FanGraphs Baseball

Comments

RSS feed for comments on this post.

  1. Love this! Terrific stuff.

    Comment by WilliaminMaine — January 12, 2012 @ 1:33 pm

  2. Do these graphs represent the average of all players in a given age group, or are these the average differences between one age group (the top one) for each player seperately? In other words, is this saying that the average player has an ISO .020 lower when he is 29 than when he is 24; or is it saying the average 24 year old has an ISO .020 higher than the average 29 year old?

    Comment by TK — January 12, 2012 @ 2:28 pm

  3. This new data contradicts what I read in the old Bill James books. Peak performance was 27, with power peaking a year or two later.

    Comment by David — January 12, 2012 @ 2:37 pm

  4. From my understanding, it’s the average of players that played in both age brackets. But Zimm used the delta method to weight the numbers, which is in the link below the first graph.

    Comment by Eno Sarris — January 12, 2012 @ 2:37 pm

  5. peak performance at 27 is consistent with this, but power is different. I tend to think that GB% and ISO linking up at 25 lends some strength to this.

    Comment by Eno Sarris — January 12, 2012 @ 2:39 pm

  6. It is the summed weighted (more PA, more weight) difference of each player from one age to the next.

    For the 24 to 25 year bracket, I took all the hitters that played when 24 and 25, took the difference of the stats, weighted it to the harmonic mean of their PAs and then found the average.

    Comment by Jeff Zimmerman — January 12, 2012 @ 3:02 pm

  7. peak performance has also been found to peak at a younger age. The above link has the “WAR” aging curves.

    Comment by Jeff Zimmerman — January 12, 2012 @ 3:04 pm

  8. IIRC, other studies with superficially similar methodology have shown the power peak at around 29. Why were these not cited, with some time spent explaining why the present methodology is better?

    Comment by dcs — January 12, 2012 @ 3:36 pm

  9. “Power” (ISO) includes, of course, extra bases on doubles and triples, which depend to some extent on speed, which peaks early. HRs, the simpler measure of power, are created by strength alone with no speed component; they peak later – at least that’s widely believed.

    Comment by Mr Punch — January 12, 2012 @ 3:47 pm

  10. there’s this too: http://www.insidethebook.com/ee/index.php/site/article/best_fit_equations_for_component_aging_curves/

    I’m curious which to believe now.

    Comment by slash12 — January 12, 2012 @ 3:52 pm

  11. Do you recall what Bill James meant by “power”? HR? SLG? ISO?

    If he was talking about HR or SLG, then it could still fit with the above graphs. Younger players hit more doubles and triples based on speed. As they age, some doubles clear the wall, the leg doubles are now singles, and the triples are far less frequent. ISO could in fact go down over time while HR% remains consistent or even goes up. SLG could go up, too, with more singles due to fewer Ks.

    Just a theory. Point is that the best indication of power is debatable.

    Comment by mentalmeat — January 12, 2012 @ 4:00 pm

  12. IsoP is going to be strongly influenced by K%, which peaks at the same time. More contact = more power. If you want to look at power independent of contact rate, you should use TB/H

    Comment by Yirmiyahu — January 12, 2012 @ 4:07 pm

  13. Are these major league numbers, or major and minor league numbers? Most 21 year old players are obviously not in the major leagues and are facing very different conditions.

    Comment by Breadbaker — January 12, 2012 @ 4:08 pm

  14. That’s why wine is the one used in sayings by older men verifying their remaining veracity.

    “Veracity”? I had no idea wine of any age had any correlation with truthfulness. (Wine consumption, on the other hand, likely has an inverse relationship at best… with both veracity and virility)

    Comment by joser — January 12, 2012 @ 4:09 pm

  15. I wonder if the time period throws off the data.

    If it goes back to 2002, we’re getting some impressive old-folk seasons, aren’t we?
    Bonds hit .370 with 198 walks that year, for example.

    I’d be curious about 1980-1990 (prior to the steroid era).

