Relief Pitching, 538 way

[Mike Sixel]

This is a great article about how RP and the save statistic are kind of bad for baseball.....

 

https://fivethirtyeight.com/features/goose-egg-new-save-stat-relief-pitchers/?addata=espn:mlb:index

 

and, it has some nice hearted snark, imo.

[Sssuperdave]

This is a seriously good article and new statistic. 

[TheLeviathan]

Awesome read, thanks!

[snepp]

I'm sold.

[Willihammer]

I like it. 

[Mike Sixel]

I know it is full of snark, but it is actually a great stat. What we choose to measure, and how we measure it, should be done to help us improve who/what we do. Saves incent bad behavior, imo.

[Brock Beauchamp]

Yes, this, so much this.

 

I hate the save statistic with a fiery passion.

 

Which is why I'm okay with Kintzler being the "closer". If teams are going to play around a statistic, the "closer" should be:

 

1. A competent pitcher

2. A guy who doesn't give up homers on demand

3. The team's third best reliever

[TNTwinsFan]

I can see two guys on the roster that could fill this role and do it well:

 

1. Tyler Duffy

2. Ryan Pressley

 

I think Duffy has the stuff, the mentality, and now (because of his work as a starter for the past few years) the durability to handle the position.

 

DO. IT.

[Mike Sixel]

I want them to try Duffy in this role, very much. Probably also/or Haley. 

[Oldgoat_MN]

I like it.

As long as they don't use a name that 'honors' a guy who is at best a jerk and more likely a [redacted] [redacted] [redacted].

[WLFINN]

Based on nothing but my own observation, I think Molly and/or Levine-Falvey have already been using the relievers in this way moreso than has been done in the past.  Pressly and Duffey have been the individuals used in such situations.

[nicksaviking]

I don't like it. But only because I completely agree and have been advocating for the last half decade or so that the Twins groome "closers" by inflating Joe Blow bullpen guys with save stats and flipping them. It's super easy to create a closer out of thin air; can't really do that with any other position.

 

So long dreams of riches and undeserved prospects.

[Thrylos]

Based on this little thing from there, one can use WPA to measure relievers and call it a day.  Not sure I like it (according to this Taylor Rogers and Ryan O'Rourke were the Twins' best relievers in 2016,) but WPA is a stat easy to find:

[ashbury]

People want counting stats they can see with the aid of a scoresheet at most, in preference to most rate stats more complex than BA and ERA. Win Probability Added is more in the latter part of the spectrum.

 

I do like WPA though, as it is a decent stab at combining performance with leverage.

[Sssuperdave]

 

Based on this little thing from there, one can use WPA to measure relievers and call it a day.  Not sure I like it (according to this Taylor Rogers and Ryan O'Rourke were the Twins' best relievers in 2016,) but WPA is a stat easy to find:

 

 

You could "use WPA" to measure relief pitchers, but the Goose egg stat helps put some heuristics around when the opportunities for a high WPA will present themselves.   Goose eggs inform reliever usage and WPA informs reliever effectiveness.

[Mike Sixel]

Not to mention that correlation =/= as effective or the same.....as he points out literally in the article. 

[Brock Beauchamp]

 

You could "use WPA" to measure relief pitchers, but the Goose egg stat helps put some heuristics around when the opportunities for a high WPA will present themselves.   Goose eggs inform reliever usage and WPA informs reliever effectiveness.

Yes. There is plenty of room for different stats.

 

Imagine you're in a broadcast booth. Now try to explain what WPA means because saying the number 2.1 doesn't mean squat to the audience. Was that in 31 appearances or 65 appearances? How many innings pitched? What is the player's typical role out of the bullpen, fifth or eighth inning? The number, on its own, means virtually nothing. It's super-useful for a deeper analysis and comparison of performance but not good for on-the-fly talking.

 

Whereas counting stats have their use in the quick-and-dirty analysis that often happens on live radio/television. When a guy is pitching to a batter, the announcer doesn't have time to spend 90 seconds explaining to the audience all the situations that led to that 2.1 WPA.

 

But saying "Tonkin has 21 Goose Eggs and 4 Broken Eggs" gives you lots of information in just ten words.

[Mike Sixel]

 

Yes. There is plenty of room for different stats.

 

Imagine you're in a broadcast booth. Now try to explain what WPA means because saying the number 2.1 doesn't mean squat to the audience. Was that in 31 appearances or 65 appearances? How many innings pitched? What is the player's typical role out of the bullpen, fifth or eighth inning? The number, on its own, means virtually nothing. It's super-useful for a deeper analysis and comparison of performance but not good for on-the-fly talking.

 

Whereas counting stats have their use in the quick-and-dirty analysis that often happens on live radio/television. When a guy is pitching to a batter, the announcer doesn't have time to spend 90 seconds explaining to the audience all the situations that led to that 2.1 WPA.

