keyboard_arrow_uptop

Deserved Run Average (DRA) is, as you know, the new pitcher-value metric here at Baseball Prospectus. DRA reflects our best estimate of the runs each pitcher “deserved” to allow in a given season. The number of deserved runs above or below average is used to calculate BP’s pitcher Wins Above Replacement Player (WARP).

By popular demand, we’ve now extracted the estimated effect, in runs, that various external factors have had on individual pitchers. You can get a direct link from the main site by clicking “DRA Run Values” from the Statistics pull-down menu, or by clicking here.

Overall feedback on DRA has been very positive, and we expect these run values will provide even more insight into the challenges various pitchers have faced. In other words, rather than simply knowing that DRA takes into account things like score differential and temperature, you can now see for yourself how those components have affected (or not affected) individual pitchers.

From the Run Values page, you’ll learn that Zack Greinke has benefited from terrific catcher framing, but also dealt with poor defense. You’ll see that his teammate, Clayton Kershaw, has lost about nine total runs from factors beyond his control, and that Dallas Keuchel has grabbed six free runs, largely from the parks in which he has pitched. You’ll also see a glossary on top of the entire table that briefly explains each column.

There are always caveats. Here are some.

First, this table does not contain all of the external factors that DRA considers, just the ones most likely to be of interest. The table is a bit crowded as it is, and we want to keep things readable. The “adjusted runs” column at the far right gives you the total effect of the factors in the other columns in the table. Are there other adjustments accounted for in the model? Yes. If one of them interests you particularly, feel free to ask below.

Second, don’t be surprised if some component run estimates diverge from other sites’ predictions. Remember that DRA calculates its own estimates for each factor, on a season-by-season basis, in light of the other considered factors. Thus, DRA doesn’t calculate generic park factors; rather, it calculates park factors that are adjusted for the temperature of the games played, the size of the strike zone there, and the quality of the teams that visited, among many other things. This doesn’t mean you can’t have your own opinion on these issues, or appreciate other perspectives, but it does mean that if DRA disagrees with somebody else about the effect of some component, it probably has a (uniquely) good reason for doing so.

Third, although we describe these figures as the “DRA Run Values,” they technically are the “value” mixed-model results from the first step of DRA. Those who enjoy the details may recall that we typically subject the value-model derivations to a second step, the MARS model, to give each season’s model the opportunity to consider basestealing prevention and other factors.

For this exercise, doing so is more trouble than it is worth, because the only factors DRA’s 2015 model selected as relevant were (1) pitcher value, (2) volume of batters faced, and (3) the starting-pitcher percentage. None of the baserunning stats were deemed relevant to assessing league-wide pitcher run prevention for 2015 (or 2014, for that matter). Thus, the MARS model in recent years largely just filters out the effects of plate appearances and converts the raw-value numbers to runs allowed per nine innings (RA/9). Those are both important, but neither is helpful when simply looking at total pitcher run value.

Lastly, for now we’ve posted run values for pitchers from the 1998 through 2015. We’ll push it back eventually to past seasons, and it’s not particularly difficult to do so, but we wanted to get you the most recent ones now.

As always, let us know what you think. We think you’ll have fun with these and get a much better sense of which areas constitute team strengths or weaknesses, and how it is that pitchers—even pitchers on the same team—can end up facing very different challenges.

Thank you for reading

This is a free article. If you enjoyed it, consider subscribing to Baseball Prospectus. Subscriptions support ongoing public baseball research and analysis in an increasingly proprietary environment.

