31 July 2013

Is It Possible for a Bullpen to Be Too Rested?

Scrolling through the Orioles newsfeed the other day, an article written by Derek Wetmore on the team’s website caught my eye.  The article discussed the recent trend of Baltimore starters pitching deeper into games, which gives the relievers in the bullpen more time to rest.  It’s not breaking news to tell you that the overuse of a bullpen is a bad thing, but I did find it curious that the team would complain (complain is probably the wrong word to use here) about its bullpen receiving too much time off.    

Conventional wisdom states that overuse of a team’s bullpen can lead to relievers becoming ineffective, particularly during the home stretch of the long season.  The most recent example is the 2011 Atlanta Braves, who leaned on Craig Kimbrel, Jonny Venters, and Eric O’Flaherty heavily throughout the year (each appeared in at least 78 games).  This strategy worked well for Atlanta until September, when the team finished with a 9-18 record, and blew an 8.5 game lead in the wild card to the St. Louis Cardinals on quite possibly the greatest single day of baseball in history.  Kimbrel and Venters were a key part of that collapse, as Venters compiled a 5.11 ERA in the month, while Kimbrel had a 4.76 ERA along with 3 blown saves. Additionally, earlier this year, both O’Flaherty and Venters underwent Tommy John surgery.  While their heavy usage likely played a part in their September 2011 struggles, it’s unclear how much it contributed to Venters and O’Flaherty’s injuries (if at all), but it is worth noting.  

Back to the Orioles.  As manager Buck Showalter mentions in the article linked above, the line between too much rest and overuse is a delicate balance.  You don’t want to wear your guys down, but at the same time, baseball is like everything else in life, you have to practice to be good at it, and nothing can replace the experience of pitching in a live game.  At this point in the season, the starting rotation ranks 18th in innings pitched, while the relievers rank 12th.  There’s room for improvement, but they essentially seem to be in the middle of the pack.  The addition of Scott Feldman to the rotation in early July has certainly helped, as the rotation’s average innings per start in July is the highest it’s been all year.

They’ve been even better since the All-Star break, averaging 6.43 IP per start.  Combine the increased innings of the starting pitchers with the extra days off at the All-Star break, and you should have a very well rested bullpen.  But is that necessarily a good thing?

Let’s take a look at a graph that shows Baltimore’s bullpen ERA and FIP associated with the days of rest the relievers have received.  The number in parentheses included on the x-axis represent the innings pitched in those situations.  

In 2013 as a unit, the Orioles bullpen has performed best after 5 days of rest, while they have performed worst after 3 days of rest.  However, the 9.1 innings pitched with 5 days of rest is too small of a sample to confidently say that is the preferred amount of rest.  Additionally, the entire analysis over 4 months of the 2013 season may be too small of a sample.  With that in mind, let’s see how the Baltimore bullpen performance changed relative to days of rest in 2012 and 2011, and see if we can find any sort of pattern.

It’s interesting that both ERA and FIP spike upwards at 3 days of rest on all 3 graphs.  This would suggest that 0-2 days of rest would be optimal for bullpen performance. The varied results following 4+ days of rest could be a result of the small sample of innings pitched (innings pitched for 6+ days of rest are biased high, as explained below).

Admittedly, this analysis is very basic, and it’s important to remember that while these graphs tell us a lot, they don’t tell us everything.  First is that they only tell us how often a reliever throws in back to back games (zero days rest), but they don’t tell us how pitching with zero days rest multiple days in a row affects the results.  Additionally, they don’t tell us who is pitching these innings.  This is important because your best relievers are more likely to be pitching more often with less rest (and yielding better results), while the “mop up” guys will be pitching “mop up” innings, meaning your worst pitchers should normally be pitching with more rest.  Finally, since the number rest days only looks at major league days of rest, it includes innings from recent minor league call-ups who have probably pitched more recently than 6 days ago, therefore artificially inflating the innings pitched in that category.

