Where have the Triples Gone?
Dan Werr steps out of the visitor’s dugout to look at the reduction in triples in the modern game.
Tim McCarver recently made a little splash here by lamenting the
lack of triples hit in today’s game. McCarver suggested that the
decline might be the result of players not running as hard out of the
batters’ box. Well, he might be right? I suppose his next suggestion
will be that there were fewer home runs in his time because players
didn’t hit the ball as hard, though I haven’t heard him bring it up
yet.
Either way, thank goodness for Tim McCarver, because that’s all it
took to get us thinking. There are plenty of explanations for the
increasing rarity of the three-bagger, many of which were discussed in
the associated thread.
The changing dimensions of ballparks can’t be ignored, nor can
improved outfield defense. But posters Ed and Steve Treder brought up
the intriguing subject of left-handed versus right-handed
batters?which led me to look at the transition of the triple from a
power stat to a speed stat.
But before we get to that, let’s take a look at how handedness has
changed over time. The following is the percent of at bats taken by
right-handed, left-handed, and switch-hitting batters by decade. It
should be noted that a few AB by hitters whose handedness was
unavailable were omitted (these amounted to about 2.5% of AB from
1901-1909, and a negligible amount from 1910-1929).
%RHB %LHB %BHB Years
57.4 34.6 8.0 1901-1909
59.5 33.2 7.3 1910-1919
57.6 36.1 6.3 1920-1929
59.1 36.9 4.0 1930-1939
60.9 35.7 3.4 1940-1949
62.4 34.2 3.5 1950-1959
61.9 32.0 6.2 1960-1969
59.2 30.9 9.9 1970-1979
54.0 30.0 16.0 1980-1989
53.6 28.2 18.2 1990-1999
56.0 29.8 14.2 2000-2001
The most obvious pattern is switch hitters declining and then
rising again to higher numbers than ever. In fact, this somewhat
masks the change in left versus right hand hitting over time. So here
is the percentage of AB taken by a non-switch hitter that were taken
by lefties, by decade:
% Years
37.6 1901-1909
35.8 1910-1919
38.5 1920-1929
38.4 1930-1939
36.9 1940-1949
35.4 1950-1959
32.0 1960-1969
34.3 1970-1979
35.7 1980-1989
34.5 1990-1999
34.7 2000-2001
Clearly, there’s been a decline, but not enough of one to explain
the drop in triples. Furthermore, the decrease in triples has been
across the board. Take a look at 3B/AB by handedness, by decade:
Total RHB LHB BHB Years
.0135 .0130 .0152 .0115 1901-1909
.0146 .0134 .0172 .0126 1910-1919
.0141 .0130 .0161 .0132 1920-1929
.0117 .0103 .0140 .0114 1930-1939
.0094 .0081 .0117 .0086 1940-1949
.0084 .0076 .0096 .0097 1950-1959
.0072 .0065 .0083 .0077 1960-1969
.0069 .0063 .0076 .0080 1970-1979
.0066 .0057 .0069 .0090 1980-1989
.0058 .0050 .0066 .0071 1990-1999
.0056 .0048 .0060 .0082 2000-2001
A few facts stand out from the above table. First, as expected,
left-handed hitters always hit more triples than right handed hitters.
Second, all hitters are hitting fewer triples than they used to. But
the most interesting fact is that switch hitters in the first part of
the 20th century were the worst group at hitting triples, but now are
the best at it. This could conceivably be explained as the result of
two phenomena:
- Switch hitters tend to be less powerful and faster than other hitters.
- Triples used to be more of a function of power, and now are more a function of speed.
Each of these ideas can be examined independently of each other.
To determine the validity of the first assumption, we can see how
switch hitters compare to other hitters in a clear power stat (such as
home runs) and in a clear speed stat, such as stolen bases.
