The Unthinkably Exceptional Scott Kazmir
Wouldn’t you know it, but Scott Kazmir is relevant again. Jeremy Bonderman is relevant again, too, and that’s also amazing, but Kazmir’s actually got himself a big-league rotation job and that gives me an opportunity to talk about an incredible Kazmir fun fact. I could’ve talked about it anyway, even if Kazmir were completely off the radar, but now it’ll be less insignificant. You’re thinking about Scott Kazmir, and here’s another thing to think about him.
Pitching with the bases empty is different from pitching with at least one runner on. For many pitchers, the delivery changes, and of course the situation and the intensity changes. The defense changes. Things change when there are people on base, so it can be informative to look at the base-state splits. In such situations it’s always important to remain aware of small sample sizes, but when you get to talking about whole careers, those concerns by and large go away.
Just the other day, I discovered the Baseball-Reference Split Finder. Since then I’ve practically worn it out, checking in on splits for both hitters and pitchers. I got to looking at base-state splits, and in particular I was curious about pitching with the bases empty versus pitching with men on. To cut a long story short, I examined the window between 1973-2012, for which Baseball-Reference has complete splits data. I set a minimum of 1,000 overall innings pitched, and I looked at OPS allowed, since OPS is kind of all-inclusive. I could’ve looked at just walks or strikeouts or hits or home runs, but everything matters and everything goes into OPS, even if OPS is clearly imperfect. It’s the best I could do, so I did it.
I wanted to know about guys who were better and worse with runners on base. Since 1973, 462 different pitchers have thrown at least 1,000 innings, and their average OPS allowed with no one on was .708. Their average OPS allowed with runners on was .740, or 104% the bases-empty OPS. We’ve known for a while that pitchers put up worse numbers with runners on, for a variety of reasons.
I sorted by what we’ll refer to as “OPS_on%”. OPS_on% is simply (OPS_on / OPS_empty) *100. With nothing further, the top 10 in terms of lowest OPS_on%:
| Player | OPS_empty | OPS_on | OPS_on% |
|---|---|---|---|
| Scott Kazmir | 0.805 | 0.670 | 83.2% |
| Doug Davis | 0.802 | 0.738 | 92.0% |
| Shane Rawley | 0.761 | 0.707 | 92.9% |
| Steve McCatty | 0.749 | 0.696 | 92.9% |
| Eric Plunk | 0.735 | 0.689 | 93.7% |
| Steve Trachsel | 0.798 | 0.749 | 93.9% |
| Randy Lerch | 0.808 | 0.761 | 94.2% |
| Esteban Loaiza | 0.798 | 0.755 | 94.6% |
| Bob Wickman | 0.726 | 0.687 | 94.6% |
| Freddy Garcia | 0.750 | 0.711 | 94.8% |
Within the sample, 74 pitchers allowed a lower OPS with runners on than with the bases empty. The second-best ratio belongs to Doug Davis, at 92.0%. The best overall ratio belongs to Scott Kazmir, way down at 83.2%. To give you a visual sense of how weird Kazmir is in this regard:
Kazmir doesn’t only lead — he leads by a statistical mile. In the sample, his OPS allowed with the bases empty is tied for the sixth-highest. His OPS allowed with runners on is the 31st-lowest. If you don’t care for OPS, you can look at wOBA; he’s at .354 with the bases empty, and .297 with runners on. That comes out to 83.9%, which isn’t meaningfully different.
If it’s more splits you’re after, with runners on Kazmir’s strikeout rate has gone up from 21% to 24%. His FIP has gone down from 4.79 to 3.51, and his xFIP has gone down from 4.66 to 3.94. Not that we should be looking at xFIP that much — Kazmir has samples of more than 500 innings for each split, and he’s allowed far fewer dingers with at least one runner on base. Most pitchers pitch worse with runners on. Kazmir has pitched markedly better, and exceptionally better.
In every single individual season, ignoring Kazmir’s one start in 2011, he’s allowed a lower batting average with runners on. In every single individual season, he’s allowed a lower OBP with runners on. In every single individual season, he’s allowed a lower SLG with runners on, so therefore in every single individual season, he’s allowed a lower OPS with runners on. Only once did Kazmir allow a higher BABIP with runners on, and that was in 2008, and it wasn’t by much. Overall, Kazmir’s allowed a .325 BABIP with the bases empty, and a .292 BABIP with the bases not empty.
There are three ways to interpret this. One is that it’s all meaningless statistical noise. Congratulations, you might be right, and you suck. Kazmir’s numbers do have to be regressed some, but he’s so far out there that the broader point remains. Another interpretation is that Kazmir really bears down under pressure. A third interpretation is that Kazmir doesn’t bear down when he’s not under pressure. The last two are related, and it has to do with whether you have a positive or negative perspective. Should Kazmir be praised for pitching better when it’s mattered most, or should he be criticized for maybe being too careless with the bases empty?
It might all be a mechanical thing. Kazmir has never been blessed with outstanding command, and that usually comes down to inconsistent mechanics. It’s possible Kazmir has been more consistent from the stretch, allowing him to pitch with better polish. With the bases empty, under 62% of Kazmir’s pitches have been strikes. With runners on, over 63% of Kazmir’s pitches have been strikes. And then there’s the whole wOBA difference thing, and the home-run-rate difference thing. Maybe this is the most likely explanation, but I don’t know, and research didn’t help.
Somebody else can worry about the why, I suppose. Maybe somebody could ask Kazmir himself. That’s more feasible today than we would’ve expected it to be a year or two ago, when Kazmir seemed like toast. It’s 2013 and he’ll be in a big-league clubhouse! The difference between Scott Kazmir with the bases empty and Scott Kazmir with runners on is historically exceptional. Even without an explanation, that’s wild.
In closing, just in case you’re curious, here’s the opposite of the table above — the top 10 in terms of highest OPS_on%:
| Player | OPS_empty | OPS_on | OPS_on% |
|---|---|---|---|
| Mike Timlin | 0.639 | 0.772 | 120.8% |
| Mark Redman | 0.717 | 0.866 | 120.8% |
| Eric Milton | 0.726 | 0.868 | 119.6% |
| Chris Hammond | 0.683 | 0.812 | 118.9% |
| Willie Blair | 0.745 | 0.876 | 117.6% |
| Steve Avery | 0.673 | 0.789 | 117.2% |
| Jake Westbrook | 0.693 | 0.806 | 116.3% |
| Woodie Fryman | 0.669 | 0.777 | 116.1% |
| Reggie Cleveland | 0.654 | 0.759 | 116.1% |
| Randy Jones | 0.636 | 0.734 | 115.4% |
With runners on, Timlin lost a lot of his strikeouts. Redman, meanwhile, saw his home-run rate go up by 44%, while Milton saw his home-run rate go up by 23%. “How is Mike Timlin the opposite of Scott Kazmir?” is a really terrible riddle. But at least now you know the answer to a question no one would ever ask.
Jeff made Lookout Landing a thing, but he does not still write there about the Mariners. He does write here, sometimes about the Mariners, but usually not.
This was cool stuff! thanks!
Agreed. Great find, Jeff, and great job laying it out there.
Who’s the pitcher at the bottom left of the graph?!
Mariano Rivera isn’t fair.
A grid graph would have better in this case, no?
Sure?
jeff, is there any noticeable starter/reliever split for OPS_on? perhaps attributable to the fact that relievers generally pitch out the stretch exclusively?
Don’t have that information at present
I wonder if he would consider pitching from the stretch at all times. If it’s the “bears down more thing”, that’s pretty pathetic.
this. I would love someone more smarter then me to compare his motion with runners on and off.
I would also be interested in the amount of time he takes between pitches with runners on and off.
