November 20, 2009
Early Career Splits
There's one advantage about being a fan of a franchise that never makes the postseason: I have much more off-season time to obsess about the following year. For Royals fans, one of the main questions over the last few years has been what do they have in Alex Gordon. Probably unfairly, people were labeling him as the next George Brett immediately after being drafted. On the other hand, a perennial All-Star is not too much to ask for a player who was picked second overall and considered the best college hitter that year.
When you watch him in numerous games, the one thing that pops out is how uncomfortable Gordon looks against southpaws, and the stats back this up (.217/.288/.365). Against righties, he has a slash line of .264/.349/.436 which is not as good as one would hope, but still serviceable in a Joe Randa kind of way. On the flip side, one of the Royals' newest acquisitions, Josh Fields, has an extreme split the other way: .285/.367/.580 against lefties, while just .206/.280/.348 against righties. I find myself hoping that these young hitters can figure it out against same-side pitchers, since they have demonstrated some hitting prowess against the opposite side. But is this an accurate assessment of the situation? This thought process led me to a bigger question: do young hitters who have extreme righty/lefty splits improve more in their peak years because they begin to figure out same-side pitchers better? Or are the young hitters who show minimal splits better hitters who improve equally against both.
To begin the analysis, I looked at the set of hitters in the past twenty years who met the following criteria:
I put in these minimal requirements so that we had enough PAs that the splits were more likely to be reality and not a problem with small sample size. Given that usually the number of plate appearances against a righty versus lefty is roughly three to one, this would usually give 300 PAs against lefties for most hitters in their early career. There have been 197 hitters that matched these criteria, and their overall stats are as follows:
The OPS split is defined as the difference in a player's OPS against the opposite-hand pitcher versus the same-side pitcher. So for this batch of players, on average, we see a 31-point improvement in their OPS from their early years to the peak years, but there isn't much change in their split numbers.
Defining Early Career Splits
For the next step, I categorized each player as having a low, medium, or high split during their early years. I selected breakpoints of .040 and .100 to divide the players into three roughly even-sized groups: Low (Split less than .040), Mid (Split between .040 and .100), and High (>.100 split).
Early Years Statistics # of Avg. PA Overall Same-side Opposite-side Split Hitters per Year OPS OPS OPS Low 64 509 770 762 783 Mid 72 507 789 761 826 High 61 492 822 719 890
There are a couple of interesting points out of just this initial categorization:
So how do these types of players improve as their career progresses?
Peak Year (Change) Statistics Avg. PA Overall Same-side Opposite-side Split per Year OPS OPS OPS Low 551 (+42) 794 (+24) 774 (+12) 835 (+53) Mid 562 (+55) 823 (+34) 782 (+21) 868 (+42) High 570 (+78) 859 (+37) 788 (+69) 895 (+5)
The first thing that should be noticed (and was probably to be expected) is that in the aggregate these splits start getting closer to the mean; i.e., for players with high splits, they don't seem to hit much better against opposite-hand pitchers, but they usually have their same-side OPS improve. With that said, this tempering isn't complete regression to the mean. Players who have high splits in their early careers continue to have higher splits in their peak years as well.
While players who had a high split improved a little better overall than the low-split group, the overall OPS improvements were not quite statistically significant. It should be pointed out that for players who have had high splits, on average, their same-side OPS actually improves such that it is now better than the players who initially had a low split. With this improvement, as to be expected, they are rewarded with greater playing time.
Getting back to our case of Alex Gordon and Josh Fields, this may bode well, in that players who have higher splits seem to experience more improvement than players who have lower splits. Before Royals fans get too excited, one should consider that there may be self-selection going on based on this analysis. We have selected players who were able to accumulate 1,600 plate appearances during their peak years. The simple fact may be that players who show extreme splits who are not able to improve against the opposite-hand pitchers do not stick around in the majors long enough to achieve 1,600 plate appearances between their age-28 and age-31 seasons.
The Improvement Paradox
So initially, one could draw the conclusion that players should do whatever they can to even out their splits. Instead of categorizing players based on the magnitude of the split in their early years, I next categorized the players based on how their split changed from their early years to their peak years. If the OPS split decreased by 30 points or more, I put them in a group labeled "Reduced," while if their split increased by 30 points, they were placed in a group labeled "Upped." All other players were put in the group "Same" (i.e., their splits stayed roughly the same).
Split # of Overall OPS Split Same-Side Opp-Side Change Hitters Improvement Change Change Change Reduced 66 +24 -83 +65 -19 Same 63 +25 0 +26 +26 Upped 68 +44 +85 +9 +94
The overall OPS improvement for those players who increased their split was higher compared to those that either reduced their split or kept it the same, and this difference now becomes statistically significant.
Players who reduced their split seemed to significantly improve their OPS against same-side pitchers. However, this was at the expense of their strength, their OPS against opposite-hand pitchers. On the other hand, those players that significantly increased their split saw a greater improvement in their overall OPS, and almost all of this came from being even tougher against opposite-side pitchers, yet their same-side OPS barely changed.
So the paradox is that this seems to go exactly against the previous finding that players with high splits seemed to improve as a whole because they were evening out their splits. When I separated out the players who had high early-career splits into the three buckets above, the players who had reduced or had the same split had a nice, steady improvement, but the players who had increased their splits ended up being off-the-charts good.
Some of this may just be coming from the still relatively small sampling of players. There are only five players (Jim Edmonds, Jason Giambi, David Ortiz, Benito Santiago, and Jim Thome) who had high splits in their early career and then increased their splits in their peak years, and three of them (Giambi, Ortiz, Thome) had a 1000 OPS in their peak years.
It was consistent in each Split Change group ("Reduced," "Same," and "Upped") that the players who saw the greatest improvement in each of these groups were those that were originally in the "High" split group. The table below shows the OPS improvement and the number of players in each group:
OPS Improvement (# of Players) Early-Career Split Low Mid High Reduced +3 (2) +22 (23) +26 (41) Same +18 (29) +27 (19) +34 (15) Upped +30 (33) +47 (30) +136 (5)
From this table, two things could be inferred: