Shutdowns: A Replacement for the Save Stat

The folks at Fangraphs have developed, and are advocating the use of, stats that seem to me to represent a major improvement, in evaluation accuracy, over the “save” and the “blown save” as methods for ranking relief pitcher performance.   The Fangraphs concept of the “Shutdown”, as a replacement for the save, is very simple in concept if you  understand the idea of “Win Probability Added” (WPA), which itself has been much discussed over the years at the baseball-reference blog that spun off into High Heat Stats.

WPA measures the degree to which the outcome of each plate appearance changes the standard probability of the hitter’s or pitcher’s team winning the game, compared to where that probability stood immediately before the plate appearance.  The resulting change in probability, positive or negative, is then assigned to the pitcher and hitter who participated in that plate appearance.  For example, a  PA with the bases loaded, one out, tie score in the late innings that results in a double play and no runs scored will mean a very large dollop of positive WPA for the pitcher and a big WPA deduction for the hitter. The results of this calculation for every plate appearances in a game are then  added up, with each pitcher and hitter in the game receving a WPA number for the game that is the sum of all his plate appearances WPAs.  Both b-ref and fangraphs calculate WPAs for every pitcher and hitter for every game (b-ref’s WPAs go back to 1948).

A “Shutdown” for a relief pitcher is defined at Fangraphs as any game by a relief pitcher in which his WPA for the game is plus 6% (+.06) or more.  This standard has a number of advantages over the traditional save statistic.  First, it eliminates credit for the very easiest saves.  For example, in the second game of the 2012 season, over in Japan, Grant Balfour picked up a save coming  in for the A’s at the start of the bottom of ninth inning with Oakland holding a three-run lead.  Teams win about 97% of games that they lead at the start of the bottom of the ninth, so Balfour closing out that game contributed only about 3% in Win Probability Added.  So although Balfour gets a save, he does not get a Shutdown for that game becasue his WPA was less than the 6% stnadard for a Shutdown.  Only a save that had a meaningful amount of risk to it qualifies as a Shutdown.  In addition, any sort of relief appearance, no matter when it happens in the game, can qualify as a Shutdown, if the reliever pitches successfully under enough pressure.  Indeed, multiple Shutdowns can be awarded in a single game if multiple relief pitchers meet the Shutdown standard.  In Atlanta’s win over Houston on Tuesday (April 10), Eric O’Flaherty, Jonny Venters and Craig Kimbrel all earned Shutdowns for Atlanta, having contributed similarly to Atlanta’s chances of winning, while only Kimbrel gets the save.

In short, the Shutdown gives credit whenever, but only when, a reliever actually pitches successfully in a reasonably high-risk situation.  

In 2011, the top five pitchers in the majors in saves were:
Jose Valverde 49
Craig Kimbrel and John Axford 46
J.J. Putz 45
Mariano Rivera 44

In 2011, the top five pitchers in Shutdowns (using b-ref’s Play Index):
Jonny Venters 45
John Axford 44
Heath Bell 43
Craig Kimbrel and Tyler Clippard 41

Jose Valverde, the major league saves leader in 2011, had 38 b-ref Shutdowns, good for eighth in the majors. Two Yankees, Mariano Rviera and David Robertson, tied for 10th in the majors in Shutdowns. 

Most Career Saves, All-Time:
Mariano Rivera 604
Trevor Hoffman 601
Lee Smith 478
John Franco 424
Billy Wagner 422

Most Career Shutdowns, All-Time:
Mariano Rivera 539
Trevor Hoffman 530
Lee Smith and John Franco 474
Rollie Fingers 434

Fangraphs and Baseball-reference use slightly different formulas to calculate WPA, so the Shutdown numbers using the b-ref Play Index will be a bit different than those that Fangraphs lists. But the concept is the same, and in most cases the results will be comparable.

Just as there is a negative version of the save in the form of the blown save, Shutdowns have a negative flip side that Fangraphs calls Meltdowns. I’ll discuss those in a separate post.

0 0 votes
Article Rating
Subscribe
Notify of
guest

63 Comments
Oldest
Newest Most Voted
Inline Feedbacks
View all comments
Ed
Ed
12 years ago

Birtelcom – A comment and a question. My comment is that I like it in theory but I have a hard time seeing it being adopted by the general public, particularly when (as you noted), there’s no standard formula for calculating WPA.

