FanGraphs Baseball


RSS feed for comments on this post.

  1. Great article. Can we expect to see a “Edge%” stat on Fangraphs in the future?

    Comment by Woodman — January 15, 2013 @ 9:30 am

  2. Awesome awesome stuff. The next step would be to figure out if Edge% correlates year to year, yes? Great work, Bill (and Jeff).

    Also, minor typo in the 4th point: “Throwing a higher percentage of pitches on the edges leads is associated with lower fastball velocity.”

    Comment by Matt Hunter — January 15, 2013 @ 9:33 am

  3. I ran the correlation for 2011 to 2012
    Minimum number of pitches in each season: R-squared
    Any number: 0.06
    100: 0.20
    500: 0.31
    1000: 0.35
    2000: 0.53

    Comment by Jeff Zimmerman — January 15, 2013 @ 9:35 am

  4. Good stuff Bill. I know there’s a lot more stuff/data to put together on this. But I hope this moves to pitch sequence location. Not certain how to explain. But changing location high/low and in/out is said to make a big difference also. Also, I’ve heard talk about pitchers that can’t/won’t pitch to certain quadrants of the plate. Some pitchers just can’t pitch inside for whatever reason. It would be interesting to find out the performance of these pitchers compared to those who use all 4 quadrants.

    Comment by Chicago Mark — January 15, 2013 @ 9:35 am

  5. Ah, yes, just saw that on BtB after I posted this. Thanks, Jeff.

    Comment by Matt Hunter — January 15, 2013 @ 9:41 am

  6. I would’ve thought that pitching on the edge would correlate with lower velocity because guys would be taking a tick or two off in exchange for greater control.

    Is there a way to determine if pitches around the edge are, on average, slower? (Or maybe that’s in this data and I’m terrible at reading it.)

    Comment by MrKnowNothing — January 15, 2013 @ 10:01 am

  7. Great article, I would love to see this subject explored more. One thought about the unexpected correlation between fastball velocity and edge% is that maybe you should fix on a particular ERA (or FIP or xFIP or something). Your hypothesis that pitchers who have bigger fastballs would be less able / have less need to hit the corners makes more sense if you look at pitchers with about the same level of overall effectiveness. It’s possible that the pitchers who have more overall talent both throw harder and have a higher edge%, but if you look at a fixed level of overall talent then the two become negatively correlated as you’d expect.

    Comment by Jon — January 15, 2013 @ 10:02 am

  8. I see Fister at 16.8% EDGE on the 2012 link….sorting problem?

    Comment by Scott — January 15, 2013 @ 10:04 am

  9. Fister is listed at 18.8% for 2011….

    NOTE: I am looking at the links from yesterday’s article:

    Have those links been made obsolete by a redefinition of the metric since yesterday’s article?

    Comment by Scott — January 15, 2013 @ 10:07 am

  10. Not obsolete, just two different takes on defining Edge%. It is still a work in progress.

    Comment by Bill Petti — January 15, 2013 @ 10:09 am

  11. No, they will probably become obsolete at some point in the future. Bill and I are trying to set the exact edge. It looks like it will be non-symmetrical with a larger one on the outside than the inside. I am guessing the next one of ours article will discuss it.

    Comment by Jeff Zimmerman — January 15, 2013 @ 10:13 am

  12. Absolutely wonderful.

    Comment by James Gentile — January 15, 2013 @ 10:13 am

  13. Thanks for the explanation guys. This is really exciting work and IMO baseball reserach of the best variety: confirming what we have always suspected with numbers to back it up. I have to admit that I was selfishly trying to figure out how to work this data into my fantasy projections for 2013 when the lists came out yesterday!

    Comment by Scott — January 15, 2013 @ 10:21 am

  14. I can not agree at all. When pitching in college my best control often correlated with my best stuff. Proper mechanics pitch to pitch lead to higher velocity and better control. Any time I was trying to be careful or take something off for better control it led to a change in release point and poor location.

    Reading books by Randy Johnson, Tom House, etc. They would suggest the exact same results–so I am not attributing this only to one mediocre college career.

    Comment by Patrick — January 15, 2013 @ 10:23 am

  15. Or maybe the real life results are selecting the data, and not vice versa.

    ie. pitchers that can’t blow you away with stuff NEED to live on the edges… or they get pounded, and get run out of the league… leaving you with these sorts of data.

    because they’re whats left.

    Comment by Dave S — January 15, 2013 @ 10:26 am

  16. I also looked at my measure for pitchers with >=200 IP for 2011 and 2012, and the R-squared was .48.

    Comment by Bill Petti — January 15, 2013 @ 10:27 am

  17. Great work guys. I can believe the need for a non-symmetrical contribution from the two sides too as you suggested may be required. Looking forward to more follow-up work.