HR/AB SB/(1B+BB)
Total RHB LHB BHB Years Total RHB LHB BHB Years
.0040 .0040 .0044 .0031 1901-1909 .1306 .1276 .1352 .1375 1901-1909
.0052 .0045 .0068 .0035 1910-1919 .1237 .1179 .1272 .1509 1910-1919
.0117 .0100 .0151 .0074 1920-1929 .0569 .0534 .0579 .0804 1920-1929
.0156 .0143 .0181 .0123 1930-1939 .0381 .0395 .0348 .0506 1930-1939
.0153 .0140 .0178 .0115 1940-1949 .0355 .0358 .0350 .0348 1940-1949
.0247 .0249 .0249 .0195 1950-1959 .0304 .0307 .0275 .0539 1950-1959
.0242 .0238 .0267 .0153 1960-1969 .0455 .0418 .0449 .0829 1960-1969
.0220 .0226 .0237 .0133 1970-1979 .0644 .0589 .0650 .0924 1970-1979
.0238 .0258 .0251 .0145 1980-1989 .0809 .0700 .0679 .1394 1980-1989
.0280 .0300 .0290 .0203 1990-1999 .0754 .0680 .0713 .0965 1990-1999
.0334 .0332 .0365 .0278 2000-2001 .0652 .0601 .0644 .0854 2000-2001
In every decade, switch hitters have been worse at hitting home
runs than left or right handed hitters, and in every decade except the
1940s, they have been better at stealing bases. Clearly, the concept
of the typical switch hitter as faster and less powerful than other
hitters is valid. This, combined with the change of switch hitters
from the worst to the best triple hitters, supports the theory that
triples now depend more on speed than power.
To confirm the theory, we can compare the correlation between
triples and home runs to the correlation between triples and stolen
bases, by decade. These correlations are limited to hitters with at
least 300 AB.
Correlation with 3B/AB:
HR/AB SB/(1B+BB) Years
.422 .096 1901-1909
.265 .191 1910-1919
.149 .205 1920-1929
.122 .148 1930-1939
.031 .117 1940-1949
-.097 .248 1950-1959
-.090 .259 1960-1969
-.198 .302 1970-1979
-.231 .429 1980-1989
-.291 .394 1990-1999
-.268 .422 2000-2001
On the whole, the correlation between triples and stolen bases has
increased, and the correlation between triples and home runs has
completely reversed. While this doesn’t explain the decline in
triples in and of itself, it certainly shows that a triple today means
something completely different than it used to.
Dan 'The Boy' Werr
Posted: July 30, 2002 at 06:00 AM |
29 comment(s)
Login to Bookmark
Related News:
Reader Comments and Retorts
Go to end of page
Statements posted here are those of our readers and do not represent the BaseballThinkFactory. Names are provided by the poster and are not verified. We ask that posters follow our submission policy. Please report any inappropriate comments.
1. Charles Saeger Posted: July 30, 2002 at 12:37 AM (#605659)* What is the change in triples per ball in play (AB-HR-SO)? Per H-HR? These are probably more of a base of opportunity than just AB, and would explain some of the differences. If players are striking out more, they would logically hit fewer triples.
* By decade, how many triples did the fastest (highest (SB+CS)/(1B+BB+HBP)) players hit? How many triples did the slowest players hit? The best home run hitters hit? The worst home run hitters hit?
* This may be completely worthless, but outfield assists have also been declining throughout the century. Runners may not be trying for the extra base more and more. As home runs have crept upwards throughout the century, people more and more may be stopping at second.
These are all good questions that I'll look into:
What is the change in triples per ball in play (AB-HR-SO)? Per H-HR?
That's an excellent point and something I neglected. I'll try to post that here soon.
By decade, how many triples did the fastest (highest (SB+CS)/(1B+BB+HBP)) players hit? How many triples did the slowest players hit? The best home run hitters hit? The worst home run hitters hit?
Of course, this is similar to the correlations I posted. Do you have a suggestion for the cutoff lines of number of players? Like, say, the top and bottom 20 or 50?
As home runs have crept upwards throughout the century, people more and more may be stopping at second.
Well, another thing to look at might be double rates over time. However, there are probably a great many factors affecting that, of which this phenomenon is low on the list. In other words, I imagine there is an effect, but it's buried under larger ones.
I think that makes sense. As scoring goes up, the value of the out goes up, so the cost of getting thrown out on the bases goes up. The old adage about not making the first or last out of an inning at third is even more true now.
Cheers,
Alan Shank
Double rates have increased, though those do fluctuate over time -- even as a percentage of hits, doubles were down in the 1960's. I'm not sure if there's much there, since doubles are as much a power stat as they are a speed stat.
The lack of strikeout data from 1901-1913 is why I didn't include numbers in the second category for those years.
As can be seen above, the general pattern shows through regardless of noise. Up from the oughts to the teens, down every following decade. It holds true for per AB, per BIP, per H, and per H-HR.
A triple is a different thing now than it used to be, and as such, I wouldn't expect rates to be comparable. It's nice to see the gradual shift though (especially to see it still continuing from the 90's to 2000-01... though I wonder what the effect would be removing 90-92 or 90-93 from the 90s category).
