Effect of the Home Run Derby on Second Half Performance (Pt.2)

In part 1, I looked at whether significant experience as an all-star or in HRD contributed to second half drops in performance.  Although it does, it's more likely that the inexperienced players are regressing to the mean after flukey first halves.

Here I want to look at whether the number of swings taken affects the second half drop off, to the extent there is one.  The swing counts are approximate, based on 10 outs per round and some info I found about "swing offs" and abbreviations (like when Josh Hamilton in 2008 quit swinging in the second round after 4 outs, because he had already advanced to the finals).

Of the 79 participants in the set, 39 took fewer than 20 swings.  These are pretty much the folks who bow out after the first round.  So that's one "tier".  There are no obvious breaks in the number of swings among the remaining 40 players, so I set the second tier at 21-41 swings, and the third tier at more than 41.  That also happens to give me 20 players in each of the final two tiers.

First tier (0-20 swings):


This group had a higher K% in the second half by 1.5%, lost 44 points of OPS, lost 34 points of ISO and had a reduced HR/Contact% by 1.1%.  BABIPs rose by only 6 points.

Second tier (21-41 swings):


This group had a higher K% in the second half by 0.6%, lost 21 points of OPS, lost 16 points of ISO and had a reduced HR/Contact% by 1.0%.  BABIPs rose by only 4 points.


Third tier (42+ swings):

This group had a higher K% in the second half by 0.9%, lost 25 points of OPS, lost 12 points of ISO and had a reduced HR/Contact% by 0.2%.  BABIPs fell by 4 points.


Conclusions:


Well I know what it doesn't mean.  The HRD does not wear out players.  The guys who took the most swings have negligible reductions in their stats that are consistent with what we found "experienced" players experienced in the second half.  And in many cases, the guys with more swings have less of a drop in the second half; and as explained below, some of them get significantly better.

Moreover, the middle group feels the impact in the second half less than those who take fewer swings in Tier 1.  Swinging more in the HRD is not adversely affecting second half performance.

What's most interesting is that the Tier 1 players have stat reductions in the second half that look almost identical to the stat reductions for the significantly inexperienced all-stars discussed in my first post!  Yet of the 39 players in Tier 1, only 21 fall into the inexperienced category.  It's almost an even split in Tier 1 among experienced and inexperienced players.  (For comparison, 60% of the players in Tier 2 are experienced by my filter, and 50% in Tier 3).  Moreover, the second half differentials for the experienced group in Tier 1 and the inexperienced group in Tier 1 are pretty much the same:  OPS, ISO and BABIP changes are within 1 or two points for both.

Does it get into the Tier 1 group's head that they didn't do so hot in the HRD?  If that's the explanation, you would think the experienced Tier 1 players would shake it off better; but that's not what the numbers say.

Or is it physical/mechanical:  that is, taking fewer hacks trying to hit homers is more likely to mess up your swing than taking more. Maybe taking more swings makes it more likely that your altered mechanics at the beginning of the HRD resolve to your natural rhythm by the time you've taken at least 20 hacks.  And maybe only taking 15 home run swings is just enough to throw you off, but not enough to get you back in rhythm.

The impact is still pretty small, but it seems likely that those who don't make the semis will be more affected than those who do.

Despite all that Tier 1 talk, the worst subset to be in is the inexperienced player taking more than 41 swings (Tier 3).  Those players had their K% jump 2.9% in the second half, lost 32 points of OPS, 37 points of ISO and 1.4% shaved off their HR/Contact%, and those numbers are with a boost in BABIP of 18 points!

By contrast, the experienced players in Tier 3 are the most stable.  They shave off 2% points from their K%, increase OPS by 177, increase ISO by 104 and increase HR/Contact% by 4.3%!  Some of that is because of a BABIP boost of 14 points (and a small sample size...there are only 10 of them).

