• Geeking Out: Pitch To Contact And Team Pitching

    Pitch To Contact - at this point, I think we can go with capital letters, don't you? That phrase and philosophy have drawn a fair amount of criticism, a chunk of which is just snark, but some of which at least tries to ground itself in statistical analysis. In a thread started yesterday on Twins Daily, there was a lot of debate on Pitch To Contact, what it means and what we really know about its effects.

    When Bill James unveiled several new tools for analyzing baseball, the most fundamental ones usually started by analyzing team stats.
    For instance, by studying teams' wins and losses, he found a correlation with runs scored and runs given up. And by studying team runs scored, he discovered a correlation between getting on base and total bases. From there, it's a short step to assuming that if you get a lot of players who get on base or hit for power, you'll score more runs.

    But how about teams that don't give up many runs? If we take a look at teams for the last few years, can we see a trend in those that don't give up many runs? And does that trend match or contradict a "Pitch To Contact" philosophy.

    So let's do a very quick-and-dirty back-of-the-napkin study. I looked at 150 teams, or all the teams from the last five years, ranked them by runs/game (R/G) and then searched the stats that most closely correlate with it. I did this using the "CORREL" function in Excel, which generates a coefficient between 0 and 1. 1 means a perfect correlation. 0 means it is entirely random. The full results are at the bottom, but here is a summary.

    Those who are critical of the Pitch To Contact philosophy are usually reacting to the reduced emphasis it puts on striking people out. The K/9 correlation to runs per game is .54, one of the lower correlations on the list. That's lower than I would have expected. Clearly, having a staff that strikes out a ton of batters isn't especially important.

    The defenders of the Pitch To Contact philosophy like to say that it's essentially saying "throw strikes." Presumably, that would imply not walking people, something that the Twins have certainly emphasized. However, the correlation of BB/9 to R/G is about the same: .56. Like strikeouts, not walking players is good, but not great.

    Combing the two gets us a little closer. SO/BB has a .69 correlation. Close to that correlation is something else the Twins have been especially good at this year: not giving up home runs. HR/9 has a .65 correlation.

    But the winner, without question, is hits. H/9 has a correlation of .88, crazy high compared to the other traditional stats in the list. That might seem obvious - if you don't give up hits, you shouldn't give up runs. So the question changes....how do you avoid giving up hits?

    Well, it isn't just "strike people out". In fact, we proved just the opposite - that's fine, but a very small part of the story. So the difference is...defense? Luck? Secret sauce?

    The truth is we don't know what it is. Sabrmetrics is still remarkably poor at predicting pitching. But we know what it isn't - it isn't as easy as gathering Ks. And whatever philosophy one adopts, its goal better be limiting hits.
    ~~~

    Here is a link to the spreadsheet to check things or play with it yourself:

    https://docs.google.com/file/d/0B-8j...it?usp=sharing

    ~~~
    And here are the final numbers....

    Stat Correl
    R/G 1.00
    R 1.00
    ERA 0.99
    ER 0.99
    WHIP 0.93
    H/9 0.88
    H 0.86
    ERA+ -0.81
    BF 0.80
    SO/BB -0.69
    L 0.66
    W-L% -0.66
    W -0.66
    HR 0.66
    HR/9 0.65
    tmSHO -0.65
    SO -0.57
    BB/9 0.56
    SO/9 -0.54
    #Pitch 0.54
    BB 0.53
    SV -0.53
    IP -0.51
    cgSHO -0.36
    GF 0.32
    CG -0.32
    HBP 0.27
    WP 0.27
    BK 0.20
    PitchAge -0.13
    IBB 0.01
    This article was originally published in blog: Geeking Out: Pitch To Contact And Team Pitching started by John Bonnes
    Comments 38 Comments
    1. Linus's Avatar
      Linus -
      Thank god for this post. The "strikeouts are everything" discussion was going way to far - especially concerning potential free agents. Baseball history is filled with successful pitchers who strike out a lot of batters and the opposite; same thing with walks (although less so I believe) as there have great pitchers who walk very few and some that walk a surprisingly high amount. Clearly there is no magic formula (at least that we have discovered yet). My guess is that if they could ever quantify a stat for keeping the ball off the fat part of the bat, you'd have your answer.
    1. Oxtung's Avatar
      Oxtung -
      I think it is important to note that there is still a correlation between team strikeouts and team runs allowed. It might not be the only important piece of this puzzle but it is still a piece. In general if you strike out more batters you will be more successful.
    1. Willihammer's Avatar
      Willihammer -
      By my math, the correlation between K% and H% going back to 2007 is -.824.