Anyway...Charles, the reason I didn't include HBP data in my speed denominator is that for some reason I thought I didn't have it all the way back to 1901. My mistake. I'm more curious about using SB+CS as opposed to just SB. Are attempts a better indicator of speed? At this scale, I imagine it matters very little, but I'm curious. I can see arguments both ways.
I believe that parks have played a factor in triple reduction. However, I believe outfield defense is a much more important factor.
If you look at Stephen Jay Gould's essay on .400 hitters (and I realize this is the billionth time I've brought it up, but only because it's so useful for this kind of thing), he shows patterns a lot like this one. Basically, one would expect outfield defenders to be getting better, just as athletes are in every other sport. And this improvement typically occurs at a diminishing rate, which is exactly what the above shows (except at the very beginning. I would suggest that that is caused by power improving more quickly than outfield defense, but I could be wrong).
I think the shape and gradual nature of the curve suggest, therefore, that defense is the primary factor. I can easily see where that would lead the transition from power to speed.
I have no doubt, however, that ballparks have also been a factor.
I'll try to get the power and speed leaders and trailers triple rates up soon.
Ed, I don't have the data of how many triples were hit by ballpark, though I suppose a reasonable surrogate would be triples hit by team. However, that would be a fairly serious undertaking.
As always, engaging in questions instead of dismissing them with the simple gut-level explanation has the possibility to lead to further knowledge, though it could also of course lead to wasted time.
The question is not whether such study is useful, because it obviously is, the question is whether the triples rate is a promising path to study, or whether other areas of study might yield more and therefore be better venues for study.
He is not counting singles and walks as speed indicators. The speed indicator is stolen bases per time of first, and he is using singles+walks to approximate times on first.
Gilbert, everything you mention is a valid cause and has probably played a part. Unfortunately, it's having that many factors that makes it so difficult to isolate even one. I wish I had the time and the data to try to control for all of them. Perhaps someone with more resources will sometime.
Pete, I think you're misunderstanding where singles and walks come into the equation. They're not used as indicators of speed, but rather as the denominator for stolen bases. This gives a better indication of speed, because if we just use at bats as the denominator, the rate can be affected by how often the player gets to first. So singles and walks are just there for stolen base opportunities. To give an example from players this year:
A straight SB/AB would show Dunn as speedier, but a SB/(1B+BB) identifies Mondesi, which I would argue is the correct choice. It's still not perfect, but it's better.
Brian, the switch hitting pattern is very interesting. I wonder if it's been looked into much.
GregD, I think looking into triples further is very worthwhile. What I think it means so far is this:
At the beginning of the 20th century, power was low enough that triples were more common than home runs. Power was low enough that triples weren't hit as often as later because of the lack of power. As power increased, both the home run rate and the triple rate increased, because more balls were hit toward the deep outfield and over the fence.
At some point, I'd say in the teens or 20s, hitting the ball that far became routine. There were no longer depths of the park too far out for any hitter. In short, increasing power wouldn't mean too many more triples, because there wasn't a lot of problem hitting balls out to the deep outfield. Furthermore, like all athletic increases, the rate of improvement fell off.
Meanwhile, outfield defense was improving at a high rate. There was very little counter-advancement, because power was increasing more slowly, and players were already hitting to the limits of ballparks. Therefore, defenders weren't subject to the typical pattern of an opposed activity (batting average, ERA, etc.), of following a constant mean while variation in skill shrinks. Instead, the pattern was one of an unopposed activity (fielding percentage, track and field events), where skill follows a curve of improvement at an always-diminishing rate. That's the same curve shown in the triple rate.
As a result, while I think all of the mentioned factors matter, none matter like outfield defense. To put it another way, if every baseball stadium since 1901 had been an exact replica of Atlanta-Fulton County Stadium, I think the numbers would be different, but the curve would look basically the same.
That is just my theory... I can't necessarily prove it, but, as I see it, the data fits it remarkably well.
As I said up above, my concern about this data is that I suspect a lot of other factors have affected doubles. But here it is:
I'm not sure what to make of this. Obviously, when comparing to the triple table, you can see that the vast majority of the above is doubles. You can also see that doubles are through the roof lately (especially considering that triples are at an all-time low). It does seem that this table tracks the basic swing of offense vs. defense fairly well.
But I think there's so much noise here beyond the triple effects...