    Comment by bSpittle — January 12, 2012 @ 4:09 pm

  16. oops wrong word haha, read right over that twice

    Comment by Eno Sarris — January 12, 2012 @ 4:10 pm

  17. Major league only, but the sample is ok for 21 year olds cause this is done by PA. Or ‘better than might be expected.’

    Comment by Eno Sarris — January 12, 2012 @ 4:12 pm

  18. Isn’t the drop in ISO due to an increase in AVG from a lower K%? That is, Stanton could hit 40 homers when he’s 23 and 45 homers when he’s 27, but his ISO would drop if he strikes out significantly less.

    I would like to see the above graph with SLG included, so better see what’s going on.

    Comment by skemmis — January 12, 2012 @ 5:35 pm

  19. Exactly what I was going to say.

    Comment by Phils_Goodman — January 12, 2012 @ 6:32 pm

  20. I’m confused about this chart. The axis says “change” and is always negative, which would mean the peak is at age 21, and regression simply slows down at older ages. Should I read these as if the % changes start at zero (instead of -2.3% for walks, for instance)?

    Comment by Kinanik — January 12, 2012 @ 6:50 pm

  21. I’m a bit confused. Isn’t this data biased by age of promotion to the majors? Players who debut at early ages are expected to produce much better age-26 seasons than players who are rookies at age-26, right?

    Comment by SeanP — January 12, 2012 @ 6:51 pm

  22. the relationship between avg and k% is not linear enough to support this I don’t think. Adding in Hr/FB could perhaps suss out a little more of the difference you and others are talking about I believe.

    Comment by Eno Sarris — January 12, 2012 @ 8:09 pm

  23. Think of the %s as ‘cumulative change from the peak’ since the top axis is zero and is the peak. It’s a little tricky on K and GB, because decreasing is a good thing, but treat the plateau part as the peak in order to read this correctly.

    Comment by Eno Sarris — January 12, 2012 @ 8:11 pm

  24. Players are only compared to themselves.

    Comment by Eno Sarris — January 12, 2012 @ 8:12 pm

  25. Yes I think perhaps adding HR/FB or TB/H would be interesting.

    Comment by Eno Sarris — January 12, 2012 @ 8:13 pm

  26. Was wondering the same thing.

    Comment by Chris — January 12, 2012 @ 8:25 pm

  27. That’s ok, it seems to be a FanGraphs tradition (and for some reason it falls to me to be Vocabulary Nazi)

    Comment by joser — January 12, 2012 @ 8:26 pm

  28. This is bookmarked. Great reference.

    Comment by Frank Campagnola — January 12, 2012 @ 11:00 pm

  29. great point.

    Comment by bstar — January 12, 2012 @ 11:42 pm

  30. Yes, this first graph took a lot of time for me to digest, and I’m pretty good with that kind of stuff.

    Comment by bstar — January 12, 2012 @ 11:45 pm

  31. So your saying the sample might be elevating the power of older men. in other words, the fall off could me even faster on power than this graph? Interesting!

    If that proves to be true then we may never have another 500 home run hitter.

    Comment by kick me in the GO NATS — January 13, 2012 @ 1:34 am

  32. I can’t help but feel like the power numbers yare skewed a bit because of young players that get called up for a look and keep getting time because of hot streaks. Playing time is “earned” more by younger players than established players. At least that’s my impression. Is this corrected for?

    Comment by Jon S. — January 13, 2012 @ 1:46 am

  33. Great analysis!

    I’m going to point to this whenever people start to bring up Prince Fielder around me as a viable DH.

    Comment by Michael F — January 13, 2012 @ 2:02 am

  34. I’ve always wondered that, “difference from peak” might be a good axis label.

    Comment by Barkey Walker — January 13, 2012 @ 9:07 am

  35. There’s likely two types of young “power” hitters to consider: speedier guys heavy on 2B and 3B, and sluggers heavy on HR. The early ISO peak may be largely due to the first group as the length of their non-HR hits lessens and they get more hits.