      Stronger relationship to Runs and ERs using K%, and weaker relationship using BB%, as opposed to /9 stats.

      K%-R correl is -.69. BB%-R correl is .13

      https://docs.google.com/file/d/0B5pI...it?usp=sharing
    1. John Bonnes's Avatar
      John Bonnes -
      Quote Originally Posted by Willihammer View Post
      By my math, the correlation between K% and H% going back to 2007 is -.824.

      Stronger relationship to Runs and ERs using K%, and weaker relationship using BB%, as opposed to /9 stats.

      K%-R correl is -.69. BB%-R correl is .13

      https://docs.google.com/file/d/0B5pI...it?usp=sharing
      I uploaded my spreadsheet to Google Docs and provided a link in the story. Thank Willihammer for reminding me we could do that.

      That BB%-R number is a little alarming. When you say "going back to 2004" are you saying you took all 30 teams and combined each team's numbers back to 2004, so you only have 30 items in your correlation? If so, that doesn't seem like enough to trust, though I don't know what that threshold should be. I did team-years, so I grabbed 30 teams for five year for 150 data points, just because I wanted to have at least that many.

      (Or it could be that I have an error in my spreadsheet, too. It's certainly possible, which is one of the reasons I provided a link to it.)
    1. Willihammer's Avatar
      Willihammer -
      Quote Originally Posted by John Bonnes View Post
      When you say "going back to 2004" are you saying you took all 30 teams and combined each team's numbers back to 2004, so you only have 30 items in your correlation? If so, that doesn't seem like enough to trust, though I don't know what that threshold should be. I did team-years, so I grabbed 30 teams for five year for 150 data points, just because I wanted to have at least that many.

      (Or it could be that I have an error in my spreadsheet, too. It's certainly possible, which is one of the reasons I provided a link to it.)
      Yeah I didn't think that smelled right either and you have pointed out why. I did a simple fangraphs query using combined totals, so my set was just 30. I did it again using season-totals instead and sure enough the BB%-R correlation comes up a more meaningful .43. K%-R is still higher than K/9 correlation though, at -.65. H%-R comes in at .80.

      I really only bring it up becuase of a pet peave with /9 stats which are going to discount any relationship you try to glean from strikeouts, as opposed to PA-based stats. Issue more walks, give up more hits, face more batters, strikeout more batters per 9! and vice versa.

      https://docs.google.com/file/d/0B5pI...it?usp=sharing
    1. Brodin4Calder's Avatar
      Brodin4Calder -
      Nobody is complaing about pitching to contact, the reason us Twins fans arent a fan of it right now is because the pitchers that we have that do it just plain suck. There are strikeout pitchers that are just as bad. Our problem isnt pitching to contact, our problem is we have horrible pitchers that pitch to contact. Theres alot of really good pitchers that pitch to contact, one of them is Johny Cueto. Is probably the best groundball pitcher, but he throws hard and gets Ks too. Another thing to note is almost every roto has contact pitchers but they also have strikeout pitchers, we only have contact pitchers. Our contact pitchers struggle to hit 90mph so they are throwing meatballs that they want the batters to hit ha. Alot of the best contact pitchers have an avg fastball around 92-93 our guys avg fastball is around 89mph. We have bad contact pitchers, plain and simple.
    1. Anorthagen's Avatar
      Anorthagen -
      This is so much to comprehend at one time.
    1. PopRiveter's Avatar
      PopRiveter -
      This is a really fascinating angle. Strikeouts are really really fun. Very useful too, but how are they better than an infield fly? Weren't Correia's 4 double-play balls last night more valuable than 4 strikeouts would've been?
      I've heard all the complaints about Pitch to Contact as a strategy, but I've watched a lot of Francisco Liriano starts. I don't get how a fanbase who watched him strike out soooooo many while simultaneously giving up sooooooo many runs can have so many members who exaggerate the value of the strikeout and minimize the value of all the other outs.
      I think your number tells the truth that a strikeout is a very good out. Not necessarily the best. Sadly, a number can't tell how fun they are.
    1. Rick Niedermann's Avatar
      Rick Niedermann -
      Seth is right. It's WHIP. Examples: Peavey Lifetime WHIP 1.175- worst WHIP of 1.423 resulted in 6W 7L 4.52 ERA, Verlander Lifetime WHIP 1.174-worst WHIP of 1.403 resulted in 11W 17L 4.84, C Lee Lifetime WHIP 1.207-worst WHIP 1.521 resulted in 5W 8L 6.29 ERA.
    1. jay's Avatar
      jay -
      Quote Originally Posted by Seth Stohs View Post
      Strikeouts just get expensive!
      This is on point.