As basic as this study was, it could easily be expanded to give more useful information.  It could include how pitching with zero days rest multiple days in a row affects results.  It could be expanded to include all teams over a longer period of time to give more confidence in the findings.  Individual pitchers could be looked at to see how each pitcher performs when given certain rest days.  Or you could even get into the details and see how rest days affect the velocity and movement of individual pitches, as Josh Kalk attempted to do for the Hardball Times.  It’s amazing how one simple question could lead to so many different studies. 

As for this study, it just reinforces the thought that Buck Showalter knows what he’s talking about.  But you probably knew that already.

29 July 2013

How Valuable Is Francisco Rodriguez in the Orioles' Bullpen?

A version of this article appears at Baltimore Sports and Life.

I wrote a version of this article a few days ago for Baltimore Sports and Life.  However, I wanted to share it here as well as my writing here will see an uptick over the upcoming playoff run for the Orioles.  So, with some moderate excitement, I am back in a deadline deal sort of way.

Tommy Hunter / credit: Keith Allison
Is there really anything more to say about the Orioles' acquisition of Francisco Rodriguez?  I certainly think so and it does not have to do with the pall cast over the deal with the known and alleged incidents of violence that are attached to the team's new relief pitcher.  No, I do not wish to discuss my perspective Rodriguez pleading guilty to attempted assault to avoid jail time when he attacked his at-the-time fiance’s father or about the time he attacked Mets’ bullpen coach during a game or the allegations of repeated abuse that the prosecution came up with or about the Milwaukee abuse case that was dropped when his at-the-time girlfriend (who was allegedly attacked) and the sole alleged witness (a housekeeper) went back to Venezuela and were no longer around for the trial.

What I want to discuss is how this deal rates as a baseball acquisition.  To what extent does Rodriguez improve this team.  What are best case scenarios in a general sense using simple numbers to see how big a deal acquiring Rodriguez was for the success of the team.  To do with, I want to assume the following:
  1. With 58 games left, there are about 25 innings left for the 7th inning righty, 8th inning righty, and middle relief / junk inning righty.
  2. Francisco Rodriguez is a 3.00 ERA pitcher (let’s assume he is really solid down the stretch).
  3. Tommy Hunter and Darren O’Day are 4.00 ERA pitchers (let’s assume they struggle down the stretch to the point that they are underperforming their current ERAs of 2.80 and 2.20, respectively).
  4. Jairo Asencio and whoever are 6.00 ERA pitchers (this is likely a pessimistic view as to what the default 4th righty in the pen could do).
So, here is what we get without Rodriguez using the above assumptions.

7th Tommy Hunter 25 4.00 11
8th Darren O’Day 25 4.00 11
Junk Jairo et al. 25 6.00 17


This is what happens with Francisco Rodriguez as the 8th inning righty, again, according to the set assumptions:

7th Darren O’Day 25 4.00 11
8th Francisco Rodriguez 25 3.00 8
Junk Tommy Hunter 25 4.00 11


These two tables note that the acquistion of Rodriguez saves the team nine run.  That is roughly worth one win based on standard sabermetric practice (that number jumps around year to year).  However, is it adequate to simply look at those nine runs and assume it is worth a win.  Are runs more or less important are the end of the game or in situations where these pitchers pitch to the point where grouping it all together makes for unfair comparisons?

The Leverage Index suggests that maybe we need to drill down a bit further.  To those of you who do not know, the Leverage Index calculates how important a situation is in a game based on the number of outs, baserunners, men on base, and the score.  A neutral scenario would have a Leverage Index of one.  A consequential plate appearance would be greater than one.  Here are the Leverage Indices of Orioles relievers when they enter a game:

Pitcher gmLI
Jim Johnson 1.81
Darren O’Day 1.42
Brian Matusz 1.15
Pedro Strop 1.12
Troy Patton 1.01
Tommy Hunter 0.95
Jairo Ascencio 0.89
Kevin Gausman 0.78
Steve Johnson 0.76
Luis Ayala 0.75
TJ McFarland 0.67
Alex Burnett 0.61

Those numbers pass the sniff test.  Johnson, O’Day, and Matusz are often placed in on the mound in tough situations.  Strop may seem like an outlier, but the team tried their best to keep him as a setup man and then 7th inning man until his performance found the team shipping him to the Cubs.  The point I am trying to make here is that even though the different between the two assumed righty sets above amount to 9 runs, 6 of those runs are saved in junk innings that normally would have gone to guys like Burnett, Johnson, Gausman, and, yes, Asencio.  Those scenarios were ones where the game was largely in the Orioles’ favor or their opponents’.  With that in mind, those runs given up in low leverage situations are not very meaningful because of the run differential in those game situations.