The effects of outfielders backing up at the close of the dead ball era is greater than you think. It completely changes the risk/benefit balance of trying to get to 3B vs. holding 2B. With Tris Speaker often playing close behind 2B, you're not going to score from 2B on a single up the middle, but you will from 3B. As BAs increased in the 1920s, with outfielders moving farther away from the infield, singles became more likely and more likely to score a runner from 2B.
Many, if not most, errors are muffed ground balls. On those plays, you score from 3B but not from 2B. Fielding pct has been increasing throughout the history of baseball. I assume that the same is true for PBs and WPs, although I don't have the data. Again, these are events which will allow you to score from 3B but not from 2B. As these events became less frequent, runners became less inclined to risk to being thrown out at 3B attempting to stretch a double into a triple.
The strategies in baseball change much more slowly than the conditions. Most contributors to this this site revere Billy Beane for his innovations, but does he value anything Earl Weaver didn't 35 years ago?
The first chart is raw 3B/AB for the top and bottom sixth in HR/AB and SB/(1B+BB+HBP):
And this next table is normalized, so it might be more informative. It's the numbers above divided by their respective decades' overall 3B/AB rate. A number over 1 is a rate greater than the overall rate, and a number less than 1 is worse:
It's certainly interesting that in the 1940s and 50s, all four groups were better than average at hitting triples. I guess that might mean that at that time, both power and speed meant an advantage, but power and speed were inversely correlated.
Anyway, I remain fairly convinced that the smooth, uninterrupted, constant pattern reflects exaclty what you'd expect to see from the defense-based cause I described. It needs to explain why triples have continued to decline in very recent years, why the change has been so gradual and smooth, and why the rate of change has diminished. The defense-based explanation is the only one I see that meets those criteria. That doesn't mean I don't think all the other causes have had a sizeable effect.
Let me try to synthesise and summarise all the perceptive answers and questions I've been reading on this thread & try to put it into two paragraphs with my own conclusion on it...
While park factors and improved outfield defense standards have both trimmed triples frequency, to quote Alan Shank, "As scoring goes up, the value of the out goes up, so the cost of getting thrown out on the bases goes up. The old adage about not making the first or last out of an inning at third is even more true now."
IN GENERAL doubles, triples and homers (extra-base hits) are hits that require first an element of power (with exceptions noted). According to Dan Werr's July 31 table, (2B+3B)/H change remarkably little over decades (20% +/- 2%) But as HRs account for an increasing %age of run-scoring, the incremental advantage (benefit) of advancing from 2nd to 3rd balanced against the risk of being put out (cost) yields a poor benefit/cost ratio. In sum, I think it's not the parks or the OFs, but that some extra base hits that would have been triples are being hit out of the park by Gouldian better athletes, and some extra base hits that would have been triples are turned into doubles by cautious third-base coaches not wanting to risk the out for low benefits.
###
Matthew, thanks for the updated chart. I'm going to try to do a study in which I control for park, and hopefully it will be up here soon (it will be of more limited scope).
Jeff, I think that's certainly part of it, but I still think it's a smaller part, though that might be eminently unprovable. Of course, for both of our explanations, the likely result is more runners staying at second. It would be hard to discern between the two reasons (even if I successfully control for park).
Jim, you make good points, but I think that switch-hitters as a whole are more the key to what I've shown above, and even in the decade where you show one long-term switch hitter, switch-hitters had 34,173 AB. That's certainly plenty of sample size, even if the group was somewhat transient. Also, as I showed above, switch-hitters as a group have always been less powerful and faster than everyone else (with the one exception of the 1940s, when switch-hitters were about league average at base-stealing). So I think it's not right to say there's no set at all; the players might not have stuck around as long as would be ideal, but there were plenty of ABs for meaningful results, in my opinion.
Of course, it's really somewhat beside the point. The switch-hitters data is just an interesting associated data point (and happens to be how I got to here from the original discussion). The same study can exist without the switch-hitter data in there at all, but I thought it was interesting and relevant.
As I said above, I'm undertaking a project currently where I account for park. Hopefully, I'll get it out soon.
Just glancing at Matthew's chart, and with an acknowledgement of the limitations of it, it appears for example that the 40s were actually a better triple-hitting environment than the 30s, yet the decline in triples was quite severe from the 30s to the 40s. Also, since the 70s, and even more so, the 80s, there appears to have been little change in size of parks, but triples keep decreasing anyway.
Hopefully, I can get something going here soon... I'm not sure how much data I can scrap together, but it will be something.