    For example, Pujols’ peak extra base hits:
    Triples = 4 at age 21, 4 TOTAL since his 26th birthday.
    Doubles = 51 at age 23-24
    Homers = 49 at age 26.

    Comment by Dan — January 13, 2012 @ 9:57 am

  36. So does this mean Anaheim is paying $240 million for 10 years worth of moldy cheese?

    Comment by Dustin — January 13, 2012 @ 11:07 am

  37. I agree with this post and it is good analysis, but it sounds eerily similar to a comment i made on another Eno Sarris post over at rotographs

    http://www.fangraphs.com/fantasy/index.php/upton-gonzalez-and-mccutchen-tier-two-nl-outfielders/

    could i have been your muse eno???

    Comment by Pat Golden — January 13, 2012 @ 11:07 am

  38. Looking back at the figure, I’m really confused now, why are there two years on the x-axis if this is deviation from peak on the y-axis?

    Comment by Barkey Walker — January 13, 2012 @ 11:15 am

  39. Can you add another line to show Bonds’s career? Just for fun.

    Comment by Nathan — January 13, 2012 @ 11:40 am

  40. Other commenters have pointed out that ISO is probably biased due to reduced K% and higher 2B and 3B totals from younger players. I’d be reluctant to say that a 26 year-old like Kemp will likely see a decline in power.

    I did this quick check: there have been 42 seasons of 50 or more home runs. Only four of them were by players of age 24. Only one player did it younger than 24 (Prince in 2007 at age 23). That’s 5/42 by players 24 or younger. And, only two seasons by 25 year olds, 7/42 by players 25 or younger. Conversely 35/42 were by players 26 or older.

    If we include seasons of 48 or 49 HRs, we get 30 more seasons in our sample. None of those were by 24 year-olds, and one was by a 23 year old (McGwire in 87). Again, two seasons were by 25 year olds. So 3/30 of the 48/49 HR seasons were by players 25 or younger.

    Only 10/72 of the best HR seasons were by players 25 or younger. Even if we exclude ~30 seasons due to steroids (in which case we have to throw out A-Rod and Beltre’s 25 year-old season), we get a number of 8/42. Not great evidence that the power peak is 24-25.

    Comment by Dan — January 13, 2012 @ 1:27 pm

  41. Since the graph is cumulative, this would only be true if you only used players that played until at least age 39.

    Comment by Barkey Walker — January 13, 2012 @ 1:58 pm

  42. Over at Baseball HQ, they did a study last year that suggested peak has more to do with when a hitter arrives in the majors than it does with his actual age. Their eventual suggestion is that a hitter takes a couple years to get acclimated and then performs at his best. Their assertion was that we only see a peak in age because of the correlation between age and when a guy reaches the majors. For example, most players enter at 24 so we think 26 is the peak but it’s only because the largest number of players have been in the majors for two years by then.

    It’s a pretty radical theory and one I’d love to see explored here at fangraphs.

    Comment by bjoak — January 13, 2012 @ 2:35 pm

  43. because of playing time issues with young players, I wouldn’t use HRs to suss this out further. Hr/FB maybe.

    Comment by Eno Sarris — January 13, 2012 @ 2:38 pm

  44. I think the fact that BB and K peak later jives with their idea in a way. I still think our athleticism peaks way earlier than our learned skills, so it’s possible that both theories are true: our bodies peak early, but our approach to the game gets refined the longer we play, until our bodies begin to fail us.

    Comment by Eno Sarris — January 13, 2012 @ 2:39 pm

  45. Gosh, I’d argue that Kemp will see a decline but it will have a lot more to do with simple regression rather than his age.

    Comment by bjoak — January 13, 2012 @ 2:39 pm

  46. Certainly seems true with pitchers who, as a group, see velocity peak at 21 or something but tend to have their best years later.