    Comment by Jon Roegele — January 15, 2013 @ 10:35 am

  18. This is nice, but what about breaking it down even further? Edge percentage inside vs away?

    Comment by MakeitRayn — January 15, 2013 @ 10:50 am

  19. Fantastic work. Since strikeouts and avg seem to be better in the 75+ percentile than the 90+ percentile, perhaps there is such a thing as pitching too much on the edge for certain types of pitchers. The interesting thing will be looking at more pitchers progression over time as you did with Price and Fister.

    Comment by Lee Panas — January 15, 2013 @ 10:54 am

  20. This is excellent work! I particularly like the Price- Fister references.

    Fister is an interesting pitcher to study. You would think that such a tall right hander would be relatively easier to run on, yet he has allowed a grand total of ten stolen bases in his career, now over three seasons. He’s actually very quick to the plate with a slide step delivery, and he obviously is able to generate a lot of downward plane on the ball with a 6 foot- 8 inch frame.

    He has quietly posted a 3.33 FIP from 2010- 2012, which is sixth lowest in the AL- also 9th in ERA and 8th in WAR. With all that going for him, if he’s able to paint the corners as well, he’ll be scary good.

    What we’re really measuring here with the Edge% is command. The count could often dictate location, such as in the case of a pitcher throwing one off the plate on an 0- 2 count, or coming down the middle on a 2- 0 count. But overall, you’d think that a pitcher that is in the higher range of Edge% is demonstrating better command.

    Nice work, guys.

    Comment by tigerdog — January 15, 2013 @ 11:30 am

  21. Hey, this is great. Are the samples large enough to look within season for an individual pitcher? It seems possible that a higher edge% is driven by better mechanics which leads to better “stuff” making it not causal but just associated. Maybe comparing pitches on the edge vs. off the edge during the same time period could help to get at the causality a little bit more. Also it is completely possible that you have done that or are planning to do that and I haven’t read the article well enough to notice.

    Comment by Macek — January 15, 2013 @ 12:51 pm

  22. I just realized that I was confusing. I meant on the edge vs. in the middle of the plate. Off the edge makes it sound like I was talking about balls and not strikes.

    Comment by Macek — January 15, 2013 @ 12:54 pm

  23. This is fascinating stuff. I really appreciate it. You mentioned that the correlations were small but when you add a little to the K rate, subtract a little from the BB rate, and subtract a little from BABIP, the 3 combine to have a pretty large impact on a pitcher’s success. A pitcher can go from an average pitcher to a good pitcher just by doing those little things. My point is that even though the correlations are small, they add up to be something much larger.

    Really great. Now I need some time to just sit around and look through all the data. Thanks.

    Comment by chuckb — January 15, 2013 @ 12:56 pm

  24. Obviously there’s a lot more work that can be done with this, and I agree with everyone else that this is fantastic stuff. You’re going to be restricted some what by sample, but I would love to see how well Edge% correlates with K% or ERA in the subsequent season.

    It’s great to explain maybe why David Price’s BABIP was so low in 2012, but it’d be even more valuable to see what this metric tells us about future performance. I figure both of you have thought of this/plan on doing it, but if it adds significant predictive value, that would be awesome.

    Keep it up, fellas

    Comment by Glenn DuPaul — January 15, 2013 @ 1:08 pm

  25. No, sorry, you can’t do that.

    Comment by Baltar — January 15, 2013 @ 3:34 pm

  26. Holy cow do I love this. Making observations, setting forth a hypothesis, developing a metric to test it, then testing it! It’s so… scientific! Great work, as always.

    So, if this stat correllates fairly well with BABIP, do you think it could be used as part of an explanation for pitchers with hit suppression? And if so, do you foresee a way to use Edge% to determine a pitchers ‘true BABIP’?

    Comment by SirCub — January 15, 2013 @ 4:06 pm

  27. This is outstanding. I wish I’d had this when I was doing some of my expected BABIP articles in FanGraphs Community Research (article 1 and article 2). I think I’ll have to see how well I can integrate this into my research.

    Will you look into the vertical equivalent of this?

    I think it would also be interesting to see the relationships between Edge%, Heart%, and Outside% vs. BABIP (and other factors, like SwgStr%) when broken down by pitch type.

    Comment by Steve Staude. — January 15, 2013 @ 7:37 pm

  28. Great article. Just one small nitpick, but in the table “> 25″ percentile should be “< 25" percentile. Can't wait to see where this goes.

    Comment by Sean — January 15, 2013 @ 10:49 pm

  29. It would be interesting to see what pitch selection correlated with a successful ‘edge pitcher’. For instance, here, in the Bay Area, it’s something of a truism that Lincecum’s decline seems to have paralleled his inability to get his 2 seam FB in the edge zone.

    Comment by chanelclemente — January 16, 2013 @ 4:54 pm

  30. Very cool. Any guesses as to whether or not Edge% is consistent enough for individuals to be predictive or if it’s a tool best served to autopsy prior performance?