 

But saying "Tonkin has 21 Goose Eggs and 4 Broken Eggs" gives you lots of information in just ten words.

 

Correct, I'd say this stat is elegant, whereas WPA is not.....I prefer elegant. 

[Brock Beauchamp]

 

Correct, I'd say this stat is elegant, whereas WPA is not.....I prefer elegant. 

Yeah, elegant is one way to put it... Mostly, I view them as having different roles.

 

Sometimes, I want to say Eddie Rosario has an 80.3% contact rate, swings at 43.7% of pitches out of the zone, 78.4% of pitches in the zone, and is hamstrung by a .213 BABIP.

 

Other times, I just want to say Eddie Rosario has a .621 OPS.

 

In some situations, the deep dive leads to better, more nuanced analysis. In other situations, you just want to keep it (somewhat) superficial and talk about results.

 

There is no right answer for all situations.

[Mike Sixel]

It's funny how uptight some people get, when you use counting stats to talk about the past and outcomes, as if more predictive stats tell the same story. They don't. different stats are used for different stories. 

[Sssuperdave]

 

It's funny how uptight some people get, when you use counting stats to talk about the past and outcomes, as if more predictive stats tell the same story. They don't. different stats are used for different stories. 

 I couldn't agree more with this.  There is nothing wrong with using counting stats to talk about what actually happened.  They help tell the story of what happened in a game, a season, etc., and we're all baseball fans so shouldn't we enjoy that?  The problem is when a non-predictive stat is used as a defense for who should or should not be signed for next season.

 

Take something like RBIs.  They are heavily situation-dependent, aren't always predictive, and don't even do the best job of differentiating a players skill.  But, if a guy has 120 RBI in a season it means a lot of exciting stuff happened while he was at bat that year, and I hope I was watching some of it! 

[Otto von Ballpark]

 

Yes. There is plenty of room for different stats.

 

Imagine you're in a broadcast booth. Now try to explain what WPA means because saying the number 2.1 doesn't mean squat to the audience. Was that in 31 appearances or 65 appearances? How many innings pitched? What is the player's typical role out of the bullpen, fifth or eighth inning? The number, on its own, means virtually nothing. It's super-useful for a deeper analysis and comparison of performance but not good for on-the-fly talking.

 

Whereas counting stats have their use in the quick-and-dirty analysis that often happens on live radio/television. When a guy is pitching to a batter, the announcer doesn't have time to spend 90 seconds explaining to the audience all the situations that led to that 2.1 WPA.

 

But saying "Tonkin has 21 Goose Eggs and 4 Broken Eggs" gives you lots of information in just ten words.

WPA/LI is better, and addresses many of your concerns.  I'd say it's no different than citing ERA or OBP/SLG.  All stats need some context.

 

 

Goose/Broken eggs don't appear to be frequent enough for most pitchers to be terribly useful either.  Your Tonkin example is nice, but according to Jeremy's article, that sample threshold was reached by just two Twins pitchers last year.  How useful is it to know that Tonkin was 5-for-6?

 

Forget WPA/LI -- is this thing telling us much that ERA and IP isn't already telling us?

 

[Otto von Ballpark]

 

Based on this little thing from there, one can use WPA to measure relievers and call it a day.  Not sure I like it (according to this Taylor Rogers and Ryan O'Rourke were the Twins' best relievers in 2016,) but WPA is a stat easy to find:

 

 

Actually WPA/LI would be better, because it would account for leverage.  Rogers and O'Rourke were racking up their WPA in a lot of lower-leverage situations.  Here are 2016 Twins relievers ranked by WPA/LI:

 

http://www.fangraphs.com/leaders.aspx?pos=all&stats=rel&lg=all&qual=0&type=3&season=2016&month=0&season1=2016&ind=0&team=8&rost=0&age=0&filter=&players=0&sort=12,d

 

Of course, it's probably not all that different of a ranking from RA9, or better yet for relievers, RA9 plus inherited runners.

[drjim]

This stat is a clear improvement on a save/hold, but I'm not sure it would lead to that significant of a difference in bullpen usage. And the gains it would lead to are pretty much happening organically. If it was adopted, would probably lead to more money for middle relievers in arbitration years, but not sure it would change free agent contracts or other player valuation all that much.

[Oldgoat_MN]

 

Actually WPA/LI would be better, because it would account for leverage.  Rogers and O'Rourke were racking up their WPA in a lot of lower-leverage situations.  Here are 2016 Twins relievers ranked by WPA/LI:

 

http://www.fangraphs.com/leaders.aspx?pos=all&stats=rel&lg=all&qual=0&type=3&season=2016&month=0&season1=2016&ind=0&team=8&rost=0&age=0&filter=&players=0&sort=12,d

 

Of course, it's probably not all that different of a ranking from RA9, or better yet for relievers, RA9 plus inherited runners.