You have not been reading. He is not using walk rates to show speed, but rather to provide an opportunity for a stolen base.
Most triples are hit to right field, or to right-center. There is more area in right/right-center now than there was at the turn of the century, and less in left/left-center. There is less area in straightaway center, which would cut the number of triples, but right field is the key. Plus, we have artifical turf now -- not as much as in the 1970s and 1980s, but the turf increases the number of triples.
The greatest triples park ever (by reputation) is Forbes Field, which was deep in both right and left fields. (The 1969 Retrosheet data does confirm Forbes's high-triple reputation, increasing triples 84% as a percentage of AB-HR-SO and 72% as a percentage of H-HR.)
I could well be wrong, but the totals above look like we should be awash in triples today.
I don't agree that the effect of outfielders playing deeper towards the close of the dead ball era is "greater than think," since that IS what I think. The closer the outfielders play to the infield, the easier it is for balls to go over their heads (or get between them and roll towards the wall), the longer it takes for them to run the ball down, and the more time the batter has to run around the bases. When the lively ball came into play, it forced outfielders back, and one result was a reduction in triples. Maybe I'm just dense, but I really don't see anything much more complicated than this going on. The deep center fields are an additional factor, but if the outfielders are all playing deep, this will have much less effect than it would when they were playing further in. Whatever changes there are in switch hitters, RH vs LH batters, etc., is insignificant compared to this basic factor.
Andy,
I could accept that explanation if it were just one event... triples up during deadball, and down afterwards. But the decline has remained consistent. For example, triples have declined steadily every five-year period from 1975 to now. I don't suspect that outfielders have been steadily moving back a little further every five years. However, I would expect that every five years, outfield defense is a little better.
Charles,
Just out of curiosity, where are you getting your park factors?
I don't think this is quite right, Dan. The curve would clearly be flatter.
There should be some way for you to isolate the elements involved in the decline of triples. It would take a lot of work to do it, because you'd have to get park factors for triples for every ballpark, and that info (as Charlie will tell you) hasn't been compiled yet.
Possibly one approach would be to look at the range of variation in triples per team and look at the extremes. Examining the park configurations for those teams might lead you toward a way to determine how much of the effect is based on defense, and how much is based on the movement toward standardization of stadia in the 60s.
T/BIP rates for ballparks would be an interesting corollary to the data presented here. One question that would be interesting to answer is whether the handedness element is distributed evenly or if it?s skewed by park. IOW, do the generous triples parks of the present (Coors, Comerica) and past (Forbes) provide more of a haven for RHB, upping their T/BIP averages, or is it uniform across all parks?
BTW, one interesting adjunct to our ways of characterizing outfield defense can be seen in H/A splits on triples. For example, so far in 2002, the Mariners have allowed only three triples at home while hitting 18 themselves. On the road, they've hit five and allowed six. Since triples are so scarce, the data is subject to distortion if we take it too literally, but it's interesting nonetheless.
I don't think this is quite right, Dan. The curve would clearly be flatter.
Yes, it undoubtedly would. The shape would be the same, or very similar (if you stretched out the Y-Axis to make it higher, that is).
There should be some way for you to isolate the elements involved in the decline of triples.
As I've noted above, I've undertaken the beginnings of a study on triples (now in the hands of the administration) that isolates for park. The gigantic limitation of this study is that it only goes back as far as 1975. However, it's interesting nonetheless.
Possibly one approach would be to look at the range of variation in triples per team and look at the extremes.
A good idea... still noisy, though. But if a team is consistently extreme, especially when the roster has completely (or nearly) turned over, that'd be a good indicator. The other problem I've run across is that stadiums have changed like crazy. You'd want to minimize that.
BTW, one interesting adjunct to our ways of characterizing outfield defense can be seen in H/A splits on triples.
Yes! I noticed that in a big way. Of course, while the Mariners having Cameron and Ichiro's speed in the OF and at the plate probably helps in that regard, it can't be discounted. To get the park-independent numbers, I combined home and visitor numbers, and every single time, the home numbers were better. I don't know why... maybe the home OF have a better sense of how a ball will interact with their home park's walls... which way it will ricochet, etc.
Why do I mention this? 3b PO are mostly popups and line drives. If the left fielder is grabbing those plays, then he is probably playing closer to the infield. (Or the ball could be traveling farther, which isn't so outrageous; the pitcher appears to be taking popups from the catcher.)
You must be Registered and Logged In to post comments.
<< Back to main