Tuesday, October 30, 2018

SHREDDING THE BULLPEN: A WORLD SERIES SUCCESS STORY

Following up on our discussion of what happened to the Astros in their ALCS matchup with the 2018 World Champion Boston Red Sox, we thought it might be worthwhile to see if playoff teams--and World Series winners in particular--showed a pattern of hitting well against relief pitchers. (You should know that the Dodgers, the first losers in two consecutive World Series since the Rangers did in 2010-11, had their bullpen shredded by both of the teams that beat them--the Astros in 2017, and the Red Sox this year...the Dodger relievers posted a 5.48 ERA in the just-completed Fall Classic.)

Thinking more globally, we went back to the data at Forman et fils (baseball-reference.com) to look at the overall performance vs. relievers over the past decade. (That gives us nine years of data to work with--a total of 270 data points--just shy of what's needed to win twelve slightly used Dodger Blue™ cupcakes.) Do teams that make the playoffs hit better than average against relievers? And do teams that win the World Series exceed the average of "garden variety" post-season teams?

The answer to both these questions is "yes." The table at right breaks it all out for you. Teams that
made the post season have their OPS+ vs. relievers displayed in bold type. Teams with a 120 "sOPS+" (Sean's acronym, not ours!) are shown in scalding orange; we've also color-coded teams with 110-119 (pale orange), 100-109 (yellow), and--on the opposite side of the spectrum--teams whose "sOPS+" is less than 85 (pale, pale blue).

At the bottom of the chart you have some averages--these are yearly "sOPS+" averages for the playoff teams. As you can see, the figures are uniformly above league average: for the nine years in question, the average "sOPS+" is 107.

Finally, note the cells with the double-thick lines around them. These are the World Series winners. And, yes, the World Series winners (as seen in the double-thick-lined box at the very bottom right of the table) are better yet on average than their post-season also-rans. Over the past nine years, World Series winners have an aggregate "sOPS+" of 112 vs. relievers.

Now, doing well in this statistic doesn't guarantee you a trip to the post-season; after all, it's only one component of team performance. There are many examples of teams doing well in this statistic who didn't make it to the playoffs at all. But if you do make it, having an offense that is able to do damage against the opposition's bullpen seems to give you a measurable advantage with respect to winning the World Series.

(And to complete another historical tidbit that was given a teaser above: teams that lost consecutive appearances in the World Series include not only the 2017-18 Dodgers and 2010-11 Rangers, but the 1991-92 Braves, the 1977-78 Dodgers, the 1963-64 Yankees, the 1952-53 Dodgers, the 1936-37 Giants, the 1923-24 Giants, the 1921-22 Yankees, and two teams--the 1911-12 Giants and the 1907-09 Detroit Tigers, who are the only teams to lose three World Series in a row. By doing it this year and last, however, the Dodgers have joined the Giants as the only teams to have three instances of "two-time loser" syndrome in the World Series. To match their Bay Area rivals, they'll need to make it back to the World Series next year--and hit the skids again...)

Friday, October 19, 2018

REGARDLESS...

...of your opinion about the call that clearly affected the outcome of Game 4 in the ALCS between the Red Sox and the Astros, there is one incontrovertible (and ironic) fact.

The Astros' bullpen, which had posted only one sub-par month (July, 5.19 ERA) during a season of exemplary achievement, picked a most unfortunate time to regress, giving serious ground in three games during the ALCS. Their overall ERA for the series (5.79) was actually better than their overall performance.

The Red Sox bullpen, considered suspect by many, managed to bend but not break during the series--and that made all the difference.

Peeking out from the stats is the fact that both pitching staffs were having trouble with their control. Red Sox pitchers averaged 5.09 BB/9 during the series, which looks a lot more like 1949 than 2018. The Astros were better (four walks per 9 IP), but this is still well above the regular season MLB average.

Thursday, October 18, 2018

POST-SEASON SNAPSHOT: A SIGH OF RELIEF FOR THE BULLPEN

So, OK, this is not really a "post-season" snapshot. There are no stats on post-season bullpen performance in this post.

What we do have, however, is a meditation on the changing perspective on the bullpen and its strategic importance for success that reverts back to more straightforward stats in order to capture those changes.

No one needs to be reminded that relief pitching is undergoing a transformation--the Tampa Bay Rays have made sure of that. We can expect more relief innings over the next couple of years as other teams attempt to emulate their "opener/delayed-starter-in-relief" strategy that was seemingly such a success.

But the value modeling that sabermetrics has imposed upon the game doesn't see it that way. Those numbers suggest that relievers did less to help their teams win games in 2018 than was the case in the previous seasons. Those modeling stats presume a different reality than what people see when they watch an individual game. And they tell us, year in and year out, that relief pitching has a net negative value in the overall model. This season, relief pitching had its most negative overall value according to Wins Above Average than has been the case in nearly half a century.

Is this hard to believe? Not for some. We find it hard to imagine, however, that teams whose bullpens post similar ERAs over a season can have significantly different WAR values. Of course, ERA has been "proven" problematic at the individual pitcher level by the recent attempts to use batter vs. pitcher (BvP) stats the go-to measure; but individual pitchers are not the only measurement aspect that we need to define and evaluate.

In fact, with the increase in reliever innings, it actually becomes more important to develop better aggregate measures for overall team performance in this area. And part of that effort should be to more tightly relate it to actual wins and losses.

And that's what we can at least start to do...by putting together these various measures in scatter chart relationships. The ERA+/WAA scatter correlation shows a lot of discrepancy in the -2.5 WAA range, with ERA+ values being all over the chart. Some high-achieving WAA teams are actually have sub-par ERA+ values.

How, then, is this any real advance over a ERA+/WPCT scatter correlation? While there are always teams who "beat" their WPCT projections based on their ERA+ (due to the fact that such teams give up extra runs in games already lost.

