November 30, 2006
A Second Look at First Contact
"Going over hitters is something you do before each series, and before we went against the mighty Angels, Sal Maglie had a great hint for one of their weak hitters Vic Davalillo. 'Knock him down, then put the next three pitches knee-high on the outside corner, boom, boom, boom, and you've got him.'
"Everybody laughed. If you could throw three pitches, boom, boom, boom, knee-high on the outside corner, you wouldn't have to knock anybody down. It's rather like telling somebody if he'd just slam home those ninety-foot puts he'd win the tournament easily." --Jim Bouton, recounting some sage pre-game advice in Ball Four
Several weeks ago we discussed the difference between "first contact" situations and subsequent plate appearances. We were looking at the set of players (437 positions players and 562 pitchers) who debuted during the 2000 through 2004 seasons, comparing how they fared against one another in their very first plate appearance versus all subsequent plate appearances.
As some readers noted, and as mentioned in the previous article, this doesn't take into consideration that players might have faced each other in the minor leagues and no, we didn't include rookie hitters versus veteran pitchers and vice versa. While expanding the data set to include those categories would also show "first contact" situations, we didn't include them because using only players who debuted is in some sense a more "pure" sample, and on a more practical level, it allows us to draw a line in the sand, thereby making the processing of the data simpler.
What we found was that in the "first contact" situations the outcome of the plate appearances equated to a batting line of .258/.331/.405, while in subsequent plate appearances it was .270/.333/.426 (deemed a statistically significant difference in terms of batting average at the 1% level). When translated to run creation, that's an increase in production of 7.8%. On the basis of this informatio, I drew the conclusion that pitchers do have the advantage the first time hitter and pitcher meet, just as we might all be prone to intuit, since pitchers initiate the play.
Nevertheless, science marches on and intuition can't always be trusted, so this week we'll look at a competing effect that substantially impacts our first look at first contact.
One of the cardinal sins of a researcher is to ignore previous research that may bear on the question at hand. In this case I had read, but had forgotten until reminded by a reader, that David Smith of Retrosheet fame published a study in SABR's publication The Baseball Research Journal #34 in 2005 (there is also an older version of this study using less data published in 1996 available online) that has bearing on this question.
In that study titled "Do Batters Lean During a Game?", Smith examined all plate appearances from 1960-2005 for batters against starting pitchers and broke down the results by the plate appearance number (omitting the fifth and subsequent plate appearances, since there are so few) as follows:
PA AVG OBP SLG 1 1,530,593 .259/.328/.393 2 1,456,880 .269/.331/.416 3 1,151,387 .274/.336/.427 4 394,251 .275/.335/.418
As you can see, hitters clearly perform more poorly in their first plate appearance in a game, and as the pitcher tires and/or the hitter adjusts his approach (Smith opts for the latter), the results trend upward until the fourth plate appearance, when both on base percentage and slugging percentage dip a bit. This latter dip is almost certainly caused by the fact that if a pitcher has the opportunity to face a hitter four times in a single game, the pitcher is likely pitching well in that particular game, and would be that much more likely to have done well in the previous three plate appearances against the batter.
If we condense Smith's table into two rows by weighting the final three ordinal plate appearances by the number of plate appearances we get:
As you can see, the outcomes for plate appearances after the first differ from the first by approximately the same amount as in our first contact results. This should raise a flag.
Where this intersects with the discussion of two weeks ago is in the fact that by definition our "first contact" plate appearances occur as the first plate appearance in a game between a hitter and pitcher. With that in mind, not all of those first contact plate appearances occur early in a game, as is the case in Smith's study, and the subsequent plate appearances also include first plate appearances in subsequent games. Even with those considerations in mind, if there is an effect causing hitters to hit more poorly in their first at-bat in a game against a particular pitcher (as seems the case), our "first contact" results would be conflated with that effect.
In order to get a feel for how much that effect has influenced our results, we'll start by restricting our view to 2000-2005 data and running the numbers in a similar fashion to the Smith study. Within that data set we can then calculate a baseline by looking only at the following criteria:
This gives us a baseline for the effect of first plate appearances versus all others in the same game within the date range of our first contact data set.