“Now you know the answer to a question no one would ever ask” would be a good motto for every Jeff Sullivan post… which is why we love him so!
Very interesting. BTW, Mike Timlin’s inability to strand inherited runners was well known – did people recognize Kazmir’s pattern?
I’m going to go look up Hochevar.
The fact that Ricky Nolasco is not on the Top 10 Highest OPS_on% list amazes me.
Kazmir has suffered a lot of injuries, yes? And my understanding is pitching itself is painful and unnatural, so pitchers that suffer injuries are probably in pain for a long time before the actual injury? And the stretch puts more strain on the arm because it eliminates some of the body action? So “bearing down” in this case may be a very good way to put it, in that Kazmir must increase his effort/strain, and maybe pitching is so painful for him, and injuries so common, he’s wary of “bearing down” very much?
I don’t know the answers to any of these questions. That’s not true about the first question.
This is probably the kindest explanation for Kazmir, and it also happens to mesh with what I saw during his Angels stint. I remember he always seemed to look so labored when he pitched, it did look painful anyway, and was even painful to watch, even when he was doing well. Maybe he was in a lot of pain all the time and had to save his best efforts for higher leverage situations, and nobody really knew it but himself and maybe his teams.
Kazmir’s strategy this season should be obvious: Walk the lead-off man every inning.
Seems to be his “strategy” (or outcome) fairly often.
Then how about the alternative. Put your first baseman on first and pretend he’s holding somebody on. Maybe that will trick Kazmir into pitching like a stud.
Another interesting find from B-R’s split -finder is that Tim Lincecum’s ops allowed in a 4-2 count is .000.
Wonderful post. This is why we read this site. Nice work.
Amazing work, this is exactly why I love the site as most analysis deals with intangibles like “he’s a really fast player whereas fangraphs community will say he’s x standard deviations faster than the average player at his position.” It’s been a while since I was in school and took any stat classes, do you know of any websites that could explain how I could chart and analyze something the same way?
I found it interesting that Jake Westbrook, who made his career as a sinkerballing ground-ball specialist finds himself on the list of the ten worst pitchers with runners on.
Doesn’t Darvish pitch from a stretch at all times? Maybe that’s the ticket for Kazmir.
Well researched, well written, and informative. Great article.
Great job.
Hows his old teammate Hell-boy not on the list?
He’s as stingy as anything with runners on also.
Minimum of 1,000 IP.
Is there a way to look up the average # of pitches per AB with runners on and with bases empty? Having watched Kaz a lot when he was with the Rays, my guess would be he was extra careful with runners on, throwing a bunch more pitches nibbling around the corners and getting pitches fouled off. He was constantly running up high pitch counts and I am sure was told to throw more strikes to keep the count down, so maybe he was just grooving it in more trying to pitch to contact with the bases empty. Could be completely wrong though. Cool article.
Sarah Bernhardt reportedly could make an audience cry by reading from the phone book. I won’t say you made me cry, but you wrote an article that interested me about a trivial statistic.
One question to throw into the “statistical noise” column is the fact that some coaches may signal their pitchers to pitch to contact to get the batter to ground into double plays. Just a thought.
Kazmir is one of my alltime favorite pitchers, so this article means a lot to me. Beautiful statistical analysis. I can only hope somebody relays this data back to Kaz, so we can find the “why”. At first read, one would think that switching to the stretch full-time would be the ticket for success for Kaz. Who knows? I just hope he can rectify his career in Cleveland this season!!
What’s interesting is that his strike % doesn’t change, so maybe it’s actually the hitter’s approach with men on base that’s different in this case?
Great stuff. Because I have a prejudice against walks I wondered if Kazmir was forced to pitch more in the zone with runners on and thus got better results, but that doesn’t appear to be the case as his walk rate is better but not significantly. The home run rate is almost cut in half, though, which suggests to me that the difference really is in his delivery from the stretch.