Grow Your Tool Box: Split Stats

Baseball announcers incorporate statistics into their broadcasts to provide the listener with additional insight.  Listeners can use this information to form their own predictions of what to expect as a game unfolds.  There is no guarantee a .330 hitter will collect two hits, but the listener can deduce the player is performing at a high level just by hearing his batting average.

Split stats, commonly referred to as “splits,” divide a team or player’s total numbers into two or more related categories.  Splits help identify performance trends that the overall totals can’t reveal on their own..

Imagine the following situation: The tying run is on second with two outs late in the game.  First base is open, and the current batter is hitting .330.  The on-deck batter is hitting .250, but the defensive team elects to pitch to the current batter.

At first, fans of the defensive team might be wondering why the current batter wasn’t intentionally walked at such a critical juncture.  However, the broadcaster reveals that while the current batter is hitting .330 overall, he is hitting just .190 against left-handed pitching, and the defensive team happens to have a lefty on the mound.

I’ll leave the hypothetical outcome to this hypothetical scenario up to your imagination as a means of emphasizing that the process and methodology behind this type of decision is always more important than the actual result (expect a future post on this subject).  What’s important is that the batter’s splits (in this case, his left/right split) influenced the manager’s strategy.

Thinking along with the manager and all of the game’s possibilities is one of the most enjoyable aspects of watching or listening to a game.  If certain splits impact a manager’s decision-making process, then providing the listener with the same information will dramatically improve their experience.

Before we get into which split categories are most relevant to a baseball broadcast, there are three very important caveats to keep in mind.  First, like any statistic, split stats can illustrate a correlation, but they are in no way absolute predictors of future performance.  Left-handed batters who have a hard time against left-handed pitching can still hit a home run off of Clayton Kershaw.

Second, even if a wide differential exists within a particular split, it doesn’t mean the specific category is the sole reason for the difference.  A batter’s numbers at home could be very different from his numbers on the road, but factors above and beyond the sheer ballpark dimensions could be contributing to the split.

Third, there are many examples of players or teams exhibiting “reverse splits,” or splits that show the opposite of what you would normally expect.  A batter could have more home runs on the road than at home, even though his team plays its home games in a bandbox.  A right-handed pitcher could have better numbers against left-handed batters, even though the majority of pitchers perform better against arm side batters.  Reverse splits can be a coincidence or the result of a more plausible explanation.  In either case, you need to be aware of reverse splits.

So which splits are most relevant to the broadcast?  The following are Growcasting’s “Big Four” split categories, each of which can improve the listener’s experience if used with the proper context.

Home/Road Split
The home/road split divides a team or player’s total numbers into performance at the home ballpark and performance everywhere else.  The home/road split is very useful in measuring the effects of a home ballpark.  Since teams play an even number of home and road games, the sample size on both sides of the split will be roughly equal at any point during the season.  This creates a fair comparison, especially late in the year when the total sample size nears its peak.

Additionally, a team will generally (not precisely, obviously) play the same number of road games at each of the visiting ballparks around the league.  This means the road side of the split is collectively neutral, further validating any differences that might exist.  In other words, if a player’s home park was hitter-friendly, but the majority of his road games took place in other hitter-friendly ballparks, it would be difficult to attribute any difference in his home/road split to stadium dimensions.

The physical differences and nuances in each ballpark make the home/road split more important in baseball than any other sport.  Every football, hockey, and basketball surface is cut exactly the same.  And of the three other major sports, only football mixes indoor and outdoor stadiums on a regular basis.  Baseball is predominantly played outdoors (especially if you include the minor leagues), so the combination of local climate and park dimensions play a significant role in a team or player’s home/road split.

In our caveats, we mentioned the split category itself is not necessarily the sole reason for a wide differential or an individual game performance.  This warning is most applicable to the home/road split where performance is affected by environmental factors just as much as it’s affected by the outfield wall’s distance from home plate or the stadium’s height above sea level.