The solid correlation in the 90-110 ERA+ region with WPCT demonstrates that there's general set of principles that remain in operation in the mid-range of the distribution, but that it frays a bit at the extremes. That's a more natural set of relationships based on real-life game situations. It suggests that there's more meaning in reliever WPCT than has been claimed for the past twenty years.

And the game is changing in ways that will likely reinforce this. When we look at the final month of the 2018 season, we see several interesting aspects of how this is manifesting itself.

In September 2018, we can see that certain teams experienced "make-or-break" months with respect to the post-season in terms of bullpen performance. The table at left sorts relief pitching in descending order of ERA.

Looking at it, we can see how one team (the Brewers) clearly rode their bullpen performance into the post-season.

And we can see how two teams (the Cardinals and Diamondbacks) wound up falling short of playoff appearances due to the poor performance of their reliever in the final month. (The third team coded in green, the Mariners, were caught and passed earlier in the year by the A's, who rode their bullpen into the playoffs.)

The color coding here is of some interest. Seven of the teams with the best performance from their bullpens (seven of the twelve with better-than-average ERAs) wound up in the post-season. Three of the top four teams in September bullpen performance are still competing in the post-season at this time (Brewers, Dodgers, Astros--only the Red Sox had a subpar performance from their relievers in September, and they managed to keep a lid on things in enough of their appearances to generate more relief wins than losses).

Relief pitching is not broken out sufficiently in the otherwise overly-parsed situational data for us to know why the Braves could go 9-3 with a 5.05 ERA, but we can make an educated guess: their mop-up relievers in already lost games gave up a lot of runs. A team like the Indians (who suffered a virtually complete reversal in bullpen performance in 2018 after a fine season the year before) managed to blow leads at crucial moments, saddling themselves with losses, but they did not pitch poorly in already lost games.

What we can tell you is that the playoff teams in 2018 posted an aggregate 65-34 record in September games where decisions were picked up by relief pitchers. We'll let you decide if you think that is as meaningless as many still seem to think is the case.

Sunday, October 7, 2018

VITAL SIGNS: END OF SEASON

Ah, the September song. Is it a preview of "coming attractions" as regards offense?

Joe P., who felt the urge a couple months back to double down on the "take and rake" offense, would doubtless point to the fact that run scoring levels are still relatively robust (4.45 per game in 2018, and 4.44 in September) and dismiss the dip in batting average (BA) for the month (.243) as being meaningless. (After all, batting average is meaningless, n'est-ce pas? So long as isolated power (ISO) can remain at all-time highs, offense can remain "robust enough.")

But such is not going to be the case if one other factor continues to follow its trend line. The rise in strikeouts--more specifically, the rise in the percentage of strikeouts that occur in plate appearance where a batter has two strikes on him--will at some point have a cratering effect on batting average, which will domino in to on-base percentage (OBP) and slugging average (SLG).

Our chart at right shows the K-to-2-strike percentage as it's evolved from 1988 to 2018 (these are the only years where the play-by-play data is detailed enough to capture this info). As you can see, this percentage is slowly but inexorably on the rise and has risen by 30% over thirty years--with the fastest rate of gain occurring in the past decade after a long lull due to the "baked-in" effects of the offensive explosion.

Additionally, BA and tOPS+ values for two-strike situations have been decaying over this time frame. (The tOPS+ stat measures the OPS value of the two-strike PA against the overall OPS. This was close to fifty percent in 1988 and remained relatively constant throughout the explosion until 2009; now, however, that figure has moved down into the low forties. BA is declining along with it, with two-strike PAs reaching a new low this year (.173).

The direction of the trend lines means that the so-called "smart adjustment" that Joe P. touts (admirably defending the tenets of an increasingly senile sabermetrics) is pushing everything into perilous territory. Not only is the game becoming more two-dimensional offensively, but it's skating into a risky region where additional pitcher adjustments will bring BA down to levels seen in two months during 2018 (September: .243, June: .245) for the entire season.

Note that HR totals did recede this year (from the absurd 1.27 in 2017 back down to 1.15). But keep in mind that such levels have to at least be sustained in order to keep offense "robust enough." Pitching adjustments, in two forms--experiments with in-game pitcher usage, and analyses to counteract the "launch angle" phenomenon that was partially responsible for the HR spike--are beginning to make themselves felt. It's unlikely that hitters are going to adjust to such alterations by pitchers in a short period of time--leaving it highly likely that home runs will drop and strikeouts will continue to rise...

...Which will result in batting averages that look disturbingly similar to what we saw in the mid-to-late 1960s.

How far can HRs drop? That's harder to predict until we see more evidence of pitching staffs improving. In 2018, we had an unusually high number of really bad teams and really good teams. Several of the really bad teams barely improved their HR allowed rates, while the really good teams showed a higher rate of improvement. When such improvement becomes more uniform, it will begin to effect teams that managed to improve their HRs hit in 2018 and a more pronounced decline will set in.

As you can see in the final comparison chart for R, HR, and BB (each month in 2018 is compared with the R/G, HR/G and BB/G from its corresponding month in 2017), the decline here was consistent but relatively uniform. (BB/G has a frequent pattern of being higher in April and September, due to weather and/or roster issues...as you can see, June was the biggest outlier, but that's because the HR rate was simply insane in June 2017 and it drove R/G up toward "offensive explosion" levels.)

A uniform year, such as was the case in 2018, is often followed by a more jagged change in the following years.

Next year we'll run these numbers for two parallel years, showing the months of 2019 against their analogous months for 2018 and 2017. And we'll be back a bit later this month with a look at big swings in BA, OBP and SLG at the league level over the history of baseball. Stay tuned...