      If you have two pitchers with the exact same stats/results but one guy gets more strikeouts, that guy is going to make more money. You almost hate to say it, but I have to believe this is at least a part of the rationale behind the Twins' strategy.
    1. Willihammer's Avatar
      Willihammer -
      Quote Originally Posted by Seth Stohs View Post
      Walks are annoying, but figure equally to a single. (all else being equal, as in, didn't drive in a run or anything).
      That's a heck of a caveat.
    1. old nurse's Avatar
      old nurse -
      How a pitcher grips a ball would be an indication of what pitch is coming. Disguising and varying the speed and movement of the ball then becomes a key component in preventing the ball from meeting the sweet part of the bat. How you would measure that to prove it would be a long task. That slugging percent rated higher than batting average in scoring runs might confirm the notion. Carl Pavano's 2010 season would show that it can be done without a lot of strikeouts. All the good strikeout pitchers would prove that it might be easier to find a great strikeout pitcher than a great pitch to poor contact pitcher.
    1. gswanson's Avatar
      gswanson -
      Quote Originally Posted by John Bonnes View Post
      But we certainly shouldn't treat strikeout rate as anywhere near as important as OPS. They're not in the same solar system.
      I don't think it should be surprising at all that hits are the strongest predictor of scoring. After all, hits are the primary way of getting on base as well as the only way to hit for power. I don't think any sabermetrician would doubt that hits correlate strongest with runs scored.

      Instead, I think sabermetrics would predict that just as OBP and SLG correlate strongest with hitter's scoring (as you said in your introduction), OBP-against and SLG-against would correlate strongest with a pitcher allowing runs to score.

      The correlations you ran of H/9, BB/9, HR/9 and SO/9 are basically serving as proxies for OBP-against and SLG-against when compared to ERA. And just as we might expect, H/9 is the best proxy for both OBP- and SLG-against because, again, hits are the most common way to get on base and the only way to slug. Of course, HR/9 was also highly correlated with scoring because of its major relationship to slugging (and it's minor relationship to getting on base).

      Predictably, BB/9 and SO/9 did not correlate as well with runs allowed because of their more tangential relationship to OBP.

      So all in all, your study basically showed (through proxies of varying quality) that whereas OBP and SLG are the best tools around which to build an offense, their prevention is the best way to guarantee a low ERA.

      Just for fun, I downloaded your table and made a down-and-dirty version of OBP-against by taking H, BB, IBB, and HBP over batters faced. The correlation between H/9 and R/G per your table was .882. The correlation for OBP-against? .886.

      Similarly, I made a REALLY down-and-dirty estimation of SLG-against. I tried to estimate total bases by taking all hits and adding the three extra bases for each home run allowed. Then I divided this by batters faced minus BBs, IBBs, and HBPs. While this "semi-slugging" percentage was much lower than the actual slugging percentages, I figured the differences among teams would hold up alright. The result? A .869 correlation, much higher than HR/9 alone.

      Now your last question was your best: What prevents hits (or more accurately, OBP and SLG)? Here is where K rate becomes important. But all sabermetricians would agree that BB rate and GB rate are important too. The problem with a "pitch to contact" philosophy is its reliance on the shakier idea that pitchers can control their BABIP and the hitting profile of the batters they face (ie their power). But if you're willing to concede that pitch-to-contact-if-you-induce-groundballs is a philosophy, then I'm sure many sabermetricians would take that just as often as high strikeout rates.
    1. Badsmerf's Avatar
      Badsmerf -
      I'm with Oxtung on this. Using teame stats isn't the best indicator. I'd feel like the stats would tell a different story when using individuals. The elite strike-out pitchers, Aces, allow less runs. I can't think of many elite strike-out pitchers that are just average at allowing runs. Once you dilute the field and include all pitchers, you are making their impact minuscule. Most MLB pitchers are going to strike out between 5 and 7 batters per 9. These guys far outweigh elite pitchers in a compilation stat. The numbers look interesting as a whole, but tell us little for individuals.
    1. USAFChief's Avatar
      USAFChief -
      Quote Originally Posted by gswanson View Post
      I don't think it should be surprising at all that hits are the strongest predictor of scoring. After all, hits are the primary way of getting on base as well as the only way to hit for power. I don't think any sabermetrician would doubt that hits correlate strongest with runs scored.