Now my question: Do you know why they’re advocating the 6% levels? Why not 4%? Or 5 or 7?

bluejaysstatsgeek
bluejaysstatsgeek
12 years ago
Reply to  birtelcom

IIRC, Tom Tango developed the metric, but Fangraphs was the first to utilize it. Here’s a link to the discussion regarding the threshold:

http://www.insidethebook.com/ee/index.php/site/comments/shutdowns_and_meltdowns/

bstar
12 years ago

Are you the same bjstatsgeek on fangraphs? just curious!

bluejaysstatsgeek
bluejaysstatsgeek
12 years ago
Reply to  bstar

Yes, also on BlueBirdBanter and Tom Tango’s blog.

Lawrence Azrin
Lawrence Azrin
12 years ago
Reply to  Ed

I agree with Ed, I don’t think “Shutdown” will catch on as a measure of relief pitcher effectiveness with the general public, if it is tied to something as esoteric as WPA. OBA, SA, OPS, and even OPS% have the advantage of building on the easily calculated and accepted statistic ‘Batting Average’.

And why 6%?

Neil L.
Neil L.
12 years ago

birtelcom, a fabulous idea to expand on the Fangraphs suggestion here in a HHS blog. If this doesn’t generate at least 50 comments then I’ll eat John Autin’s Brooklyn Superbas baseball cap, even without ketchup. 🙂 I’m at work, so I only have time for a quick readthrough and a superficial reply. We have all recognized for a while the flaws in the save-blown save statistic. Does the possibility of multiple shutdowns per game cheapen the statistic, in your opinion? How often, in a single game, does a pitcher enter a game and add 6% to his team’s chances of… Read more »

Neil L.
Neil L.
12 years ago
Reply to  Neil L.

Sorry, I was still typing when you posted, #2, birtelcom.

Doug
Editor
12 years ago
Reply to  birtelcom

Shutdowns (I’m guessing) may have a similar frequency to the sum of Saves and Holds. Lots of attention paid to the former, very little to the latter.

Thus, a common stat for all relievers would be better in that equal attention would be drawn to closers and setup men.

The other nice thing about a Shutdown is that, if adopted, would presumably lead to a common defintion of WPA (not that the differences between the calculations appear very significant).

bstar
12 years ago
Reply to  Doug

That’s another problem I see: when’s the last time a new stat came out that didn’t have multiple versions of it across the internet? Just look at WAR. The fact that there are so many equations out there has lowered its acceptance value to commentators then to the masses.

Personally, it saddens me a bit that forming a unified opinion on WAR calcs, WPA calcs, etc. are not at the VERY TOP of the agenda at SABR conferences. If we can’t even agree amongst ourselves, why should anyone else?

Ed
Ed
12 years ago
Reply to  bstar

Agree 100% Btsar!

Neil L.
Neil L.
12 years ago
Reply to  birtelcom

By cheapening the save statistic, birtelcom, I guess I meant it will require a readjustment of our interpretation of seasonal numbers. I know that a 40-save season is pretty good for a closer, but would a 40-shutdown season represent the same level of excellence? Perhaps after a few seasons of shutdowns, I could re-calibrate my baseball intuition, but the automatic cap of one save per game prevents the profileration of the stat. Agreed with others than the slightly different methods of calculating WPA presents an obstacle to universal adoption of shutdown as a legitimate statistic. And I concur with others… Read more »

Neil L.
Neil L.
12 years ago
Reply to  birtelcom

Yeah, but a lot of the comments are mine, so …. pass the ketchup. 🙂

Actually, a great topic, in my opinion. And, birtelcom, without any input from the august John AUtin.

Dave J.
Dave J.
12 years ago

Does anyone know when BR begins publishing WAR stats? I don’t see anything for 2012 as of yet.

Lawrence Azrin
Lawrence Azrin
12 years ago
Reply to  Dave J.

It’s too early for 2012 WAR stats, in the same way that I can’t take this year’s league leaders seriously until no one is batting over .400.