There is the home crowd factor.  A pitcher could take the mound in a spacious ballpark, but blocking out the hostile noise of thousands of fans requires a very high level of discipline and concentration.  When a team is playing at home, its players are able to sleep in their own beds and work on their normal schedules.  A good night’s sleep can be more valuable to a player than whatever structural benefits a stadium might offer.  These are just two of the many environmental factors to consider.

Nevertheless, the same ballparks tend to produce similar results from year-to-year.  This reliability, combined with the 50/50 nature of the sample size make the home/road split an essential part of a broadcaster’s tool box.

Left/Right Split
The left/right split is by far the most relevant because of how often it influences in-game strategy.  A manager might alter his starting rotation to ensure a specific pitcher makes his next start at home, but that decision is made ahead of time and locked in place once the game begins.  As we noted in our hypothetical exercise at the beginning of the post, the left/right split constantly plays a role for both managers up until the final out is recorded.

The left/right split divides a team or player’s total numbers into performance against left-handed pitching (or hitting) and right-handed pitching (or hitting).  The wider the differential, the more likely a manager will try and exploit the weaker side of a player’s split in a key match-up.

For switch hitters, the left/right split is the perfect way to identify which side of the plate the batter has more success from.

Pitchers generally have more success against arm side batters, but if a pitcher has a reverse left/right split, a quality change-up might be the answer.  Change-ups are usually only thrown to glove side batters due to the pitch moving down and away (the same pitch would move down and in to arm side batters, which is why you generally don’t see many right-on-right or left-on-left change-ups).  So if a pitcher’s best weapon is his change-up, don’t be surprised if he has a reverse left/right split.

If a pitcher has a dramatic left/right split, it might also lead the opposing manager to “stack” his lineup against the weaker side.

Another thing to keep in mind is whether or not a team has any left-handed pitchers in its bullpen at all.  Every manager wants to have at least one left-handed reliever, but circumstances could lead to an all right-handed pen for a period of time.  If this is the case, the left/right split becomes much less relevant.  You can bring up the fact that the opposing team lacks a left-handed reliever early in the broadcast, but specifically mentioning a batter’s left/right split once the starting pitcher has exited isn’t necessary.  There is no possibility of a left-on-left match-up, so fill your downtime with a better quality of information.

Situational Splits
There are multiple situational splits that divide a team or player’s total numbers into categories based on the number of outs, the number of baserunners, or both.  Some common examples of situational splits include:

– Leading off an inning (or facing the leadoff batter for pitchers)
– Runners on base
– Runners in scoring position
– Bases loaded
– Runners in scoring position with two outs

Because situational splits are more specialized than home/road or left/right, the sample sizes are naturally going to be smaller.  This needs to be considered right away, because it might take until the halfway point of a season (or even later than that) for a player to build up enough at bats or innings where you can interpret the meaning of a situational split with a reasonable level of confidence.  For this reason, situational splits are best used to illustrate what a player has done up to that point in the year as opposed to using them to try and glean any predictive value.  For example, if a batter is hitting .240 for the season but .350 with runners in scoring position, you can say something along the lines of “he has really made his hits count so far this year.”  It would be premature, however, to declare “he is one of the most clutch hitters in the league” if the player only has 20 at bats with runners in scoring position.

Other situational splits that might be noteworthy enough to mention, but offer very little predictive value include day/night, record or performance by day of the week, performance by inning (although, this can offer some value for identifying when a starting pitcher tends to fatigue), or performance by position in the batting order.

Trend Splits
The last of our “Big Four” are trend splits, which can provide just as much insight to the listener as the left/right split.  When a .333 hitter steps to the plate, it’s highly unlikely he has gone 1-for-3 in every single game.  Virtually every player experiences hot and cold streaks, and no matter where a player’s final numbers end up, a player is probably experiencing one of those streaks in the present moment.  Trend splits allow you to provide the listener with this valuable information.

