Mulligan ERA-- A What if Scenario

by dannijd on Jun 3, 2010 1:53 PM EDT

EDITOR'S NOTE: Lost in the shuffle over the offensive collapse and the Perfect Game. Promoted from the FanPosts. - WC

Last week, there was a post on this site discussing how good Jamie Moyer's ERA would be if the two innings where he gave up five runs each were removed. It got me to thinking-- even the greats have bad innings (I am sure Doc Halladay would like the sixth inning of last Sunday's game back), that raise their ERAs. What would happen if they were allowed a mulligan or two?

This led me to the concept of Mulligan Earned Run Average-- MERA-- an entirely new creation. This figure is determined by removing both the two worst innings (or partial innings) over the course of the season, along with two innings in which they gave up no runs (in order to shave off two of their best-- the goal is to force us to look closer at the typical run of the mill events over the course of the season. Calculations were run for each member of Phillies rotation. J.A. Happ (ERA 0.00) was left off both due to the impossibly small sample size and the fact that he has not given up an earned run. I also left Nelson Figueroa out, as his sample size as a starter was too small to be predictive.

In addition to determining the new ERA numbers, I determined the degree to which this recalculation affected the pitcher's ERA, as this percentage seemed to be indicative of a high ERA being caused by a couple of bad innings as opposed to a tendency to give up runs. While Jamie Moyer received the greatest benefit of any full time starter (ERA 4.55, MERA 3.20 for a 30% change), Joe Blanton received the most benefit (ERA 5.06, mERA 3.52, for a 36% change), this is probably due to a small sample size. Roy Halladay, with a 28% drop in his ERA (ERA 2.22, mERA 1.60) showed the third largest drop, followed closely by Cole Hamels with a 27% drop (ERA 3.92, mERA 2.86). The lowest percentage change, by far was that of Kyle Kendrick, with a 23% change, from a 5.66 ERA to a 4.37 mERA. This makes sense given his propensity to give up several runs over the course of a game.

Below is the math behind the numbers. I was amazed at the amount of change that was caused in general by the impact of one or two bad innings. It will be interesting to look back at these numbers later in the season, and see if the ERA trends closer to the median as innings are added and hopefully bad innings are avoided, or if true ERA is more predictive of player performance.

MERA

Halladay
4/5/2010- 7 innings pitched, 1 earned run (1 in first)
4/11/2010- 9 innings pitched, 1 earned run (1 in the 3rd)
4/16/2010- 8 innings pitched, 1 earned run (1 in the 4th, 1 in the 8th)
4/21/2010- 9 innings pitched, 0 earned runs
4/26/2010- 7 innings pitched, 5 earned runs (****2 in the 1st****, 1 in the 2nd, 1 in the 6th, 1 in the 7th)
5/1/2010- 9 innings pitched, 0 earned runs
5/6/2010- 7 innings pitched, 1 earned run (1 in the 7th)
5/12/2010- 6.1 innings pitched, 2 earned runs (1 in the 1st, 1 in the 7th)
5/18/2010- 9 innings pitched, 2 earned runs (1 in the 2nd, 1 in the 6th)
5/23/2010- 5.2 innings pitched, 6 earned runs (1 in the 2nd, 1 in the 4th, ****4 in the 6th****)

Total Innings Pitched: 77
Total Earned Runs: 19
ERA: 2.22

2 Worst Innings
4/26/2010- 2 run 1st
5/23/2010- 4 run 6th (2/3 of an inning pitched)

MERA
Innings Pitched: 73.1
Earned Runs: 13
MERA: 1.60

Percent Change: 28%

Hamels:
4/7/2010- 5 innings pitched, 2 earned runs (2 in third)
4/12/2010- 5.2 innings pitched, 4 earned runs (1 in the 2nd, ****3 in the 3rd****)
4/18/2010- 8 innings pitched, 2 earned runs (1 in the 2nd, 1 in the 9th)
4/23/2010- 6 innings pitched, 6 earned runs (****5 in the 4th****, 1 in the 5th)
4/28/2010- 6 innings pitched, 4 earned runs (1 in the 5th, 3 in the 6th)
5/4/2010- 8 innings pitched, 1 earned run (1 in the 9th)
5/9/2010- 5 innings pitched, 3 earned runs (3 in the 5th)
5/16/2010- 6.2 innings pitched, 2 earned runs (2 in the 6th)
5/21/2010- 7 innings pitched, 1 earned run (1 in the 1st)