    Comment by jdbolick — January 16, 2013 @ 5:30 pm

  31. No offense intended, and I am not a traditionalist who is anti sabermetrics, but isn’t this getting just a little silly? You seem to have gone through several hoops and formulae to explain to us that it’s a better idea to try and not groove pitches down the heart of the plate.Uh,no kidding.This is not stop the presses news.Some baseball people learned that a while back.

    Comment by Ballpark Frank — January 16, 2013 @ 6:29 pm

  32. Maybe you can combine this with the the earlier article of strikezone by count and actually physically adjust what is considered the “edge” to reflect what the effective edge is in that count. Certainly pitchers don’t calculate precisely what an umpire is likely to call, but I’m also certain they know that a hitter’s count leads to a larger strike zone and a pitcher’s counts lead to a small strike zone. Therefore a pitcher who is good at pitching to the edges would purposely use a bigger zone and compensate for a smaller one.

    Am I correct in thinking that Edge percentage is as a percentage of all pitches thrown? Maybe you could try the analysis again but as a percentage of all pitches in the zone. When a good pitcher throws a pitch out of the zone, it’s often the case that he meant to do it. As it is, doing this is penalizing good pitchers, since bad pitchers will throw more balls, but they also get into fewer favorable counts and then have to throw more pitches in the zone to compensate. Many good pitchers have low Zone%, so it seems to me that if we’re assuming a pitcher throws to the edge on purpose, it should also be assumed he meant to throw it in the zone, and pitches intended to be out of the zone are noise.

    Comment by Bip — January 16, 2013 @ 6:42 pm

  33. There are usually good reasons to do a statistical test that confirms common knowledge. Here are some, in this instance:

    -With statistics, instead of getting a qualitative answer (e.g. yes it is good to pitch to the edges) we get a quantitative answer. We see exactly what percentage of ERA variation is explained by a pitcher’s ability to pitch to the zone, as well as the magnitude of the effect.
    -We can measure which pitchers actually do it and how well. Instead of having general knowledge, we have specific knowledge about each pitcher
    -We can see if Edge% is a skill that is repeatable or possibly a randomly varying result that explains some randomness in ERA
    -There is a chance that we may actually debunk common knowledge, so it’s worth checking even if we don’t stand to learn anything if it’s confirmed

    There are probably more.

    Comment by Bip — January 16, 2013 @ 6:47 pm

  34. One assumption is a pitcher needs at least one of good velocity and good command to stay in the big leagues. A pitcher with bad FB velocity and bad command as represented by Edge% would be out of the league. So what’s left are the guys who have one or the other or both. This would create a negative correlation.

    But if there is a stronger negative correlation between vFB and Edge% among pitchers with similar “effectiveness” as you suggest, that would be quite interesting indeed…

    Comment by Bip — January 16, 2013 @ 6:54 pm

  35. It’s possible that once you get into the 90th percentile you get to the guys who, knowing they don’t have the stuff to blow guys away, have focused on command to an extreme degree. So having below average stuff makes a guy more likely to improve his command, meaning that command and stuff, as represented by K%, are not independent.

    Comment by Bip — January 16, 2013 @ 6:57 pm

  36. King Felix is a good example. His K%, BB% and GB% are all good but not great. However, being above average in all three makes him an elite pitcher. Pitchers more commonly excel at two and lag in a third.

    Comment by Bip — January 16, 2013 @ 7:00 pm

  37. The vertical case seems a little more difficult. It seems that the effectiveness of pitching up vs down in the zone depends more on the pitcher, including his delivery, release point and the types of pitches he throws. Different pitches from the same pitcher probably have different effectiveness high vs low. The horizontal edges are are probably more reliably good for the pitcher when contrasted with the middle of the zone.

    Comment by Bip — January 16, 2013 @ 7:04 pm

  38. Good point, but that’s why I’d also be interested in the pitch type breakdown when it comes to the vertical equivalent of this.

    Comment by Steve Staude. — January 17, 2013 @ 2:12 am

  39. But there’s a difference between “blowing it out” and staying within yourself, maintaining your tempo and trying to work downhill.

    While on ‘average’ the differences between the ‘average’ pitcher in each percentile seems marginal, that’s the ‘average’. If Mr. Petti had used more examples, the differences in edge% and velocity, and general levels of success (I suspect) would be more highlighted. Perhaps I’m wrong.

    This is pretty great stuff. I think the things that pitchf/X developments are allowing the baseball industry to do are remarkable. I like pitch f/X…it’s where metricians and “throwback” types alike can be looking at ‘advanced’ data, yet know that it still fits in easy-applicable and ‘imaginable’ game situations (for those aforementioned ‘throwback types’).

    Comment by Adam — April 22, 2013 @ 8:58 am

Leave a comment

Line and paragraph breaks automatic, e-mail address never displayed, HTML allowed: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>

Close this window.

0.249 Powered by WordPress