If you play fantasy baseball, odds are that you won't dump any experienced player just because of HRD.  But if you have an inexperienced player who had a hot first half, made the All-Star team and is participating in HRD, you might be a little worried.  Based on these numbers, ideally those guys (Hart, Swisher and Young) make it past the first round and lose in the semis.  The worst thing that can happen is for those guys to make the finals and take a bunch of swings (think Bobby Abreu and Garret Anderson).

For the experienced players, the further they go in the HRD, the better off you'll be.  :)

Effect of the Home Run Derby on Second Half Performance (Pt.1)

There's a lot on the Web today about whether participation in the All-Star Home Run Derby (HRD) affects second half performance (theoretically, by messing up the player's swing).  Many have commented that the second half performances are just regression to the mean, because players who make the all-star team are often already playing over their heads.

The bloggers at Pinstripe Alley, a Yankee fan blog, looked at the first and second half OPS of all HRD participants since 2000 (except IRod in 2000 because he barely played in the second half).  Their conclusion was that the OPS dropped .042 points, which they described as "not huge".

I wanted to slice the numbers a little differently, using essentially the same data set that Pinstripe Alley used.

I wondered whether HRD participants with significant all-star game or HRD experience were affected differently than inexperienced players.  I categorized experienced players as those who, at the time of the HRD appearance in question, had made the all-star team at least twice before or had participated in an official HRD before.  Of the 79 players in the data set, 40 were experienced using that filter, and 39 were not.  That worked out pretty nicely.  :)

I used 5 stats:

Strikeout percentage (SO/AB):


Experienced players had their K% increase from 19.2% to 19.9% in the second half, for a 0.7% increase.  A difference not worth commenting on further.

Inexperienced players had their K% increase from 20.3% to 21.8% in the second half, for a 1.5% increase.  It's not a huge difference from first half to second half, but the difference is twice as large as for experienced players.

OPS (OBP+SLG):

Experienced players had their OPS decrease from 1.011 to .987 in the second half, for a .024 decrease.  That's not a big difference.

Inexperienced players had their OPS decrease from .956 to .913 in the second half, for a .043% decrease.  Again, not a huge difference from first half to second half, but the difference is almost twice as large as for experienced players.

I also wondered how much OPS was affected by high first half BABIPs and regression to the mean in the second half.  As an average, neither group had a significant BABIP change from first half to second half.

Nine (9) of the experienced players had their OPS drop 100 points or more, and five (5) of those had their BABIPs drop 25 points or more.  By contrast, fifteen (15) of the inexperienced players had their OPS drop 100 points or more, and nine (9) had their BABIPs drop 25 points or more.  So inexperienced players were hit more dramatically, but essentially the same percentage had significant BABIP adjustments as a contributor.

Five (5) of the experienced players had their OPS rise 100 points or more, and three (3) of those had their BABIPs rise 25 points or more (and the other two had rises of 14 and 19 points).  Only three (3) of the inexperienced players had their OPS rise 100 points or more, and all of them had a rise in BABIP of 25 points or more.

BABIP seems to be a partial explanation for significant drops in OPS in the second half, but seems to be a more significant factor to significant OPS gains in the second half.

ISO (SLG-AVG):

Experienced players had their ISO decrease from .293 to .279 in the second half, for a 14 point decrease. No big deal.

Inexperienced players had their ISO decrease from .277 to .243 in the second half, for a 34 point decrease.  More significant, but again, the difference is two and a half times as large as it is for experienced players.

BABIP:

Neither group experienced a significant change in BABIP.  The inexperienced players actually got a little 5 point BABIP bump in the second half.  Both groups started with substantially equivalent BABIPs.  See note above under OPS.

HR/Contact Percentage (HR/(AB-SO)...essentially, how many balls contacted flew out of the park):

Experienced players had their HR/Contact% decrease from 9.3% to 9.0% in the second half, for a .3% decrease. No significance at all.

Inexperienced players had their

HR/Contact% decrease from 8.7% to 7.2% in the second half, for a 1.5% point decrease.  That strikes me as more notable, but again the most glaring difference is that the change is five times bigger than for experienced players.  