Instead, the important runs in the optimistic scenario that was detailed earlier in this post would be the focused on the three runs that do not score thanks to 3.00 ERA Rodriguez on the mound instead of 4.00 ERA O’Day in the eighth inning, but we need to determine how valuable those runs might be.  As mentioned earlier as well, a general appreciation of what a run is worth (about 1/9th of a win) would suggest that having Rodriguez is worth about a third of a win due to those three runs saved.  However, runs scored in high leverage situations are likely to be worth a different amount than runs scored in less important situations

Here is the score differential of the July outings when O’Day entered the game:

Dif Runs
July 24th 0 1
July 21st 2 0
July 19th 2 0
July 13th -4 0
July 11th 1 0
July 9th -3 1
July 7th 1 0
July 5th 1 0
July 3rd 0 0

Five (or 56%) of his July outings were games where the differential in score was 0 runs or the Orioles being up by 1.  If that stays the same then of those 25 8th innings to come, 14 of them will be games where the score is tied or the team is holding on by one run.  That difference now is about 2 runs (slightly less) in what you could expect between a 3.00 pitcher and a 4.00 pitcher.  That suggests than maybe the value of a third of a run more important, might be lower than that.

However, assume that for 12 of those innings pitched, Rodriguez and O’Day would be the same pitcher, but for the other two innings Rodriguez would give up no runs and O’Day would give up a run in each performance.  Well, how does that impact win probability.

Dif Win Probability
1 85
0 48
-1 14
Dif Win Probability
1 89
0 63
-1 30

Why are the probabilities different?  It almost all comes down to the home team having two innings to score while the visitor having only one.  If we look at the worst case scenarios (as a vistor going +1 to 0; as a home team going 0 to -1), we have decreases in win probability of 37% and 33%.  In other words, not letting runs score in the worst cases possible detailed above means a 3.00 pitcher is worth about 0.7 wins more than a 4.00 pitcher.  In other words, the runs saved in the 8th inning in high leverage conditions are roughly twice as important than they are in general based on the assumptions I have made here.

Where we are left is that under ideal circumstances, it is probably safe to assume that the addition of Rodriguez will benefit the team with an extra game in the win column, at most.  More likely though, his impact will be far less felt.  Differences in runs saved in high leverage innings will likely be rather inconsequential and applying Hunter to even lower leverage innings will likely be even more inconsequential.

In the end, the team gave up a second tier prospect in Nicky Delmonico.  A position player who was 100th ranked in Baseball America’s top 100 prospects has a 1 in 3 chance of being a useful starter in baseball.  At best, Delmonico was just south of 150th and probably no worse the 200th.  More than likely he will never make a dent on a MLB roster.  Even more important, Delmonico would have had zero impact on the Orioles in 2013 when they are competing for a playoff spot.  In other words, one would be pressed to call him a meaningful prospect.

However, Delmonico does not need to be a meaningful prospect to have value.  I’d suggest that his ability to augment a trade package is probably worth more than him being traded straight up for a reliever who at best makes the 2013 Orioles a +1 win team.  It is difficult to ever know what is possible in trades and it is an illness in the brain that us followers of the game tend to want to believe in an abundance of trade opportunities, but I do think that a package of Delmonico plus one would likely bring back to the Orioles a player of greater importance than Rodriguez.  Of course, this contention is somewhat unfair.  It is difficult for the addition of Rodriguez to stand up against a comparison of something that does not exist.  The vagueness of the unknown is certainly a draw for many to embrace and questioning Oriole front office authority has been a talent that has been thoroughly developed over the past decade and a half.  That said, I maintain that the health of the franchise is better served when second tier prospects are stacked instead of being doled out one at a time.