Yesterday, I got to talking to former MLB pitcher Danny Graves on Twitter. He’s part of the team at the newly-launched 120Sports, a brand new sports network. It’s a cool concept, focusing on video analysis that is typically about 2 minutes long for each story, getting you up-to-the-minute updates on everything that’s going on.
Anyway, I asked Danny about his unusual career:
@HighHeatStats I liked to pitch to contact. Unless runner on 3rd w/ less then 2 outs. Formula was to try and throw max, 3 pitches per batter
— Danny Graves (@dgravy32) July 7, 2014
He did, in fact, average 3.41 pitches per plate appearance over his career, during which time MLB had an average of 3.73. That’s an 8% difference.
If we focus down on just one year, we can see a bit more detail. Take 2000, when Danny averaged 3.65 pitches per batter. He was the Reds’ closer that year, and one of 24 pitchers to save at least 20 games that season. The other 23 closers averaged 3.85 pitches per plate appearance, about 5% more than Graves. In fact, John Rocker, Armando Benitez, Troy Percival, and Dave Veres all averaged over 4 pitches per batter, while only 4 of the closers averaged fewer pitches than Danny (Antonio Alfonseca, Jeff Shaw, Jose Jimenez, and Steve Karsay).
So, it’s clear that Danny’s assertion that he tried to pitch to contact and limit pitches per batter is correct. This had another effect, though, which was first pointed out to me yesterday by Adam (@baseballtwit). Throwing fewer pitches per batter allowed Graves to face more batters and pitch more innings. Looking again at 2000, Graves threw the most pitches among those 24 closers. He tossed 1418 times against 388 batters. Compare that to Benitez, who tossed 1313 times against 304 batters. That’s only 105 fewer pitches, but against 84 fewer batters, which explains why Benitez averaged about 2/3 of a pitch more per batter.
In that 2000 season, the 388 batters that Graves faced produced 274 outs, meaning that he pitched 91.1 innings, and these were spread over 66 games. Benitez (who, remember, threw nearly as many pitches as Graves), produced only 234 outs. His 78 IP were spread over only 77 games. Benitez was the very model of the modern closer, tossing exactly one inning per appearance just about as surely as the sun came up every day. Graves, though, was used frequently in the 8th inning to record a 4, 5, or even 6-out save.
Did Graves’ desire to pitch to contact lead to more hits? Not in 2000, no. He allowed 7.98 H/9, whereas the other 23 closers allowed 8.06 H/9, and NL-wide average in 2000 was 9.20. So Graves was similar to other closers, who as a group were a lot better than league average. It did, however, affect his walks. In 2000, he issued 4.14 BB/9, as compared to 3.45 BB/9 from the other 23 closers. And if you throw out John Rocker, who issued 48 walks in 53 innings himself, the other 22 guys averaged 3.29 BB/9, putting Graves 25% above their average. This had a dramatic effect on his FIP, which was 4.59 as compared to his actual ERA of 2.56. That difference of 2.03 was by far the largest of the 24 closers, with the other 23 averaging a 0.44 difference (3.35 ERA, 3.79 FIP).
What, then, was Graves’ formula for success? For starters, in 2000, only 19% of the balls put in play vs. him, were line drives, as compared to the NL average of 25%. This is consistent with the notion that he tried to pitch to contact, presumably trying to be on the edge of the strike zone. This would tend to lead to poorer contact, but also more walks, both of which were true for Danny. He also induced double plays in 13% of opportunities, well above the NL average of 10%. The lack of good contact by batters also limited his extra base hits allowed to just 28% of total hits, as compared to 35% NL-wide and 30% among the other 23 closers.
It’s interesting that come 2001, Graves dropped his walk rate considerably, from 4.1 per 9 in 2000 to 2.0 in 2001 and 2.3 in 2002. The cost? More hits. His strike percentage jumped from 59.4% in 2000 to 68.1% in 2001 and 67.7% in 2002. As he was around the strike zone more, batters made better contact, with their line drive percentage climbing to 24% each season, and his H/9 rose from 8.0 in 2000 to 9.3 in 2001 and 9.0 in 2000. Thanks to the reduction in walks, his FIP over 2001-2 was 3.66, much closer to his actual ERA of 3.62, as opposed to the huge gap in 2000.
In his final years of 2004-2006, Graves was used like a modern reliever, averaging exactly 1 inning per appearance (121 games, 121 IP). He kept his walk rate down at 2.8 per 9, but his K rate stayed low, at 4.7 per 9. His extra-base hit percentage soared to 38% as a lot of balls left the yard (11.6% of his fly balls allowed were homers, well above league average). I’m not sure why Danny’s approach changed. In might have been out of necessity, but clearly as he threw more strikes, he was hit harder. Issuing fewer walks did not benefit him, and in fact led to more hits and a higher overall WHIP.
@HighHeatStats BB bother me more now that I don’t play. When I was playing, I didn’t mind BB. Confident I could get ground balls.
— Danny Graves (@dgravy32) July 7, 2014
As Adam pointed out to me, Graves was pretty much the last MLB closer who consistently averaged over 1 inning pitched per appearance. In his prime, he wasn’t afraid of walking people and made up for it by getting double plays and not allowing extra-base hits very often. By the end of his career, when he pitched more like a modern closer (strictly 1 inning pitched, low walk rate) he was actually less successful. Maybe this is a sign that today’s relief pitching isn’t as optimized as many folks think.
Andy, that’s a condemnation of fip as a concept in general! fangraphs beware 😉
I don’t understand “pitching to contact” as a phrase. If you’re working the corners, you’re not looking to get hit. “going right at him” is throwing strikes. It seems like a better phrase would be “working the corners” or “nibbling” rather than “pitching to contact” which implies to me not wasting time with balls the batter will take and instead giving him something he’ll likely swing at almost all the time.
Yeah, that’s reasonable. I mean, I don’t think Graves means he is pitching with the hopes of the batter making SOLID contact. I’m sure he could’ve grooved every pitch at 80 MPH down the middle if he wanted. He means pitching to WEAK contact, which is achieved, as you say, by pitching to the corners. Nibbling, to me, is something else–it’s throwing pitches near the edge of the strike zone that the pitchers hope are called strikes.
I did a whole analysis on glavine here but the computer gods ate it so I’ll try to refresh more concisely.
Glavine, a famous “working the corners” guy was extremely consistent at out-performing his FIP. Here’s ERA-FIP (negative is unexpectedly good) for his post-strike years…
1995 -0.41
1996 -0.51
1997 -1
1998 -1.03
1999 0.05
2000 -0.63
2001 -1.2
2002 -1.23
2003 -0.22
2004 -0.64
2005 -0.14
2006 -0.48
2007 -0.41
2008 -0.48
That’s an average of -0.60 and a standard deviation of a scant 0.39. You can say good defense but the braves were not really that strong up the middle and glavine threw a lot of grounders. His stud outfield defenders were far less valuable to him than say Smoltz. Also, you see no discernible change when he went to the Mets. I point this out for consistency sake. FIP is clearly not seeing the pitcher that he was. Glavin was not exceptional at avoiding the walk and his whip is fairly high for the ERA that he managed during those years. ’99 sticks out a bit, but look at BABIP. He averaged .285 over the period and 99 had by far the highest at .318. He clearly had some bad luck that year having his ERA merely at his FIP instead of well below.
I bring this up as supporting evidence. Telling a pitcher to throw more strikes may indeed be bad advice on occasion. Lowering your BB/9 can make you an inferior pitcher if you don’t have swing and miss stuff. Glavine was walking about 3.1 per 9IP over the period shown and not exactly a strikeout artist at 5.3 K/9 yet still dominated an ERA+ of 122 over nearly 3000 IP. The strategy of minimizing walks can go too far, like you said, if lowering it increases hits.
It might be possible to get pretty close to an ideal strike percentage for a pitch knowing how often hitters will swing and miss at it. If they can’t hit it in the zone, err towards throwing it in the zone more. If they can hit it in the zone, there’s probably a lot of wisdom in erring on the side of being outside more often. At least for guys with proven track records of keeping the ball down and inducing ground balls, preferably with stud middle infielders.