Conclusion

One could conclude from the numbers above that the effect on second half performance is not tremendously significant for any HRD participant, and is basically non-existent for the players I've named as "experienced."

On the other hand, the difference between "experience" and "inexperience" is probably just regression at work.  Consider that everyone in the experienced data set was appearing in at least their third all-star game, with one exception of a player who participated in HRD twice before ever making his third all-star appearance.  So for the most part, these guys are true all-stars and less likely to be in the game because of a statistical fluke.  

The so-called inexperienced players have more trouble in the second half because their first half performances were surprises.  Brandon Inge in 2009's HRD was the best example.  He went from having the best first half of his life, to being literally one of the worst hitters in MLB during the second half (how do you like a OPS of 541?).

It's not that all-star game or HRD experience makes a player seasoned enough to not let his swing get screwed up by HRD. It just means those are the really good players -- the perennial all-stars -- who naturally have more consistency.  After all, the first half numbers of experienced players are better than the inexperienced players across the board...lower K%, a full 60 points better on OPS, 16 points better on ISO and a higher HR/Contact %.

I have a quick part 2 to post, because I also sliced the data by the number of swings the player took in the HRD.  To come...

Griffey as Home Run King?

I was thinking the other day how nice it would have been if Griffey was the all-time home run leader instead of Bonds. Not that I hate Bonds. I just like Griffey so much more. So smooth in the outfield. Such a beautiful swing.

Griff stands at 593 homers. I know it is a big part of the accomplishment to stay healthy. Bottom line, Griffey didn't do it. Still, I thought it would be fun to try and project what his home runs would have been without the injuries.

Griffey's principal injury years were 94-95 with Seattle, and 01-04, and 06 with Cincinnati. To project the number of games he should have played in the injury years, I'll average the healthy year prior to the injury year, and the healthy year after the injury year, and assume the same rate of plate appearances and homers as during the injury year.

Year      Actual HR     Revised HR
89           16             16
90           22             22
91           22             22
92           27             27
93           45             45
94           40             52
95           17             35
96           49             49
97           56             56
98           56             56
99           48             48
00           40             40
01           22             27
02            8             16
03           13             34
04           20             33
05           35             35
06           27             33
07           30             30
            ----          ----
            593            676

That's 83 more homers. He would only need to hit 87 more to pass Bonds, if Bonds retires at the end of the year. And Griffey is five years younger than Bonds.

Alas, wishin' don't make it so. Something to keep in mind when you see that A-Rod is "on pace" to break the all-time record.

Weird stat: Griffey has grounded into more double plays than Bonds, despite 3 fewer seasons and all the injuries. That's the advantage of Bonds' high walk rate.

Weirder stat: A-Rod has grounded into only 3 fewer double plays than Griffey, and more than Bonds, despite being only 31.

500 HR, 300 Wins or 755 HR?

Which is the toughest? I don't mean which is the toughest to achieve in a career. That's obvious.

I mean, if you are Rodriguez at 499, Glavine at 299 or Bonds at 754, how hard is it to get that one extra notch?

I'd say Rodriguez has it easiest. He's the youngest. He's going to get #500 and may eventually get 755. It's just a matter of time for him. He ought to take the same approach at the plate as always. It's a great accomplishment, but it will pale in comparison to other records he will eventually achieve.

Bonds has more media scrutiny, by far, and he's also the one of this group that is least likely to have fan support for his effort. That's some serious pressure. That pressure increases in San Francisco, because he wants to hit it for the home fans. But, like Alex, Barry's going to hit at least one home run before its over. He is going to break the record. So it's tougher for him than Rodriguez.

But I'd argue Glavine has it the toughest. Of the three, his skills have declined the most. He's hanging on as a 4th starter. In addition, he has less control over the situation. He can pitch great and get no run support. His defense can let him down, etc. It is likely he'll get the win, but it is possible that he strings together a bunch of no decisions or losses. The risk of injury is also higher for a pitcher.

I predict they happen in that order: Rodriguez to 500; Bonds to 755; Glavine to 300.