Total innings pitched: 57.1
Total earned runs: 25
ERA: 3.92

2 Worst Innings
4/12/2010- 3 run 3rd
4/28/2010- 5 run 4th

MERA
Innings Pitched: 53.1
Earned Runs: 17
MERA: 2.86

Percentage Change: 27%

Kendrick:
4/8/2010- 4 innings pitched, 5 earned runs (3 in the 1st, 2 in the 4th)
4/14/2010- 1.2 innings pitched, 6 earned runs (3 in the 1st, 3 in the 2nd)
4/20/2010- 8 innings pitched, 0 earned runs
4/25/2010- 5 innings pitched, 5 earned runs (****5 in the 5th****)
4/30/2010- 5 innings pitched, 4 earned runs (3 in the 3rd, 1 in the 5th)
5/5/2010- 7 innings pitched, 0 earned runs
5/10/2010- 6 innings pitched, 4 earned runs (1 in the 1st, 1 in the 3rd, 2 in the 5th)
5/17/2010- 8 innings pitched, 2 earned runs (1 in the 1st, 1 in the 5th)
5/22/2010- 4.2 innings pitched, 5 earned runs (1 in the 4th, ****4 in the 5th****)

Total Innings Pitched: 49.1
Earned Runs: 31
ERA: 5.66

Two Worst Innings:
4/25/2010- 5 Run 5th inning
5/22/2010- 4 Run 5th inning

MERA
Innings Pitched: 45.1
Earned Runs: 22
MERA: 4.37

Percent Change: 23%

Moyer:
4/10/2010- 6 innings pitched, 5 earned runs (****5 in the 3rd****)
4/17/2010- 6 innings pitched, 5 earned runs (****5 in the 1st****)
4/22/2010- 6 innings pitched, 0 earned runs
4/27/2010- 6 innings pitched, 4 earned runs (2 in the 2nd, 2 in the 5th)
5/2/2010- 6 innings pitched, 5 earned runs (3 in the 1st, 2 in the 4th)
5/7/2010- 9 innings pitched, 0 earned runs
5/14/2010- 6.1 innings pitched, 4 earned runs (3 in the 2nd, 1 in the 7th)
5/19/2010- 7 innings pitched, 2 earned runs (1 in the 3rd, 1 in the 7th)
5/25/2010- 5 innings pitched, 4 earned runs (1 in the 1st, 1 in the 2nd, 1 in the 4th, 1 in the 5th)

Total Innings Pitched: 57.1
Total Earned Runs: 29
ERA: 4.55

Two Worst Innings
4/10/2010- 5 run 3rd
4/17/2010- 5 run 1st

MERA
Adjusted Innings Pitched: 53.1
Adjusted Earned Runs: 19
MERA: 3.20

Percent Change: 30%

Blanton:
5/3/2010- 6.2 innings pitched, 4 earned runs (1 in the 3rd, ****3 in the 7th****)
5/8/2010- 6 innings pitched, 3 earned runs (3 in the 6th)
5/15/2010- 7 innings pitched, 5 earned runswm (1 in the 5th, 1 in the 6th, ****3 in the 7th****)
5/20/2010- 7 innings pitched, 3 earned runs (1 in the 5th, 2 in the 7th)

Total Innings Pitched: 26.2
Total Earned Runs: 15
ERA: 5.06

2 Worst Innings
5/3/2010- 3 Run 7th (2/3 of an inning)
5/8/2010- 3 Run 6th

MERA
Innings Pitched: 23
Earned Runs: 9
MERA: 3.52

Percent Change: 36%

Tweet Comment 27 comments | 0 recs |

Do you like this story?

Comments

Great post, great thoughts

But they’re aint no mulligans in baseball.

Moyer kind of reminds of an inverse phillie… he’s a big inning kind of guy, just in the bad way.

A two run inning is the 2nd worst inning Halladays has had all year. crazy

"Although I may not agree with what you're saying, I'll fight to the death for your right to say it. Good day, sir." Pete Griffin

by RiotJuice on May 26, 2010 11:05 PM EDT reply actions

Yeah… I have to admit that running Halladay’s numbers gave me a bit of a reality check on him— maybe he is as good as the hype.

by dannijd on May 27, 2010 10:27 AM EDT up reply actions

As the resident baseball metric creator, I actually don’t find this idea half bad. Certainly interesting, and worth playing with. There are two small thoughts and then one real test that you want to run. Small thought number one is that you should have a certain percentage of innings removed instead of just two innings because obviously Joe Blanton’s two worst innings make up more of his overall innings than a guy who has already thrown 80 innings. Small thought number two is that you shouldn’t call it median earned run average, because median and average is an oxymoron. If you were doing a median earned run prevention measure, then I would call it ERM (earned run median), but yours is really Mulligan ERA. Call it that. It’s catchy and informative.

Next, this is the main thing… the question is “how useful is this metric?” The answer comes down to whether it gives you better information than ERA is giving you. What you want to do is take a stat like Baseball Prospectus’ SNVA (Support Neutral Value Added) which tells you what a pitcher’s winning percentage should be if he had average run support. You run a correlation of mERA with SNVA and of regular ERA with SNVA and see which does better. If getting rid of a couple mulligan innings gives you a better estimation of how often a pitcher wins games, you have a useful statistic to market. If getting rid of those innings hides a measure of effectiveness that is contained in that data, then you don’t have much to work with.