Voomo Zanzibar
Voomo Zanzibar
12 years ago
Reply to  Lawrence Azrin

Yoenis Céspedes is on track for 393 Total Bases and 255 strikeouts.
Justin Verlander is on track to go 0-17 with a .55 WHIP

Lawrence Azrin
Lawrence Azrin
12 years ago
Reply to  Voomo Zanzibar

Austin Jackson, on pace to bat .500, slug .850, with 324 hits, 550 total bases, and 162 runs scored! Back-to-back Tigers MVPs, baby!

kds
kds
12 years ago
Reply to  Voomo Zanzibar

I need a LoL button!

Lawrence Azrin
Lawrence Azrin
12 years ago
Reply to  kds

kds,
Thanks, I’ll be here all day! Sadly, that is more true than I’d like to admit…

topper009
topper009
12 years ago

This still seems a little too random, basically the same as a regular save except it just raises the standard. Why not JUST use the WPA, guys should get still get credit for 3 run saves, just not as much as a 1 run save. And they should get extra credit for saves worth more than 6 WPA points.

Here are the 2011 leaders:
5.227 Tyler Clippard
4.645 Jonny Venters
4.227 John Axford
4.195 David Robertson
4.064 Jose Valverde
3.633 Chris Sale
3.398 Mariano Rivera
3.260 Jonathan Papelbon
3.179 J.J. Putz
3.103 Greg Holland

Lawrence Azrin
Lawrence Azrin
12 years ago
Reply to  birtelcom

This going to seem incredibly obvious to most of you here, but: Teams wins and losses are the result of the performance of a number of individuals in any particular game. That is why any “binary” (as birtelcom astutely points out above) category such as pitcher wins or losses, saves or blown saves, batter BA, RBI or runs scored, will never completely capture the player’s performance isolated from the rest of his team. The general public wants to have one “go to” easily understood stat to rate players, such as Wins, Saves, BA, or RBI, but those aren’t as accurate… Read more »

Lawrence Azrin
Lawrence Azrin
12 years ago
Reply to  birtelcom

Everything you say is true, but binary stats can also have certain arbitrary cutoff points for standards of excellence, that may affect people’s perception of their meaning. For example: Let’s say that in an a June game, a batter comes up with the bases loaded, and hits a rope down the right field line that curves foul at the last moment. Next pitch, harmless fly caught in left field. The batter ends up the year with: 29 HR, 96 RBI if that ball in June stayed fair, it’s: 30 HR, 100 RBI Sounds better, right? (even though we know it’s… Read more »

Doug
Editor
12 years ago
Reply to  birtelcom

Ortiz went 5 for 15 with the bases loaded in 2011, including one HR and 10 RBI. He had 5 strikeouts, 4 groundouts (incl. 3 DPs) and one flyout.

His net WPA for those 15 PAs was -0.3.

topper009
topper009
12 years ago
Reply to  birtelcom

@18, you can say that about anyone, I’m sure each player on the list had 1 chance to make a big swing in their WPA…but they didn’t. You cant assume Sale could have upped his WPA to 4.4 without also assuming Clippard could have upped his to 6 with 1 extra key strikeout.

birtelcom
birtelcom
12 years ago
Reply to  birtelcom

topper @47: That’s precisely my point. WPA season totals are subject to wide swings based on the results of a single game, perhaps more than is ideal. A stat such as Shutdowns, which is accumulated one game at a time, can balance that out. From a Shutdown point of view, Sale’s failure on September 28 just means he is awarded one less Shutdown for the season than he would have had he gotten out of the inning. That may put that one-game failure in a helpful perspective that shifting his WPA for the whole season from 4.4 to 3.6. I’m… Read more »

Hartvig
Hartvig
12 years ago
Reply to  birtelcom

If I understand WPA correctly, a pitcher who got a 2 inning “save” would get more credit than one in a similar situation but only pitching 1 inning. Is that correct?