So... Escobar should be pitching more?

That would be more fun than watching Tonkin.

[Teflon]

The modern use of the closer in only statistical-save-potential starting-the-9th inning situations is pure buffoonery.

 

For older Twins fans out there who remember Doug Corbett, the 538 article makes you appreciate his 1980 season in comparison to modern relief pitching all the more. According to the 538 article, Corbett had the 14th highest single season in Goose Egg performance in MLB history. I would add that he was first in the American League that year in win-probability added, as well. He pitched 136.1 innings in 73 games allowing only 102 hits. He inherited 80 runners that season and only 19 scored. The average pressure situation when he entered games was 1.780. (Jeurys Familia who led the Majors with 51 saves last year had an average pressure index in games he entered of 0.42) Looking at his game log that season I found a stretch in August where he pitched an iinning on the 5th, an inning-and-a-third on the 6th and six-and-two-thirds innings on the 7th! He entered the game with one out in the 8th inning and pitched through the end of the 14th.

 

I'm not advocating that teams use their best relief pitcher like Corbett was used in 1980 - but we should ABSOLUTELY use the best pitchers in the highest leverage situations. Like Gossage says, if there is a pitcher on the staff that can't protect a three-run lead pitching one inning, he shouldn't be on the staff.

 

 

[Brock Beauchamp]

 

Goose/Broken eggs don't appear to be frequent enough for most pitchers to be terribly useful either.  Your Tonkin example is nice, but according to Jeremy's article, that sample threshold was reached by just two Twins pitchers last year.  How useful is it to know that Tonkin was 5-for-6?

It's useful in that it tells us Tonkin is a low leverage reliever.

 

But use someone like Kintzler instead. He qualified for 23 Goose Eggs last season versus 20 save attempts.

 

Will all relievers qualify for a bunch of goose egg opportunities? No, of course not... because there are some parameters on the goose egg, they're simply not as nonsensically dogmatic by inning and number of outs.

 

Which means some throwaway "saves" won't count but some important 7th and 8th inning outings will count.

 

And isn't that the original intent of the save statistic in the first place? To try to give value to relievers and give them credit for helping their team win baseball games? It seems to me the goose egg strikes a pretty good balance between a counting stat and being more useful than the save statistic.

[Otto von Ballpark]

 

The average pressure situation when he entered games was 1.780. (Jeurys Familia who led the Majors with 51 saves last year had an average pressure index in games he entered of 0.42) 

 

I think you're looking at the wrong column or something?  Familia had a 1.67 gmLI last year (game-entering leverage index), pretty comparable to Corbett's 1.73:

 

http://www.baseball-reference.com/players/f/familje01.shtml

[Otto von Ballpark]

 

It's useful in that it tells us Tonkin is a low leverage reliever.

 

But use someone like Kintzler instead. He qualified for 23 Goose Eggs last season versus 20 save attempts.

 

Will all relievers qualify for a bunch of goose egg opportunities? No, of course not... because there are some parameters on the goose egg, they're simply not as nonsensically dogmatic by inning and number of outs.

 

Which means some throwaway "saves" won't count but some important 7th and 8th inning outings will count.

 

And isn't that the original intent of the save statistic in the first place? To try to give value to relievers and give them credit for helping their team win baseball games? It seems to me the goose egg strikes a pretty good balance between a counting stat and being more useful than the save statistic.

 

I didn't need to know that Tonkin had 6 goose egg opportunities, in 65 total appearances, to know that he was a low-leverage reliever last year.  I could just look at his average game-entering leverage index of 0.80.

 

The goose egg sounds more useful than the save stat, no doubt. But the save stat isn't very useful at all.  I think the bar should be set higher for its potential replacement.

[Thrylos]

 

Yes. There is plenty of room for different stats.

 

Imagine you're in a broadcast booth. Now try to explain what WPA means because saying the number 2.1 doesn't mean squat to the audience. Was that in 31 appearances or 65 appearances? How many innings pitched? What is the player's typical role out of the bullpen, fifth or eighth inning? The number, on its own, means virtually nothing. It's super-useful for a deeper analysis and comparison of performance but not good for on-the-fly talking.

 

Whereas counting stats have their use in the quick-and-dirty analysis that often happens on live radio/television. When a guy is pitching to a batter, the announcer doesn't have time to spend 90 seconds explaining to the audience all the situations that led to that 2.1 WPA.

 

WPA is a counting stat      "A" stands for "added", which is the epitome of counting.

 

I'd say that if you are not able to explain WPA, you do not belong in an MLB broadcasting booth in 2017.

Sorry DickNBert.