Subscribe now
You need to be logged in to comment. Login or Subscribe
brownsugar
8/26
I look forward to diving in. Let's start with a question. Per the DRA leaderboard this year, Zack Greinke leads with 22.7 RAA. If I use the rule of thumb that 10 runs = 1 win, that means he's somewhere around 2.5 wins above average. When I look at his player card, he is sitting at 6.1 WARP. This implies that a league-average pitcher with his number of innings is worth 3.6 wins (6.1 - 2.5 = 3.6). Is that accurate, or does the 10 runs = 1 win rule of thumb not apply here?
mcquown
8/26
It's 9.2 Runs/win in 2015. Replacement RA for both pitchers is 4.97. This column can be added to the sortable reports for pitchers (full season reports only, not by-team).
brownsugar
8/26
Good information, and thank you, but my central query remains unanswered. Using 9.2 runs/win, Greinke's 23.41 RAA equates to 2.54 WAA. Per Greinke's player card, he has accumulated 6.0 WARP in 172 innings. Using these two win values, a league-average starting pitcher throwing 172 innings is worth 3.5 wins (6.0 - 2.5 = 3.5). So my question is: is a league-average starting pitcher worth 3.5 WARP with a month to go in the season?
mcquown
8/27
Yessir, I didn't mean that as a complete answer. We have definitely been discussing this. For now, WARP (taken from the sortable reports) is being displayed in place of RAA.
bachlaw
8/26
Hi Randy, WARP is wins above replacement level; RAA is runs above average. So, it's a matter of a different reference point.
brownsugar
8/26
I understand that, but using Greinke as the example again he's currently worth about 2.5 wins above average (if 10 RAA = 1 win) and 6 wins above replacement. That is a 3.5 win gap between value above average and value above replacement, and implies that a league average starting pitcher is worth about 4 WARP over the course of a full season. Is that an accurate understanding?
bachlaw
8/26
Ok, I see what you're saying. We'll take a look. Thanks for writing in.
jfcross
8/27
It could be that since pitching performance gets shrunk/regressed in the mixed model (where pitching is a random effect) it then get essentially unshrunk in the MARS model that predicts RE24 from RAA and other factors. So, while there's only a 25 run difference between Greinke and an average pitcher in RAA there's something like a 40 run difference in their RE24 predictions and thus their WARP's (average full time pitchers are checking in at about 1.5 WAR).
brownsugar
8/26
Because Red Sox aside, no teams are paying league average starting pitchers a salary equivalent to a 4-win player.
philly604
8/26
I have a question about the glossary definition of FRAA. Maybe I'm reading it wrong, but it's not clear to me. It states: FRAA - The total runs lost from good or bad defense (created by averaging the total FRAA/PA of the defenders on the field for each play). Is the FRAA that is being used generated specifically by the on field action during that PA or is it the general defensive performance for the 9 fielders over the course of the season scaled down to the specific PA?
bachlaw
8/26
Excellent question. DRA uses the defender's FRAA for the entire season divided by the number of events they've participate in. So, FRAA/PA. As we learn more about the defender's play each year then the context of every event is adjusted accordingly.
Nantrin
8/26
Would something like GCAA be useful in this context? Are some pitchers helped/harmed by game calling in a way that moves the needle?
bachlaw
8/26
Stay tuned. :-)
harold
8/26
What is the baseline that the 'Adj Runs' is applied to? I ask in regard to the situation between Francisco Liriano and Gerrit Cole. By most metrics (RA, ERA, FIP, bWAR, fWAR), Cole has pitched better than Liriano this season. Both WAR models have Cole at least one Win ahead of Liriano. DRA is an outlier, as it has Liriano ahead of Cole. I expected that meant that the DRA model is able to flesh out differences in the environment between the two pitchers such that Cole was either penalized more or helped less than Liriano. But looking at the breakdown of the DRA Run Values, Liriano has actually been penalized slightly more than Cole (-14.21 vs -13.52). So that seems to imply to me that they have different baselines from where the 'Adj Runs' is applied (and Liriano's baseline is higher than Cole's) in order for Liriano to have more value despite facing few batters and having worst adjustments. I think it would be good to also publish the un-adjusted baseline runs number for each pitcher (which uses base-out run expectancy instead of raw runs, right?) to provide more context.
Nantrin
8/26
Wait. Is the first number a starting value, that is then adjusted by the others? Or is it the final value that comes after the other adjustments have been applied?
TangoTiger1
8/27
It looks to me that WARP is based on DRA. And RAA is based on a REGRESSED DRA. His DRA is 2.05, which is 2.92 runs per 9IP better than replacement. 2.92/9*172 = +56 runs better than average. Divide by 9.2, and you get 6.1 WARP. But in no way is 2.05 only +23 runs better than average. 2.05 is 35 or 40 runs better than average. However, the hidden thing that no one is talking about is that the data is being regressed. This is huge. So, they are regressing someone with 700 or whatever PA at 30 or 40%, and so, bringing down the +35 down to +23.
TangoTiger1
8/27
The second paragraph should read "better than replacement".
bachlaw
8/27
I think the bottom line is that the RAA column was mislabeled. It's the regressed output of the value equation and not the same as the RAA used for PWARP, which is further down in the process. So, it's confusing for that reason. It also unduly confuses people as to whether they need to account for the first column when looking at the rest of them (they don't). The only reason I added the RAA column was as a means of sorting the pitchers by ultimate total value; essentially it was a proxy for PWARP ranking. But all it is doing is confusing people, so I expect we'll find some other means of sorting the table very shortly. I'll post another comment once we have done that. As I noted, the remainder of the columns are independent of the current "RAA" column and you can use them without regard to it. Thanks to everybody for pointing this out. As usual, our readers are great.
bachlaw
8/27
The table is now sorted by WARP (DRA-based). Hopefully this makes much more sense.
brownsugar
8/27
Thanks for diving into it gents. The regression explanation makes sense; helps me feel like I've got a better sense of what I'm looking at while I poke around the leaderboard.
bachlaw
8/28
We've now changed the "bat_home" column to "Home/Away," which is actually plain English. The column shows which pitchers have been particularly fortunate (or unfortunate) in drawing home versus away games. The glossary at the top of the DRA Runs page has been updated accordingly.
bachlaw
9/04
we've now adding Opposing Batter (Opp Bat) runs, to tell you the strength of the batters each pitcher faced. We also consolidated the "LHB" and "RHB" park factors into one "Stadium" factor, and removed "Innings," which nobody seemed to find useful. Finally, Home / Away now includes not only the general home-field advantage, but also the way home field affects team defense and the benefits a home team derives from batting in the bottom frame of innings.