      Instead, I think sabermetrics would predict that just as OBP and SLG correlate strongest with hitter's scoring (as you said in your introduction), OBP-against and SLG-against would correlate strongest with a pitcher allowing runs to score.

      The correlations you ran of H/9, BB/9, HR/9 and SO/9 are basically serving as proxies for OBP-against and SLG-against when compared to ERA. And just as we might expect, H/9 is the best proxy for both OBP- and SLG-against because, again, hits are the most common way to get on base and the only way to slug. Of course, HR/9 was also highly correlated with scoring because of its major relationship to slugging (and it's minor relationship to getting on base).

      Predictably, BB/9 and SO/9 did not correlate as well with runs allowed because of their more tangential relationship to OBP.

      So all in all, your study basically showed (through proxies of varying quality) that whereas OBP and SLG are the best tools around which to build an offense, their prevention is the best way to guarantee a low ERA.

      Just for fun, I downloaded your table and made a down-and-dirty version of OBP-against by taking H, BB, IBB, and HBP over batters faced. The correlation between H/9 and R/G per your table was .882. The correlation for OBP-against? .886.

      Similarly, I made a REALLY down-and-dirty estimation of SLG-against. I tried to estimate total bases by taking all hits and adding the three extra bases for each home run allowed. Then I divided this by batters faced minus BBs, IBBs, and HBPs. While this "semi-slugging" percentage was much lower than the actual slugging percentages, I figured the differences among teams would hold up alright. The result? A .869 correlation, much higher than HR/9 alone.

      Now your last question was your best: What prevents hits (or more accurately, OBP and SLG)? Here is where K rate becomes important. But all sabermetricians would agree that BB rate and GB rate are important too. The problem with a "pitch to contact" philosophy is its reliance on the shakier idea that pitchers can control their BABIP and the hitting profile of the batters they face (ie their power). But if you're willing to concede that pitch-to-contact-if-you-induce-groundballs is a philosophy, then I'm sure many sabermetricians would take that just as often as high strikeout rates.
      Awesome first post.
    1. Oxtung's Avatar
      Oxtung -
      Quote Originally Posted by Badsmerf View Post
      I'm with Oxtung on this. Using teame stats isn't the best indicator. I'd feel like the stats would tell a different story when using individuals. The elite strike-out pitchers, Aces, allow less runs. I can't think of many elite strike-out pitchers that are just average at allowing runs. Once you dilute the field and include all pitchers, you are making their impact minuscule. Most MLB pitchers are going to strike out between 5 and 7 batters per 9. These guys far outweigh elite pitchers in a compilation stat. The numbers look interesting as a whole, but tell us little for individuals.
      Just to be clear, I don't know if the numbers would change just looking at individual pitchers rather than team numbers. My argument was more a logic based one. Looking at a sum of numbers tells you nothing about the individual parts that constitute that sum.

      For example if you have 5+5 you know the resulting sum is 10. However, if all you have is the sum, in this case 10, you know nothing about the terms that were used to derive the sum. It could be 5+5, 3+3+3+1, 2+8, 1+7+2, etc...
    1. IdahoPilgrim's Avatar
      IdahoPilgrim -
      Quote Originally Posted by Oxtung View Post
      Just to be clear, I don't know if the numbers would change just looking at individual pitchers rather than team numbers. My argument was more a logic based one. Looking at a sum of numbers tells you nothing about the individual parts that constitute that sum.

      For example if you have 5+5 you know the resulting sum is 10. However, if all you have is the sum, in this case 10, you know nothing about the terms that were used to derive the sum. It could be 5+5, 3+3+3+1, 2+8, 1+7+2, etc...
      Does it really matter? At the end of the day, this is a team sport and aggregate team ability is what drives success, including aggregate team pitching. We could have one superstar and four middlin' pieces - it won't add up to a world series (think Kevin Garnett and the Timberwolves).

      I'm not trying to be argumentative - just raising the question.
    1. Oxtung's Avatar
      Oxtung -
      Quote Originally Posted by sbknudson View Post
      Does it really matter? At the end of the day, this is a team sport and aggregate team ability is what drives success, including aggregate team pitching. We could have one superstar and four middlin' pieces - it won't add up to a world series (think Kevin Garnett and the Timberwolves).

      I'm not trying to be argumentative - just raising the question.
      Since the pitcher on the mound any given night isn't some freakizoid conglomeration of body parts encompassing every pitcher on your staff, yes it matters. The only way to know if strike outs correlate to less runs is to look at it from an individual pitcher perspective.
©2014 TwinsCentric, LLC. All Rights Reserved.
Interested in advertising with Twins Daily? Click here.