395 feet vs. 380 ft.

After Guillen and Castilla's complaints about the outfield distances in RFK Stadium, I started thinking about how much an effect 15 feet might have. Someone mentioned a memory that the Astrodome moved its fences a couple of times and it drastically affected the hitters.




I looked at the Astrodome's dimensions over the years, using the data at Ballparks.com. In 1965, the first year of the Astrodome, the fences were, left to right 340-375-406-375-340. For some reason, they lengthened the power alleys by 15 feet, so the dimensions from 1966-1971 were 340-390-406-390.




From 1972-1976 they shortened all the dimensions to 330-378-400-378-330. Then from 1977-1984 they went back to the same dimensions as 1966-1971. Then in 1985 they shortened again to the dimensions from 1972-1976, where they stayed until 1992 when they were reduced a little to 325-375-400-375-325.




In sum, from 1966-1971 and from 1977-1984, the dimensions were longer: 340-390-406-390-340. From 1972-1976 and from 1985-1992, the dimensions were shorter: 330-378-400-378-330 (with a slight reduction in 1992).




Ignoring the first year of the park (1965) and the first year of the new dimensions (1966), the above data provides 13 years of play with the dimensions twelve feet longer in the power alleys (the "Long Years") and 13 years of play with shorter power alleys (the "Short Years"). The 378/390 switch isn't a perfect match for the 380/395 distinction that Castilla and Guillen were complaining about, but it's probably close enough. One of the reasons the Astrodome is a a great example -- apart from the similar power alley dimensions -- is that weather is a non-factor.




Here is how the Houston team hit at home and on the road during the seasons with longer and shorter power alleys. First let's look at how they did on the road during those years (season averages):




Road

Seasons      AB    H     2b     3b    HR
Long Years  2751  688   113     18    53
Short Years 2793  697   111     17    60

The road numbers are remarkably similar for all the years from 1967-1992, so we need not be too concerned about whether the quality or style of the various Houston teams in those 26 years are skewing the home data we are about to see. Because the road numbers are essentially neutral, they will not be of further use below. But in general, it was much much harder for the Astros to hit homers in the Astrodome compared to road parks, and much easier for the Astros to hit triples -- and to some degree doubles -- in the Astrodome compared to road parks, regardless of the Astrodome's dimensions.




Now the home data for the Astros (season averages):




Home

Seasons      AB    H     2b     3b    HR
Long Years  2586  655   114     32    29
Short Years 2684  678   120     21    44

It's a little difficult to tell the effect of the dimension changes by looking at the raw data, articularly since the batting averages are .253 for both the Long Years and Short Years, which might lead to the conclusion that the longer dimensions did not affect hitting. So let's look at some rates.




Home

Seasons      H/2b  H/3b   H/HR    TB/H     
Long Years   5.73  20.68  22.96   1.40
Short Years  5.63  32.53  15.28   1.44

Now the differences become more obvious:




Doubles: During the Long Years, one out of every 5.73 hits was a double and in the Short Years one out of every 5.63 hits was a double. So, the dimension changes did not appreciably affect doubles.




Triples: During the Long Years, one out of every 20.68 hits was a triple and in the Short Years one out of every 32.53 hits was a triple. So, lengthening the dimensions increased triples by 57%.




Home Runs: During the Long Years, one out of every 22.96 hits was a home run and in the Short Years one out of every 15.28 hits was a home run. So, lengthening the dimensions decreased home runs by 33%.




Total Bases: During the Long Years, the Astros got 1.40 total bases per hit and in the Short Years got 1.44 total bases per hit. So, the dimension changes only slightly changed total bases per hit.




Putting it all together, lengthening the power alleys (and to some extent the foul lines and center field) shifts home runs to triples.




Looking at Jose Guillen's numbers at home in RFK, he has no triples and one home run. Kind of hard to say some of his would-be homers have been reduced to mere triples when he has no triples. As I mentioned in an earlier post, using a park home run adjustment analysis, RFK suppresses home runs by about half. So based on the park analysis and the Astrodome study, at most Guillen might be able to claim he should have one additional home run at RFK. (The other complainer, Vinny Castilla, has 6 homers at home, and one triple, so he may have a claim for another 3 homers. What is odd is that Castilla has only one homer on the road in approximately the same number of at-bats).