Strike one for the ERA crowd.
Re: “Telling a pitcher to throw more strikes may indeed be bad advice on occasion” — On occasion, maybe. I’m pretty sure Glavine could not have won 300 games if he pounded the strike zone. But Glavines and Moyers are few and far between.
Here’s two quick-and-dirty studies of the correlation between low walk rates and success:
(1) I took the 92 qualified pitchers so far this year and split them into three groups based solely on BB/9, then took their unweighted averages. There’s very little difference in their K rates, but that’s where the resemblance ends:
Low walks: 1.76 BB/9 … 7.5 SO/9 … 123 ERA+ … 3.37 FIP
Med. walks: 2.53 BB/9 … 7.4 SO/9 … 110 ERA+ … 3.73 FIP
High walks: 3.65 BB/9 … 7.6 SO/9 … 100 ERA+ … 3.98 FIP
(2) OK, what about guys who’ve pitched both ways? I looked for active pitchers with qualified years at 3.5 BB/9 or higher, and with 2.5 BB/9 or lower. Five guys had at least two of each: CC Sabathia, Matt Cain, Clayton Kershaw, Bartolo Colon and Randy Wolf.
I took that group’s unweighted averages for their qualified high-walk and low-walk years:
Low walks: 2.1 BB/9 … 138 ERA+ … 3.29 FIP
High walks: 3.9 BB/9 … 115 ERA+ … 3.94 FIP
Colon is an interesting example. He has 11 years with at least 150 IP. I divided those by walk rate — 6 years at or below 2.7 BB/9, and 5 years at or above 3.1 BB/9. In the low-walk years, he’s allowed slightly more hits, but slightly fewer HRs. The big difference is in walks and strikeouts … and results:
Low walks: 2.0 BB/9 … 6.1 SO/9 … 126 ERA+
High walks: 3.6 BB/9 … 7.8 SO/9 … 116 ERA+
If you adjust for the league rates, Colon’s recent years of extremely few walks are actually much lower in Ks and higher in hits than his younger years, but the results are comparable.
I think there are very few who lack swing-and-miss stuff who can learn to succeed by Glavine’s methods.
I agree but I would wager that guys with low K rates (a huge disadvantage) are better off with a Glavine-esque strategy if they can pair it with a high ground ball percentage. Danny Graves, for example, was not going to strike many guys out but he could keep the ball down pretty well. Aiming on the edge of the strike zone and not being afraid of walking about 3 guys per 9 is probably about right to maximize his effectiveness. Who’s a guy who fits this mold now? Probably somebody like recently traded pitcher Brandon McCarthy. He doesn’t strike many guys out but does a great job of keeping the ball down and induces a lot of grounders with his sinker. He’s probably better off pushing that even further, even taking a few more walks, to help that ballooning hit rate.
I think the point is although in general walks are bad, there is some balance. You need to throw good pitches.
Career spilts, runner on 3rd, lt than 2 outs:
239 PA, 32 BB, 23 SO, 19 IBB (that seems really high – 8% IBB in those spots?)
Runner on 3rd, 2 outs:
230 PA, 36 BB, 28 SO, 7 IBB
No evidence of an increase in the ability to strike someone out more often (hitter strategy changes as well, of course), the rate went down in situations where a ball in play would be less desirable. An increase in the willingness to walk someone unintentionally with 2 outs (5.4% vs. 12.6%), which makes sense for a closer, where the value of extra runs is often meaningless to the outcome.
I didn’t see an easy way to get the splits for reliever work only. I would think it is more likely that general aging caused late-career problems, not the different use patterns, but you never know. Also, the failed starting experiment…that didn’t go well.
Yeah, I also tried to get reliever splits but couldn’t find a convenient way to do it, hence why I created the closer comp group.
I presume the high number of IBB you cited is because they were potential game-ending situations where they needed a double play, or simply because it was a walkoff situation and they wanted to face an easier batter.