    Comment by bjoak — January 13, 2012 @ 2:43 pm

  47. What, not even a hat tip? This article reeks of a month of research built off of my assumptions in regards to Justin upton and his aging curve…

    Everything you’ve written here today flies in the face of your expectations and responses for upton in the article I cited

    Be a man and own up to running with my idea

    Comment by Pat Golden — January 13, 2012 @ 3:33 pm

  48. haha, I’ve been talking to Zimm about this for like three months. The real genesis was thinking about Stanton’s strikeouts — but I will admit to thinking about that Upton post and your comment when I was looking at the young power custom leaderboard linked above for sure.

    Comment by Eno Sarris — January 13, 2012 @ 3:45 pm

  49. you’re generalizing too much…that’s not really what the prediction is. he’s saying that the average falloff might start earlier and be steeper. it says nothing about special cases…like HOF players. recall, prior to the steroid generation, there had only been fifteen hitters in 90 some years with 500 HRs. even if you include josh gibson, adjust for improvements in medicine, etc etc it’s always been and will remain a difficult club in which to gain entry.

    Comment by ole custer — January 13, 2012 @ 4:07 pm

  50. I really like these types of analysis, but I also think that just looking at ISO is highly flawed. Half of this article is dedicated to demonstrating that hitters become better at contact and plate discipline until they are in their late 20′s, and then the latter half of the article attempts to correlate a drop in ISO with a drop in power. In my opinion, this approach lacks significant measures of power, and simplifies things in a way that skews the conclusion.

    As an example, let’s take the case of power pitchers. There was an article recently on Fangraphs demonstrating that pitchers generally have a K% peak in their early 20′s and then decline from there. This could be due to several factors. One, their stuff gets worse, they lose speed on their fastball, and hitters figure them out. If that is true, then essentially all pitchers are at their best very early in their career. But we know that to generally not be the case. The best pitchers tend to be in their late 20′s or early 30′s. How can that be, if their K% is steadily dropping? Probably because they are making subtle changes to their pitching patterns and approach to hitters that decreases overall pitch counts, results in weaker contact, and produces more outs per pitch thrown. Does that mean they are less effective? No, it does not. In fact, it could be argued that they are still capable of striking out more hitters, but elect to take a more efficient approach to creating outs.

    Now back to the topic at hand. If batters are becoming better hitters through their late 20′s that would generally mean that they are increasing their OBP, SLG, AVG and other basic parameters used to measure hitter effectiveness. Since ISO is a simple subtraction of batting average from slugging, it would stand to reason that as a hitter raises their batting average, their ISO is going to be effected in one way or another. At-bats that used to end with a strikeout or weak pop up may now result in a single or walk. Thus, measuring power based purely off ISO is not going to tell you if a hitter is more or less “powerful” year to year.

    I’m not a mathematician or a statistician, and I know my own argument has it’s flaws. But ultimately if we’re going to do this the right way, we need a few more angles to determine peak power for hitters based off of age.

    That is all.

    Comment by Adam G — January 13, 2012 @ 4:40 pm

  51. Why doesn’t someone just make ISO with the weights from wOBA? That would make more sense than ISO.

    I still think ISO is helpful for evaluating how much a power a hitter is perceived to have by pitchers, fans, writers, and announcers. This is supported by the fact that the average person doesn’t have many problems with how ISO is weighted. In the context of this article, however, we care more about how good these hitters actually are rather than how they are perceived.

    Comment by Bryce — January 14, 2012 @ 12:41 pm

  52. “That’s a pretty definitive peak for power”

    No, it’s not. Add the standard deviation among hitters to get a sense of how much variation there is from person to person. I suspect that the “distinct peak” will become a 4-5 year statistically equivalent plateau. Without std dev or std error bars the graph is essentially useless.

    Comment by Stats — January 14, 2012 @ 5:05 pm

  53. sobs silently

    Comment by Michael Young — August 9, 2012 @ 8:34 pm

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>


Current day month ye@r *

Close this window.

0.126 Powered by WordPress