For an explanation of this series, please read the introductory post. The data is a hybrid projection of the ZIPS and Steamer systems with playing time determined through depth charts created by our team of authors. The rankings are based on aggregate projected WAR for each team at a given position.
Over the last couple of weeks, we’ve been going position by position around the sport. We finish up the series with bullpens today, but it’s worth noting that these projections follow a slightly different structure than the rest.
For one, projecting specific innings totals for relievers is a taller task than projecting playing time for position players or even innings totals for starters. There are numerous outside factors impacting bullpen usage, including things that we can’t really predict like the distribution of runs scored and allowed by each team. One team might play in a bunch of blowouts and rarely need their closer, while another could end up in a continuous stream of one run games and ask their best few arms to carry a lion’s share of the workload. Beyond that, the health of a team’s rotation is going to be a factor, as some relievers are also reserve starters who might be pressed into duty mid-season. And the depth charts are continually evolving, as injuries and acquisitions move guys into differing roles that come with different usage patterns.
So, for the relievers, we’ve simply assigned IP totals to each slot on a depth chart. Closers and primary setup men get 65 innings each, with the 3rd/4th relievers getting 55 innings each, and then the rest have their innings allocated in descending order according to their placement on the depth chart. And, in order to make each team’s total number of innings pitched (both starters and relievers) equal out to 1,458, we’ve added in a set for each team that makes up the missing innings in the projections. The performance projection is the same for each team, and is set to be around -0.1 WAR per 100 innings, on the assumption that the 10th or 11th reliever a team uses throughout the season is probably a little bit below replacement level. The statline in the table is just there as a placeholder – those numbers aren’t actually affecting the calculation beyond just setting innings equal and being included in the WAR sum.
Also, since we don’t have separate ZIPS/Steamer projections for guys as starters and relievers, guys who were projected as starters but are going to pitch in relief will likely be under-forecast. Aroldis Chapman, for instance, is getting his starter projections prorated to reliever innings totals, and he’ll almost certainly pitch better in relief than he was projected to do as a starter. There aren’t a lot of those types, but for guys like that, adjust their numbers upwards accordingly.
One final note: we’ve mentioned this on the other lists, but it is worth emphasizing here – the gap between many teams is so slim that you shouldn’t read too much into a team’s placement in the ordinal rank. The gap between #12 and #22 is +0.7 WAR. That’s no difference at all, really. There are good bullpens, okay bullpens, and a couple of bad bullpens, but don’t get too caught up in whether one team is a few spots ahead of another team. With margins this small, the specific placement on the list is mostly irrelevant.
On to the list.
Read the rest of this entry »
