Guarding The Lines: Dr... (04/27)
the archives are now free.
12 comments have been left for this article. BP Comment Quick Links
| |
|
|
April 27, 2017 Flu-Like SymptomsBlowin’ in the Wins (Part 2)
Last week, I looked at teams that are charged with two or more blown saves in a game. The conclusion was that the number of games with multiple blown saves is increasing, and that increase is largely due to more relievers per game (currently averaging over three per team), creating more opportunities for blown saves. (There, I just spared you reading 1,300 words).
Outstanding commenter Llarry responded:
And the next question is whether there is any consistency to the teams that multiple blown saves are *against*. I notice that two of the ones so far this year were against Houston. Does that end up meaning anything, or is it just random?
Good question, Llarry!
I started to look at games with multiple blown saves, but I concluded that sample’s too small. As I noted in the article last week, no team has ever had more than five games with two or more blown saves in a season, and that’s only happened twice.
So I instead considered blown saves more broadly. As I discussed, blown saves are not an official statistic, so data sources apparently calculate them differently; there are several discrepancies among Baseball Prospectus, Baseball-Reference, FanGraphs, and MLB.com. Since I’d gone through the labor of creating a spreadsheet with every single blown save (as defined by Baseball-Reference) since 1969, I decided to put it to work.
I looked at all blown saves from 1998 to 2016. I chose that time period for three reasons. First, it represents the entirety of the 30-team era. Second, it’s a pretty robust dataset: 570 team seasons, encompassing 11,234 blown saves. Third, since blown saves are correlated with relievers per game, there were a lot of relievers in each of those seasons, ranging from 2.46 per team per game in 1998 to 3.15 per game last year.
Returning to Llarry’s question, what kind of teams allowed a lot of blown saves? And what kind of teams were beneficiaries of a lot of blown saves?
Before looking at the overall results, let’s look at outliers. Sometimes teams at the extremes are emblematic of a trend as a whole.
Most blown saves in a season: 2004 Colorado Rockies, 34. I am shocked, shocked that a Rockies team would top the list. Especially one highlighted in the report last week as one of the only two teams ever to have five games with two or more blown saves in a season. They had 24 other blown save games, too. What distinguished this team?
Fewest blown saves in a season: 2003 Los Angeles Dodgers, 8. OK, so scratch the “blown saves as a product of the steroid era” argument.
Beneficiary of most blown saves in a season: 2004 Los Angeles Dodgers, 35. Of course, they played in the same division as the 34-blown-saves-allowed Rockies. And nine of the Rockies’ blown saves that year came against the Dodgers. But the Dodgers broke the hearts of their other divisional rivals as well: They received five blown saves against the Diamondbacks and three each against the Padres and Giants.
Beneficiary of the fewest blown saves in a season: 2010 Los Angeles Angels, 5. The team finished just 82-80, but at least their top draft pick the year before hit .341/.428/.490 in A-ball at age 18!
There wasn’t any real commonality among these teams. So to find the key markers of overall success at blown save avoidance/accumulation, I considered three drivers:
Take teams that allow a lot of blown saves. We’d expect those teams to be poor teams (negative correlation to won-lost record, both overall and in one-run games) and allow a lot of runs (positive correlation to runs allowed). It works out that way. Kind of. The correlation of blown saves with team winning percentage is -0.26. The correlation with winning percentage in one-run games is -0.20. The signs are right, but the value of those correlations indicates a very weak relationship. The correlation with runs allowed, +0.34, is barely better.
We’d expect teams that are beneficiaries of blown saves to be teams that do well overall, particularly in close games, and score a lot of runs (in order to come back from save opportunity-creating deficits). But that’s not borne out at all. The correlation of forcing blown saves with won-lost record is +0.06. With record in one-run games, it’s +0.13. And with runs scored, it’s +0.17. None of those rise to the level of significance.
And the best evidence that blown saves are pretty random? I calculated the correlation between blown saves and reliever ERA, FIP, and DRA. Bad bullpens should allow a lot of blown saves, right? The correlation of blown saves to bullpen ERA is 0.00. The correlation to bullpen FIP is -0.01. The correlation to bullpen DRA is -0.03. I would have thought, you would have thought, the teams that allow a lot of blown saves are teams with a lousy bullpen. And you and I would both be wrong.
What can we draw from this? Well, not a lot. Avoiding blown saves, or accumulating them, is either fairly random (as is a team’s record in one-run games) or it’s driven by something I didn’t measure.
In the meantime, Tuesday Joe Biagini blew a save in the seventh inning and Roberto Osuna blew another in the ninth inning during Toronto’s 6-5 win over the Cardinals. That’s the eighth multiple-blown-save game this season. The all-time record is 39 in 2006. Granted, multiple blown saves don’t have the same cachet as, say, Eric Thames homers, but we could be witnessing a record season!
Rob Mains is an author of Baseball Prospectus. Follow @Cran_Boy
|
The Stash List: Fourth Edition
Thanks for looking at this. I wouldn't have expected it to be nearly so random, but now we know.
I'm not as surprised at the results of teams blowing saves, as much as that there's not at least a *little* more in the teams that benefit.
As for the reliever stats comparison, I guess what it shows us is that bad bullpens do most of their damage when the team is tied or already losing. It's not so much giving away a game that the offense and starter have set up as winnable, as making it so that no late-inning offensive heroics will be enough heroics.
Thanks again for the suggestion!
Good point about reliever stats. That, and a couple other things that came to mind as I was writing this (eg, should I be looking at run scoring/prevention in the seventh, eighth, and ninth innings as separate predictors?), could warrant further research, but I was concerned with, as a junior high gym teacher put it, beating a dead horse to death.