Just my two cents for free…

by Matt Swartz on May 27, 2010 12:04 AM EDT reply actions

It is an interesting idea.

While the stat probably requires a lot of tweaking, I do think the idea starts to get at an issue that really is somewhat significant and useful, which is the importance of managerial decisions on a pitcher’s ERA. If you have two identical pitchers, and the first pitcher’s manager pulls him when he starts to get in trouble for a big inning and thereby minimizes the damage, while the second pitcher’s manager leaves him in to get beat up, the two pitchers could come out of their games with vastly different ERAs, even though their performances were really identical.

And some managers consistently make bad decisions with their pitchers, so this isn’t necessarily the kind of thing that will even out completely over the course of a long season.

by taco pal on May 27, 2010 10:13 AM EDT up reply actions

I should add that managerial decisions can not only affect (and distort) ERA, but also the more advanced fielding-independent stats as well.

by taco pal on May 27, 2010 10:35 AM EDT up reply actions

I agree… unfortunately, there is not (that I know of) a statistical managerial intelligence index for pitchers…. and I have yet to come up with a good enough tweak for that, so the pitcher is unfortunately going to have to statistically take it on the chin.

by dannijd on May 27, 2010 10:44 AM EDT up reply actions

The strange thing about hat is looking at some of the bad innings that I removed, they were not even the last ining that the pitcher pitched— this is true for one of Halladay’s bad innings, and both bad innings for Hamels and Moyer. While it is true that managerial bad decisions can inflate ERA (can someone explain to me why Pedro Martinez was left in both World Series games as long as he was????), they are not the only thing that leads to the inflated innings.

by dannijd on May 27, 2010 3:09 PM EDT up reply actions

That’s true. These two issues intersect a bit, but they’re not the same thing.

by taco pal on May 27, 2010 3:34 PM EDT up reply actions

Pedro

Apparently he has some Jedi mind tricks that work on managers (but not opposing hitters). See 2003 ALCS Game 7.

by phatj on May 27, 2010 8:32 PM EDT up reply actions

That was the place that I felt Manuel should have learned from. When I can look at the pitcher and tell that he needs to be removed before disaster occurs (Pedro and Hamels are the good examples… in last year’s playoffs, I saw those disasters coming and yelled at the TV accordingly)… it was like watching a train wreck (or the last 5 Phillies games)

by dannijd on May 28, 2010 9:12 AM EDT up reply actions

last 5? I was thinking 11 (with an intermission for Halladay’s Perfecto mingled in)

by Bilzo on Jun 3, 2010 2:16 PM EDT up reply actions

It was 5 on 5/28… now you are right, it is the last 11 with the exception of the perfecto and a small and limited part of the game on the Friday prior (I do not consider portions of any game that we lead to be train wrecks).

by dannijd on Jun 3, 2010 2:24 PM EDT up reply actions

The percentage thing is something that I thought of, but I was not sure what an approrpriate percentage would be— I am open to hearing thoughts and suggestions.

I liked your idea of calling it Mulligan ERA, and have in fact changed my name to fit with your suggestion. The reason that I was callingit Median had to do with the idea of finding what was in the middle for the pitchers involved— after removing the same number of innings top and bottom. But I like yours, and so it shall be named that.

I will have to think about trying your correlation theory… I see a baseball prospectus subscription in my future— I love their sister site, Football Outsiders for stat based football analysis and think I might enjoy looking at baseball the same way. I will try to find time to tweak it soon.

by dannijd on May 27, 2010 10:42 AM EDT up reply actions

You’ll find that we’re light years ahead of football performance analysis over at BP. It’s not that we’re better, but just that baseball lends itself to this kind of analysis better because there isn’t as much of an interaction between players. One player bats at a time and all.

My suggestion for Mulligan ERA would actually be to take a few different %s and see if one comes out better. You have the freedom to fine tune your own metric to make it correlate with SNVA or whatever other statistic as well as you can get it to.

by Matt Swartz on May 27, 2010 2:27 PM EDT up reply actions

It’s also easier to break down into discrete portions of the game that aggregate nicely. You can statistically measure each player’s defensive prowess, and relatively measure the prowess of a particular infield. As an example, it’s much harder to define the influence of an offensive line on a running back’s stats; my opinion is that the line matters more than the back in most situations, but it’s hard to statistically demonstrate that.