If you’ve ever watched or listened to a baseball game, you have undoubtedly heard an announcer reference what a player has done in his “last five” or “last 10” games.  The 5/10 split is the most commonly used trend split, and is made available in nearly every pre-game statistical report at the professional or collegiate level.  While using the 5/10 is common practice for announcers at all levels of baseball, you should start ignoring it right now.  The 5/10 split can be distorted too easily by one or two monster games, and leaning on it too much does a disservice to the listener.

The optimal method for identifying recent trends is to carefully examine the game-by-game logs of each player.  This allows you to locate exactly where a hot or cold streak begins and ends.  You’ll be able to provide the listener with a much more accurate snapshot of a player’s recent trend, and it demonstrates on-air proof that you’ve made the extra effort to try and improve their experience.  Consider the following two examples:

Batter #1
June 1st: 2-for-5
June 2nd: 1-for-4
June 3rd: 2-for-4
June 4th: 1-for-4
June 5th: 3-for-4
June 6th: 0-for-3
June 7th: 2-for-5
June 8th: 2-for-4
June 9th: 1-for-4
June 10th: 1-for-3
Last 10 Games: 15-for-40 (.375)

Batter #2
June 1st: 5-for-5
June 2nd: 4-for-5
June 3rd: 2-for-4
June 4th: 3-for-4
June 5th: 0-for-2
June 6th: 0-for-4
June 7th: 0-for-4
June 8th: 1-for-5
June 9th: 0-for-3
June 10th: 0-for-4
Last 10 Games: 15-for-40 (.375)

You would be accurate in saying that both players are hitting .375 in their last 10 games, and this would be reflected in the 5/10 portion of the daily stat package.  However, you would not be accurate in saying that both players are “hot.”  You can clearly see the two players arrived at their .375 clip in different ways.  Batter #1 has been consistent throughout the 10-game stretch, hitting safely in nine of the 10 contests.  Meanwhile, Batter #2’s .375 average in his “last 10” is inflated by the first four games of the stretch, and actually appears to be in the beginnings of a cold streak (1-for-22, .045 in his last six games).

The 5/10 split will not reveal this, whereas checking the individual game-by-game logs will.  By reading the game-by-game logs, you might even discover that the first four games of Batter #2’s “last 10” marked the tail end of an historically long hitting streak.  Game logs are easily accessible for all professional leagues, so there is no reason not to consult them.  They just aren’t included in the daily stat packages for the purpose of saving paper.

The 5/10 split is fine if you have limited preparation time, but if you’re relying on it too much, savvier listeners will quickly recognize that you’re not painting the most accurate picture.  Just as a .333 hitter doesn’t go 1-for-3 every night, hot and cold streaks are not exclusive to convenient five- and 10-game windows.  If you invest the extra few minutes to read the game-by-game logs, you’ll identify the most powerful trends and deliver the highest quality of information to the listener.

Finally, make sure to avoid “double splits.”  A double split divides a split category into another category.  An example would be home/road performance in day games versus home/road performance in night games.  Again, you always have to be conscious of sample size.  Even a normal split cuts the total numbers in half (at best), so the initial sample size needs to be large enough for the split to have any reliability.  A double split reduces the sample size even further, and in general, double splits offer close to no predictive value.

In virtually all cases, double splits cross the line of oversaturating the broadcast with numbers.  The further you take double splits, the closer you are to saying something like “he’s hitting .325 with runners in scoring position during overcast day games in which the team is wearing the red hats.”  Even if a significant sample size existed for this scenario, the statistic would mean absolutely nothing.

The “Big Four” are all you need to deliver a quality broadcast to the listener, and are readily available at almost every level of the game.  Properly researching, interpreting, and incorporating them will greatly enhance the listener’s experience.

2 comments

  1. Pingback: Scorekeeping Part 1 – Your Scorebook | GROWCASTING
  2. Pingback: Keeping Score Part 3 – Notes Boxes | GROWCASTING

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s