Side note on the Houston teams:




Over baseball history, singles have been worth approximately .46 runs, doubles .80, triples 1.02 and homers 1.4. Applying those to the Long and Short Year average seasons, you get approximately 385 runs for the Long Year seasons and 406 runs for the Short Year seasons, or about a 22 run per year disadvantage with the longer dimensions. That means on the hitting side, they should have been about 2 wins worse during the Long Years, all other things being equal.




Generally clubs build their teams to suit the stadium, so during the Long Years, so it is possible the Astros may have had a little less power and more speed, and in the Short Years may have had more home run hitters. However, the road data for the Long Years and Short Years is nearly identical, indicating on balance that the Houston teams in the Long Years and Short Years were equally capable in terms of power.




Houston did not have a significant number of power hitters in the Long Years. The Toy Cannon and Doug Rader were true power hitters, and combined for 5 seasons of 20+ home runs during the Long Years. The Astros also had Bob Watson, Dickie Thon, Joe Morgan and Cesar Cedeno during the Long Years, who had 9 seasons of 20+ home runs in their careers, but only two for Houston during the Long Years. Watson and Thon, strangely, had their only 20+ homer seasons during the Long Years -- and might be stretching the limits of any power hitter definition. In all, there were only 7 player seasons of 20+ home runs in the thirteen Long Years.




In the thirteen Short Years, there were 19 player seasons with 20+ home runs. The power on those clubs was Wynn (again - 2 seasons), Rader (again - 2 seasons), Cedeno (3 seasons), Lee May (3 seasons) and Glenn Davis (6 seasons). They also got 20+ seasons from Cliff Johnson, Kevin Bass and Franklin Stubbs.




It appears, then, that the Astros were at least trying to incorporate a little more power during the Short Years.




-----

*All park data is from Retrosheet.

Guillen and RFK

So Vinny Castilla and Jose Guillen think they are getting gipped in RFK Stadium and it will affect contract negotiations. See http://sports-law.blogspot.com/2005/07/rfk-stadiums-false-dimensions-are.html. Guillen said "I should have 29 homers. I should be up there with Derrek Lee, Andruw Jones. You see the balls I'm hitting in those ballparks away from here. I just want to find out the truth on my own."




Using the methods described in Bill James' book "Win Shares", the Home Run Park Adjustment for Washington Nationals players would be .785, meaning you'd divide a player's home runs by that amount. Guillen has 19 home runs, so with the adjustment, he'd have about 24 if he played in neutral parks all the time (including on the road). However, only 1 of those extra 5 home runs would be at RFK (RFK itself actually suppresses home runs by half)...the others would come in road parks.




For comparison, the Braves have a Home Run Park Adjustment of .963, so Andruw Jones would have 33 home runs if he played exclusively in neutral parks. The Cubs have a Home Run Park Adjustment of 1.025, so Derrek Lee would have 31 home runs if he played exclusively in neutral parks. Guillen is not even close.




It is easy to see why Guillen doesn't like RFK...he only has 1 home run there, which is tied for 8th best on his team, even though he is the team's best hitter. But why, when the Nationals have gone into a nosedive and have lost the lead in the NL East, is he complaining about whether his personal home run totals are high enough at home? Maybe he should be worrying about the Nationals making the playoffs. In short, he is a head case.




Last year Guillen was the best hitter for the Angels. They were in a pennant race as the season wound down. They suspended Guillen for undisclosed reasons, just when they needed him most. The Angels made the playoffs, but did not put Guillen on the playoff roster and effectively released him. How much of a clubhouse problem must Guillen have been? He was their best hitter, and rather than fine him or something, the Angels basically said "You are such a cancer to this team, we are kicking you off the team, right in the thick of the playoff chase."