Better bullpen should lead to better results for A's in 2016
Let's be clear about what's being said here: The A's aren't anyone's division favorites. Not now, and not in March, I'm going to guess. The Astros are no longer up and coming; they're up and they've come. The Rangers are coming off a huge bounce-back season, the Mariners are surrounding their core with more depth, and the Angels have baseball's best player. The AL West is going to be tough, and the A's probably have the most to prove.
But the A's were supposed to be competitive in 2015. Not great, but competitive. That didn't work out, and they've made some moves I'm sure they'd love to undo. Yet there's a road to being competitive again in 2016. Understanding it requires some understanding of what, exactly, went wrong in the past season. It's always complicated, for any team, but for the 2015 A's, the story might be the least complicated possible.
I started thinking about this after receiving an email Thursday that referred to something I wrote in July. At that point, I called the A's the unluckiest* team of the millennium. The asterisk serves a purpose -- I don't actually like the word "luck" in baseball. What happens happens. There's just not a great list of alternative terms. Anyway, this relies upon a metric called BaseRuns. I don't want to get too math-y, but BaseRuns is intended to calculate an expected team winning percentage, based on performance. What gets stripped away is the element of timing. Timing is critical, of course, but it's not always predictive. Team records tend to follow team BaseRuns records. Deviations are common over the short term, but they'll mostly balance out given time.
I've got team BaseRuns data stretching back to 2002. Which means I have actual team records, and BaseRuns team records. Below, see the teams with the greatest differences between those records, where the actual record was worse than expected:
Right there, you see the 2015 A's. Instead of winning 80 games, they really won just 68 games, giving them the biggest difference in this direction over at least the past 14 seasons. Now, you might notice in that plot that there are five teams from last season alone. That's certainly interesting, and maybe it means something, but I can't think of a good reason why BaseRuns might've suddenly become obsolete overnight. So for now I have to think it's a fluke. By this measure, the A's should've won more games. By this measure, the A's mostly played like a .500 ballclub.
What do we make of differences between actual records and BaseRuns records? Here's how they've held up, year over year. On the x axis, the difference in a given season; on the y axis, the difference for the same teams in the next season.
It's a lot of noise. You don't need to understand much about basic regression to see that. Whatever relationship there is is weak. Take the 20 least fortunate teams. In the first year, they averaged a -7.5 win difference between their real records and their BaseRuns records. The next year, those same teams averaged -1.2. At the other end, the 20 most fortunate teams went from a +8.9 win difference the first year to a +1.4 win difference the next year. Things tend to return more or less to normal. The A's should be encouraged by what BaseRuns says about what they did.
But I'll tell you why I don't like the word "luck" -- the players are responsible for what happens. Bad timing might not be sustainable, but it still takes place, and you can't just wash your hands of that. The biggest problem, by far, for last year's A's? The bullpen, for months, was a disaster. Reasonably talented relievers made the wrong pitches at the wrong times, and the team lost countless heartbreakers. That helps to explain why the record and the BaseRuns record don't match up.
I need to show you another weird statistic. This one's called Win Probability Added, or WPA. It's a hell of a story-telling statistic. At the heart of it: For any given game situation, we can calculate a team's odds of winning. So then we can track those changes in odds, over events. Say you have a team with a 50% chance of winning. Then a player hits a single, and the chance goes up to 54%. The WPA of the single was +4%. It makes sense for each individual event, and then you can add a bunch of them up. I decided to look at full bullpen WPA. It shows the effect team bullpens had on the team's chances of winning.
Here are the five worst bullpen WPA marks from last season.
The A's were worst, and by a big margin. The difference between 30th and 29th was the same as the difference between 29th and 22nd. The league-average bullpen had a WPA of +1.9, implying it was about 10 wins better than Oakland's. Late-inning meltdowns crippled the A's. They started almost immediately, and the team could never recover.
Lowest Bullpen WPA, 2015
| Team | Bullpen WPA |
| Athletics | -7.8 |
| Braves | -3.9 |
| Marlins | -3.2 |
| Blue Jays | -1.8 |
| Dodgers | -1.7 |
That's just 2015. We have WPA data stretching back to 1974. Here now are the 10 worst bullpen WPA marks since then:
The A's aren't worst anymore, but they're still close -- now they're tied for seventh-worst, out of a sample of 1,168 seasons. The point here is just to say the bullpen was historically ineffective. Some of it was bad timing, and some of it was legitimately bad pitching. It wasn't the only thing that went wrong, but it was the biggest problem for the first few months, and then trades and injuries depleted the talent level. The season went awry mostly because of the bullpen. Once it was a write-off, the front office had to act as such.
Lowest Bullpen WPA, 1974 - 2015
| Season | Team | Bullpen WPA |
| 1999 | Royals | -10.9 |
| 1990 | Braves | -10.3 |
| 1979 | Padres | -8.8 |
| 1978 | Mets | -8.6 |
| 2010 | Diamondbacks | -8.4 |
| 1993 | Mets | -8.3 |
| 2007 | Devil Rays | -7.8 |
| 2015 | Athletics | -7.8 |
| 2013 | Astros | -7.6 |
| 1991 | Astros | -7.5 |
Now, most of what we've done is talk about last year's A's. Last year's A's aren't next year's A's, but last year's A's were a half-decent team sunk by the bullpen. And now look at where the A's are, and what they're doing. Most importantly, by far, the 2016 A's should have a healthy and effective Sean Doolittle, who missed most of last season due to injury. He looked healthy again toward the end, and the last time Doolittle was a regularly healthy pitcher, he was a dominant pitcher, one of the top relievers in baseball.
Beyond that, in a quiet move, the A's exchanged Jesse Chavez for Liam Hendriks. Hendriks isn't a well-known reliever, because of his background as a mediocre starter, but last year he started throwing 95 mph, and he finished with more than six strikeouts for every walk. Most of Hendriks' profile indicates that he's a very solid reliever for the short term.
There are also reports that the A's are heavily involved in talks with free-agent reliever Ryan Madson. Injuries cost Madson three years, but last year he again looked like himself, averaging better than 94 mph. The A's could add him to the group, and even if they don't, the interest suggests they'd go get another reliever instead.
And there are a few interesting holdovers. Fernando Rodriguez made himself useful last season. Evan Scribner somehow finished with four walks and 64 strikeouts. It isn't yet all settled, but from the looks of things, the A's should have a much stronger relief group in the season ahead. And though last year's A's also thought they'd have a healthy Doolittle, that was just one of their bad breaks. If those bad breaks don't repeat, the A's should have reliable late-inning relief, and then they get to avoid last season's catastrophe.
There's other stuff that would need to happen for the A's to get into the playoffs. They don't have a true standout position player, and I'm not sure anyone knows what to expect from the Rich Hill roll of the dice. At this point, the A's definitely don't look special. But they do look like they could be good enough. And while that might sound surprising, it's amazing the difference it can make to not have to keep going to a bullpen nightmare. Fix that, and so much of everything else looks an awful lot more sunny.