If so, then I am really in favor of it in place of saves. I am so sick of all these one inning wonders…

Ed
Ed
12 years ago

BTW, Fangraphs added Shutdows and Meltdowns to their player pages a bit over a year ago. Would be nice if Baseball Reference followed suit….

bstar
12 years ago

Here’s two recent articles calling for the save to end:

http://www.grantland.com/blog/the-triangle/post/_/id/23416/blowing-up-baseballs-most-dangerous-stat

and here’s the link that probably inspired birelcom’s article:

http://www.fangraphs.com/blogs/index.php/shutdowns-and-meltdowns-should-kill-the-save/

bstar
12 years ago

The big stumbling block I have with this is how is it going to seep into the awareness of the masses? It takes a live WPA/win expectancy chart to even know if a particular situation is a SDown/MDown situation, plus a calculator to figure out the 6%. It would take the commentators to first be able to understand this(begrudgingly), and Boog Sciambi, former Braves and current underused ESPN commentator, might be the only one out there who’s SABR-savvy enough to even get this stat.

vivaeljason
vivaeljason
12 years ago
Reply to  bstar

Yeah…seeing as how even people who embrace sabermetrics don’t fully understand how WPA words (admittedly yours truly, for example), I really don’t see SDowns and MDowns ever being used by the masses. That said, I like it — especially when compared to saves. I especially think the Shutdown/Meltdown metric is really useful because it gives us better insight into other relievers. The hold is just as useless as the save is (if not moreso), so any metric that can tell you something useful about middle relievers and setup men is okay in my book. I just don’t ever expect to… Read more »

bluejaysstatsgeek
bluejaysstatsgeek
12 years ago
Reply to  bstar

If the masses watch Texas Hold’em on TV – and they post probabilities for each hand – why can MLB broadcasters add win probability and leverage to the broadcasts?

You might be surprised how quickly the masses catch on!

Neil L.
Neil L.
12 years ago

BJSG, an interesting analogy.

One immediate reaction is are the poker hand probabilities displayed on the “live” broadcast or are they added in post-taping editing? I don’t know if I’ve seen a “live” poker telecast and I wouldn’t know one if I saw one.

Provided a universal formula for WPA could be agreed on, it should be possible to put the numbers up on a batter-by-batter basis.

bstar
12 years ago
Reply to  Neil L.

You’re absolutely right, Neil. They DID show the last World Series of Poker main event live, and due to twitter and the internet and texting, could not even show what cards the players were holding. So, no, hand probabilities would not work with live poker broadcasts. I’m a little skeptical that the masses are really going to care about a WPA chart. Even a newcomer to the game understands inherently when a game is close and when it isn’t. We’re the ones who might like a WPA chart while watching a game(you can get this at Fangraphs), but not Joe… Read more »

Neil L.
Neil L.
12 years ago
Reply to  bstar

Hey, bstar, be careful when you sling around those Joe Sixpack references ….. I might resemble your remark!! 🙂

bstar
12 years ago
Reply to  bstar

I am the living epitome of Joe Sixpack, Neil, so I’m allowed. 🙂

bluejaysstatsgeek
bluejaysstatsgeek
12 years ago
Reply to  bstar

A couple of comments: The Win Expectancy used to generate WPA are simple and available on Tom Tango’s blog. (I know that the WPA values reported on FG and BR use the different tables, but the differences are minor. FanGraphs, which I almost always have on during a game, generates them in realtime – all one needs is the Inning/Base/Out state and the value is a direct lookup from the table. There is no reason why these cannot be generated in real time during broadcasts. Second, bstar, where did you pick up the “Joe Sixpack” expression. I use it a… Read more »

bstar
12 years ago
Reply to  bstar

bjsg, it’s always been a generic term for an average baseball fan, or perhaps more generally just the average American male(similar to Joe Shmoe).

While I may have your attention, I was watching the Marlins-Phillies game and was looking at the FGraphs WPA chart but could not figure out whether Papelbon entering the ninth with a 2-run lead qualified for a shutdown. Where do you actually find this number specifically?

bluejaysstatsgeek
bluejaysstatsgeek
12 years ago
Reply to  bstar

As the top of the 9th started, the Phillies win expectancy (WE) was 93.9%. Therefore, as Papelbon was successful closing out the game the WE moved from 93.9% to 100%, a change of 6.1%, which is Papelbon’s WPA.