Honor is no substitute for victory.

by The Dark on May 27, 2010 2:59 PM EDT up reply actions

More useful

If this could be perfected, intuitively, I would think it would be more useful to the average fan. ERA is supposed to tell us what kind of outing we can expect from a pitcher, on average. If two or three bad innings over the course of a season of 150 or 200 skew an ERA substantially, we are not getting the information we should from the stat.

by David S. Cohen on Jun 3, 2010 2:26 PM EDT reply actions

Dannijd… What I like most about this post is you had the balls to actually publish it. It takes guts to come up with something like this and put it out there at the risk of being criticized, torn up, and accused of stupidity or worse, like the responses to a few of my posts :)

I like it. Nicely done!

by Boundforbeach on Jun 3, 2010 3:11 PM EDT reply actions

Thank-you!

by dannijd on Jun 3, 2010 3:16 PM EDT up reply actions

Plus, we didn’t even need to send out the Matt Swartz light into the nighttime sky! He appeared without it!

Numbers always get things going around here. Good research. What I believe you’re driving at is Earned Run Mode rather than Average.

by Wet Luzinski on Jun 3, 2010 4:45 PM EDT up reply actions

Or Median, perhaps.

by taco pal on Jun 3, 2010 5:00 PM EDT up reply actions

The problem is that it does not fit as a true Mode or Median, as most pitchers would have a Mode (most often appearing number) of 0, and if they pitch enough innings, even the Median (middle number) could be. I do not know what the best term for it is— I started out calling it median ERA (mERA), but Matt Swartz suggested the name Mulligan ERA, and I realized that it was in many ways more fitting, and thus re-named it, and changed the name in this post.

by dannijd on Jun 3, 2010 5:31 PM EDT up reply actions

I’m wondering if it might be worth doing qERA (quartile ERA) – take off the top 25% and bottom 25% of innings, and see if that middle 50% is more representative than overall ERA. That might be too much to cut off, though. It would allow for a smoothing over the course of the season, since each 4 IP would mean cutting 2 IP from somewhere in the season, so it wouldn’t be a set number of innings to cut.

Another thing that came to mind would be wERA (weighted ERA), where longer outings are counted more heavily than short outings, since a pitcher doing poorly usually gets yanked fairly quickly (see Kendrick’s 1.2 IP, 6 ER outing), which doesn’t give them an opportunity to recover. This is a really undeveloped idea, though, since I’m not sure how it would be done. I know overall ERA already does some of this, since it’s a per stat, but it would be a way to emphasize that a pitcher who can consistently go long innings is more valuable than a pitcher that sometimes can go 6 or 7 innings, but other times only lasts 2 or 3.

Honor is no substitute for victory.

by The Dark on Jun 4, 2010 8:05 AM EDT up reply actions

You would probably need to do an empirical study to determine how much to leave off, by correlating various versions of qERA with future ERA. Also, it wouldn’t necessarily be the same percentage at the top and bottom — consider that ERA has a lower boundary (0.00) but no upper limit.

by phatj on Jun 4, 2010 9:36 AM EDT up reply actions

I am in the process of trying to figure out how best to handlethis dilemma, as it is true that some percentage of innings is probably more determinative than just arbitrarilly taking two out top and bottom. Quartile is definitely too much, though, as even a halfway decent pitcher probably has many more innings of not allowing runs than of giving them up.

by dannijd on Jun 4, 2010 9:53 AM EDT up reply actions

Speaking of Happ is that guy ever going to come back. o thought he was going to have like an all star year this year until he got hurt. Anybody know any knews on his status or anything. Also really hope polanco plays in the up coming series. GO PHILS!!!!

by Phanatic43 on Jun 4, 2010 2:52 AM EDT reply actions

Per the news on MLB at Bat, Happ was scheduled to throw 2 innings in an extended spring training game yesterday. If he got through that without a setback, then he could make his first rehab start this coming Tuesday. I had high hopes for him too… I saw him pitch live twice last year, including his complete game shut out of the Rockies, and was looking forward to what he would do this year.

Also per at bat, Polanco and the team believe that he will return to action tonight. Provided that he is truly feeling better (they are not rushing him back before he is healed enough), I look forward to seeing him return.

by dannijd on Jun 4, 2010 9:33 AM EDT up reply actions

One last thought

You could almost use this as a model for pitcher’s consistency. Rather than the percentage change from ERA to mERA, if you take 1-(delta ERA:mERA), you get an idea of how consistent a pitcher is:

Halladay: 0.72
Hamels: 0.73
Kendrick: 0.77
Moyer: 0.70
Blanton: 0.64

At the time this was done, Blanton was obviously the least consistent pitcher, and Kendrick was consistently bad (based on his ERA/mERA and the low difference). Hamels and Halladay are both fairly consistent, and Moyer a little less so. If this was done across a larger sample size, it would be interesting to see how well it matches up with the conventional wisdom of who’s consistent and who’s not, and if it has any “predictive power” regarding injury (i.e. someone’s consistency plummets, and we find out later they were injured around that time).

Honor is no substitute for victory.

by The Dark on Jun 4, 2010 10:04 AM EDT reply actions