2000-2005 All Hitters PA AVG OBP SLG OPS K/AB BB/PA HBP/PA First PA in a G 257858 .259/.326/.413 739 .195 .080 .009 Subsequent PAs Same G 449233 .276/.339/.449 788 .164 .076 .009
From a productivity standpoint, once again using Runs Created, moving from the first plate appearances to subsequent plate appearances in the game increases run production by 15.8% (4.81 to 5.57 RC/27). You'll also notice that as mentioned in the previous column, strikeouts go up in first plate appearances, as do walks. Because of the large sample sizes involved, this result is highly statistically significant, meaning that for all practical purposes, there's no chance that the difference we're seeing could be a matter of chance.
For comparison, we'll look only at our subset of first contact hitters and pitchers, and see how they fared in the games in which their first contact took place, and where they faced each other more than once.
First Contact Situations PA AVG OBP SLG OPS K/AB BB/PA HBP/PA First PA in 1st G 12911 .268/.337/.417 753 .201 .081 .012 Subsequent PAs in 1st G 22276 .274/.335/.434 769 .170 .072 0.010
Aha! This is a little more interesting. Here we can see that in games in which a hitter and pitcher first meet and also rack up additional plate appearances, the productivity gain in those subsequent PAs is only 4.6% (5.05 to 5.28 RC/27). This points to the conclusion that, relative to the baseline above, it is the hitters, and not the pitchers, who have the advantage in first contact situations. This is suggested by the fact that the "natural" first plate appearance advantage of 15.8% is shrunk to 4.6% when pitchers and hitters face each other for the first time. If pitchers aren't realizing their full natural advantage, then hitters must be performing better than expected.
As noted in the previous article, the odds of a hitter being hit by a pitch in their first at-bat against a pitcher is slightly increased, as is their walk rate, leading to a higher OBP. This difference in batting average does not have quite the strength from a statistical perspective, although the difference in slugging percentage is significant at the 1% level.
There's an additional interesting fact hidden here. When we look at this set of hitters and pitchers in their first games, the runs created per 27 outs was 5.20 (.272/.428/.335), while in subsequent games with the same set of hitters and pitchers, the output went down to 5.12 per 27 outs with a batting line of .271/.425/.333.
At first blush one might think this difference is caused by factors such as the negative effects of pinch-hitting in subsequent games, since it's known that there exists an inherent "pinch hitting penalty" that suppresses offensive production. But it turns out there were too few pinch-hitting plate appearances--less than 1% of the total--to have any effect in those subsequent games. However, it may also be the case that throughout the first game or games, hitters enjoy a general, albeit slight, advantage over pitchers, but that the more a pitcher faces a particular hitter, the more that advantage fades away. In other words, although hitters appear to learn during a game, might it be the case that pitchers are the ones doing the learning over the long haul?
To determine whether this latter idea holds any water, we can take a look at the performance of a subset of those first contact hitters and pitchers over time. The following graph represents the RC/27 for each of the first fifteen plate appearances for 862 pairs of hitters and pitchers who had at least fifteen head to head encounters from 2000 through 2005.
As you can see, this group of hitters performed better in their first contact PA than in the next several, but then rebounded before falling off again around plate appearance number thirteen. In lieu of a distinctive pattern, the best one can say is that this is what you might expect if both parties are constantly making adjustments. Breaking these plate appearances into three groups of five, we also see no difference in productivity in plate appearances 1-5 and 6-10 and only a slight dropoff (-2%) in plate appearances 11-15, which is not statistically significant. So in general, this provides no evidence that a trend exists in either direction and because the first game advantage discussed above is so slight, the question remains open and the most we can say is that there is no evidence that hitters are at a disadvantage in their first game against a particular pitcher.
The Developing Book
As mentioned in the initial column, I would have expected pitchers to have the advantage the first time they encounter a particular hitter at the big league level. It turns out that indeed they do, only that the advantage is smaller than the advantage normally enjoyed by pitchers the first time they see a particular hitter in a game (as in the Smith study). In addition, that first plate appearance advantage is apparently offset a bit by a slight, though not statistically significant, boost in performance in that first game. Although these conclusions turn out to be counterintuitive, that's all the more reason to ensure that those pregame meetings bring all the available intelligence to bear, and are (hopefully) more productive than the one famously disparaged by Jim Bouton.