I seem to recall Tom Tango, but it could have been someone else, commenting that they liked that a 2-run 3-out save was a shutdown, but a 3-run 3-out save, didn’t qualify.

Neil L.
Neil L.
12 years ago
Reply to  bstar

BJSG @39, Generating that kind of on-screen display during a major-network broadcast of a basball game would require that the producers of the broadcast feel viewers actually want it. Let alone that the Orel Hershisers, Tim McCarvers and Joe Morgans of the world understand WPA and shutdown. In addition, with all due respect, “Joe Sixpack” is not going to sit there with his laptop open to Fangraphs during Saturday afternoon baseball on FOX or Sunday night baseball on ESPN. You and I maybe, but the not the average viewer. What would it take for shutdown/meltdown to become an “official” stat… Read more »

bluejaysstatsgeek
bluejaysstatsgeek
12 years ago
Reply to  bstar

Neil @43: But imagine how the gaming industry would like this!

Mark in Sydney
Mark in Sydney
12 years ago

Having just watched Brian “Fear the Beard” Wilson do his thing I am not 100% sure about this whole save/shutdown thing. It strikes me as trivializing either batters or pitchers, not sure which. The Beard comes in, bottom of 9, with a 4-1 lead. So, a 3% WPA, not a save. First 2 batters up are Tulowitzki and Cuddyer. Neither easy outs. One double and an in-field single. He ends up walking a run in, and getting the win, leaving runners at 2 and 3. Under the shutdown idea, this should have been easy pickings, due to WPA. It misses… Read more »

Big Daddy V
Big Daddy V
12 years ago
Reply to  Mark in Sydney

Sure, those batters may not be easy outs, but the simple fact is that it is just not an accomplishment to avoid giving up 3 runs in 1 inning. That should be the bare minimum expectation.

Neil L.
Neil L.
12 years ago
Reply to  Big Daddy V

Big Daddy, ??

Not quite sure if I caught your meaning? You are agreeing that there are “cheap” saves and “quality” saves as the statistic is currently defined?

Yet all saves are created equal, even if it represented a somewaht ineffective outing, like Wilson’s.

Neil L.
Neil L.
12 years ago

Mark in Sydney, I am developing a lot of respect for your analytical skills and baseball acumen. Now, can you explain to me the nuances of Aussie rules football? 🙂 Seriously, in looking over your comment, Mark, are you saying that game WPA should be tweaked to take into account the threat posed by the hitters the closer has to face? In other words, a save or shutdown should take into account the batting order positions faced by the pitcher? That sounds like cutting the save/shutdown statistic into a million tiny pieces. With all of our angst about the save/blown… Read more »

Mark in Sydney
Mark in Sydney
12 years ago
Reply to  Neil L.

Neil, I am not convinced that any of it is needed, not Wins, or Saves, or ERA, when it comes to pitching. In all these situations we are tagging an individual for what is a team effort, and I am not sure that is at all meaningful. Traditional, sure, but not too meaningful. Pitchers pitch. And there are some stats that attempt to do the fielder independent thing to give an idea of their contribution. But baseball is, at its heart, a team game, though one that needs heroes. Compare two of the best pitchers over the last decade: Halliday… Read more »

bluejaysstatsgeek
bluejaysstatsgeek
12 years ago
Reply to  Mark in Sydney

I believe that WPA needs refinement to distribute credit better, not solely to the batter or pitcher. Tom Tango, as I recall, disagrees. Maybe he thinks that there is simply noise that averages out. In my view, when a catcher nails a runner attempting to steal second, a lot of thing had to happen: a quick pitch to the plate, a good transfer and throw by the catcher and a good catch with the tag applied by the middle infielder. Yet, as WPA is calculated now, only the pitcher gets credit for the out. I strongly believe that defensive players… Read more »

Neil L.
Neil L.
12 years ago

You have my attention, BJSG. Are you saying that WPA, as it is currently constituted, credits the out entirely to the pitcher? The catcher gets credit for the caught stealing. How can one ever credit defensive positioning as a minor part of WPA or WAR? What about a manager’s WPA for a defensive positioning? An example is Brett Lawrie, as a 3rd baseman, being positioned in shallow centre field against Ortiz or Gonzalez where Lawrie turns a sharply-hit ball up the midlle into an out. The pitcher gets the WPA buff because of an out. We can parse WPA down… Read more »

bluejaysstatsgeek
bluejaysstatsgeek
12 years ago
Reply to  Neil L.

I thought I posted this reply already: I am working on a paper that will create a framework for breaking down attribution for WPA, but it will be a few months before it’s finished.

Neil L.
Neil L.
12 years ago
Reply to  Neil L.

BJSG @51, perhaps I missed that reply. Where do you anticipate the paper will be published?

I admire you, BJSG, if you have the ability to write clearly about complex baseball metrics, although I understand it is not possible to completely demathematize advanced sabremetrics.

The Hidden Game of Baseball would be an example of what I mean about clear, interesting writing to explain statistics.

Shping
Shping
12 years ago
Reply to  Neil L.

I too, BJSG, will applaud your efforts in advance for such a book and analysis. But i also share Neil’s concerns about parsing it down to the ludicrous. And not sure if this would factor in your analysis, but Mark @31 brought up a good point earlier about individual situations: Should WPA for a pitcher be adjusted when a Troy Tukowitski is at the plate, instead of a Mario Mendoza? Part of me says yes. But then again, how could you or anyone possibly account for a Tulowitski or Pujols hitting only .159 on April 7, when we still “know”… Read more »

blusjaysstatsgeek
blusjaysstatsgeek
12 years ago
Reply to  Neil L.

@Shping/62: It’s not a book, but a paper. It could be construed by some as parsing down to the ludicrous, but I see how the reaction is when it is done.

I’m simply too busy in the next month to work on it. I’ve been reduced to listening to games and watching recaps.

Tristram12
Tristram12
12 years ago

I know I’m late to the thread, but I’m surprised no one pointed out how little the leaderboards changed. I really expected quite a bit more change and a rise in the ‘success’ recognition of middle relievers. All of which makes me question if it’s worth it to try to get mainstream adoption.

Neil L.
Neil L.
12 years ago
Reply to  Tristram12

I did notice that also, Tristam, but didn’t comment on it. The correspondance in career totals between shutdowns and saves is surprising. It shows that the Riveras and Hoffmans of the baseball world didn’t get many cheap saves.

bstar
12 years ago
Reply to  Neil L.

Here’s a thought: why not just tweak the Save stat itself? Eliminate the three-run lead, one inning save and any others that some find objectionable. We already have holds; if they need tweaking also, then do that. I just think introducing a stat where you need a live win expectancy chart has little chance of catching on.

Neil L.
Neil L.
12 years ago

I’m interested in the meltdown post to come, birtelcom.

Presumably there is a natural cap to the number of meltdowns a pitcher could have abd still hold down a ML roster spot.

Neil L.
Neil L.
12 years ago

My brain must be cramping up here. A shutdown is based on a pitchers total WPA for the game being >= 0.06? So it doesn’t require a win by your team to be credited? A long reliever, coming in early in a game, when his team is far behind would get a shutdown if his team came back to make it close enough for him to accumulate a WPA of 0.06? A starter is ineffective and is replaced after, say three innings in a free wheeling 5 to 5 game. The reliever pitches three innings in relief, and leaves with… Read more »

Neil L.
Neil L.
12 years ago
Reply to  Neil L.

If I’m understanding this shutdown thing properly, the Play-Index, prior to today’s games, shows 195 shutdowns so far in 2012, compared to 60 saves. That’s over a 3 to 1 ratio.

Interesting that, in the 195 shutdowns, 55 times the pitcher’s team lost the game. Puts a different slant on the shutdown compared to the save.

Shping
Shping
12 years ago

Late to this discussion, but a couple comments: Love the concept of Shutdowns for being able to eliminate cheap saves, and at first glance, i don’t think the fact there can be multiple occurrences per game cheapens it. As we all know, sometimes the true “shutdowns” occur long before the 9th inn (accurately or not, for example, i’m thinking of the Shutdown that Boston could have earned in the famous Pedro/Yankees/Grady Liddle game of 2003 — if only Grady had pulled Pedro). But then again, Neil’s numbers posted above almost remind me in some weird way of the infamous GWRBI.… Read more »