters.

In the Retrosheet event database (1920-2019) there are 328,844 starting pitchers in games.  Of these, only 10,259 (about 3%) left with a lead before completing 5 innings.  And of those, only 3,361 (which I call the ‘lost wins’) times did his team not subsequently lose the lead in the game. (In some of the other games, the starting pitcher actually got the loss, since runners left on base when he left reversed the score and the lead never changed hands again.)

As one might expect, this effect was fairly minor for a long time but has become much more common lately.  Here is a graph:

From 1920-2014, the lost win rate has been pretty constant at about 1% of starts.  There was a bump in the mid-1950’s for some reason. (I’m going to assume it was not people taking note of VF Ridgway’s article.)  But the number has begun to take off.  The three highest rates in history were 2017-2019, and each of those years was higher than the one before.  The 2019 rate as triple the historic average, at around 3% of all starts.  (I don’t have the 2020 data, but I’m not sure I’d use it even if I had it.)

No one thinks that wins for relievers are meaningful.  Indeed, reliever wins are very often the result of bad pitching rather than good pitching, when a reliever blows a save but his team takes the lead in the next inning. Many other reliever wins are pure chance; whoever is in when the lead changes hands gets the win no matter how many pitchers pitched as well or better.  The Braves’ 1-0 13 inning over the Reds in the first game of the playoffs saw the Braves use 8 pitchers, not one of whom yielded a run.  Minter got the win because he happened to be the pitcher when the Braves finally scored.  His performance was not readily distinguishable from anyone else’s that day.

But we persist in thinking that starting pitcher’s wins are important.  But especially in today’s game, they mostly reflect (a) an ability of the pitcher’s team to take a lead early, which is correlated with the pitcher’s own prowess, but in a highly context-dependent way; and (b) the ability to hold a lead, which has little to do with the starting pitcher at all, as Jacob deGrom can readily attest. 

It is true that starting pitcher wins are positively correlated overall with most other measures of pitcher’s performance, but we have those other measures of performance.  It is certainly unclear that wins have any independent value above that given by ERA, ERA+, FIP, WHIP, WAR and the like.  And while the latter 4 are newfangled sabermetric measures, ERA has been around forever.  It is indeed difficult to see anything provided by wins and losses for a starting pitcher that isn’t provided by the traditional stats of ERA, starts and innings pitched as a starter.  (That last one is sometimes a little difficult to come by for pitchers who both start and relieve, to be fair.  But adding wins and losses doesn’t help, since that generally includes wins and losses in both roles as well.)

Unsurprisingly, the pitcher with the most lost wins is Ryne Stanek.  In 2018-2019, before being traded to the Marlins,  Stanek made 56 starts and pitched a total of 83 innings in those starts.  His record was 0-3, but he left 9 times in which the Rays had a lead they never relinquished.  Frankly, a career record of 9-3 as a starter makes a lot more sense for Stanek than 0-3.  He had a ERA of 2.71, a WHIP of 1.06 and a K/BB ratio of 3.29 in those starts.  But in any case our understanding of Ryne Stanek, or indeed any pitcher, will not be hopelessly corrupted by handing out a few wins.

In the Retrosheet Era through 2019,  there are only 65 pitchers who would gain as many as 5 wins in their careers.  Here is the list:

This list is a mix of current players like Stanek, some starter/relievers who made a number of spot starts like Mike Bielecki,  and a few guys with really long careers like Jim Kaat, Lefty Grove and Andy Pettite.  I’m not sure there is much to be learned from this, and these are the top ‘lost win’ guys of all time.  Maybe Tim Hudson’s HOF case would be augmented slightly by making him 227-133 rather than 222-133, but I doubt it.

One argument against my proposal is that it will allow teams who use a bunch of openers to vulture wins for them.  For openers who only go one or two innings, though, the effect is pretty minor.  Your team has to take the lead in those first two innings and hold it for the remaining seven.  It certainly happens, but it is much harder to do than it appears.  Take Tampa Bay in 2019.  They had 73 games in which the starter did not go 5 innings, by far the most in MLB.  Most of those were opener games.  The opener left the game ahead in only 21 of those games, and in only 10 of those games did Tampa Bay win the game without ceding the lead. (6 of those are 2/3rds of Stanek’s lost wins.)

Against the objection lies a counter-objection: teams who keep a pitcher in the game only for the win.  The third game of this year’s NLCS, in which Julio Urías went 5 innings with a 15 run lead, even though, as one of LA’s most effective pitchers, his contributions to a potential LA NLCS would almost surely be higher in higher-leverage situations; he wasn’t able to show that until Game 7. (Admittedly, it’s hard to do much better than he did, but they might have won in fewer games if he’d been available earlier.) But to pull him would be to deny him a win, and this slight might well overwhelm the ultimate best interests of the team. I’m not saying Dave Roberts was wrong to let Urías pitch 5 innings — perhaps the psychological damage by denying him the win exceeded the increase in the Dodgers’ chances. But were the 5 inning rule not in place, and Urías would not have to sacrifice a win from an earlier departure, the costs of pulling him would have been lower.

We have no similar qualms on the other end of the game. A guy who pitches 4 2/3 innings of shutout ball with a one run lead doesn’t get a save if they bring in somebody else to get the last out. We don’t condition consecutive game hitting streaks on how many at-bats you have. Even the seasonal minimum rules have exceptions: you can qualify for the batting title if you’re short the required number of at-bats if assuming an out for every one of those at-bats would still get you the title.

Since we already regard wins by relievers as barely relevant, correcting the record by subtracting wins from whomever got them seems to be no problem.  Overall, I see no reason at all why this rule should be retained, why the number of wins for starting pitchers shouldn’t be augmented retrospectively to make this change.  And the best part: Chip Caray will have to think of something useful to say about a pitcher’s performance in the 3rd-5th innings.