Stat Chug: Searching for a Drinking to Batting Ability Link

Stat Chug: A specially brewed stats based draft.

Searching for a Drinking to Batting Performance Link: The quest to find the link between inebriation and batting performance.

In a Sport such as Tapey Beercone, where drinking is prevalent (in fact encouraged by the rules), it seems obvious that all of this drinking would have some effect on a player’s performance and the outcomes that player produces. A good part of what makes Chugging Percentage (CHG) such a prestigious statistic is that the components of the stat ( Batting production measured by Slugging Percentage (SLG) and Drinking Frequency measured by Beers per Inning (BPI)) are presumably not independent. The idea being that for one to remain a productive batter whilst also quaffing beer at an exceptional rate takes unique ability. Plainly put, one would expect that the relationship between batting outcomes and drinking frequency would be inversely correlated, that to drink more would on average mean to give something away as a batter. With these assumptions in mind, this author aimed to seek out this relationship with the goal being to measure it, and then with these measurements in hand create a new metric to judging the exceptional batters/drinkers who standout from the pack. What follows is a walk-through of this quest.

Let’s start by talking performance. Specifically, allow me to explain the methods I would use to measure offensive performance. I looked at two metrics: Weighted On-Base Average (wOBA) and Batter Leverage Adjusted Win Probability Added (bWPA/LI). The quick definitions are that wOBA is based on the actually outcomes of each batting play where as bWPA/LI is only concerned with what the batter can control producing a stat that is fielder and park neutral. For a more thorough explanation of each stat, click the tabs below, or check out the statistics page:

wOBA
wOBA weights the actual outcomes of each play by that type of play’s average run scoring potential, and scale the result to the On-Base Percentage scale. For example an out is given a weight of 0, a single is given a weight of .969, while a home run is weighted at 1.475.

bWPA/LI
bWPA/LI is a component of WPA/LI which measures the Win Probabilty Added by the batter by the action they can control at the plate. This could be walking, striking out, or hitting a particular quality of batted ball into play. The batter is credited with the estimated average win expectancy change that will result from that event with that change scaled by the situation’s leverage.

With statistical explanations out of the why, the quest begins.

Step 1: Compare the result of each play against a batter’s BPI.
Result: A slightly positive relationship!, but essentially no correlation, with R2 values of less than .01 for both wOBA and bWPA/LI. How could that be?

There appears to be an obvious reason for this. We haven’t yet controlled for the batter, and many of the Sport’s best batters also happen to be frequent drinkers. To control for this we need to adjust and compare each player against their overall average for each batting stat.

Step 2: Adjust each event for the batter, and just look at how the play compares to the batter’s average play when compared to that batter’s BPI.
Result: Still slightly positive outcomes, but now basically no relationship when it comes to pure batter production. As the graphs show this is just random noise, with no correlations at all, R2 below .001!

Now things are starting to get complicated. After controlling for batter one would expect to see some kind of relationship. To find nothing is a little perplexing. There’s one more obvious adjustment to make, that being to also adjust each play by comparing each player against their average BPI.

Step 3: Also adjust each event for the batter’s average drinking habit.

If you’re still following along that means that each play becomes a data point where we measure how the batting stat compares against the batter’s average outcome as well as comparing the batter’s BPI to their overall average BPI. For example if Boyd hits a single while running a 2 BPI this would show up on the graph as an adjusted BPI of 0.11 (2.00 – 1.89) and an adjusted wOBA of .352 (.969 – .617)
Result: Say what? We are back to positive again based on wOBA. But we are still basically looking at complete random noise.

Time to take a step back. Are we measuring the right statistics? Particularly when it comes to the drinking end of things? Is having a BPI of 3 in the first inning (3 beers drank after one inning played) more impressive than having a BPI of 2.5 in the sixth? (15 beers drank after six innings played) No way! Plenty of players could get off to a fast start and feel nothing three beers in, but few players take down 15 beers and feel fine even after three hours. Clearly then we need a new statistic:

Enter Inebriation or INB

Inebriation is built on the concepts of alcohol content and alcohol metabolism. As you drink you increase your body’s alcohol content, but you are constantly metabolizing that alcohol. A rule of thumb for metabolizing alcohol is that one burns off a single beer’s worth of alcohol per hour which works out to about one-half beer per inning of play. Thus INB is a simplified measure of total beers consumed above your body’s ability to remove those beers from your system.

Armed with this new statistic we move forward.

Step 4: We replace BPI with INB
Still keeping in mind that we are measuring a batter’s INB on each play against that player’s average INB over all plays.
Result: Still nowhere, just more random noise with a small negative relationship when it comes to bWPA/LI.

Now a Hail Mary.
Step 5: Instead of looking at every play individually we’ll bucket them by player based on adjusted INB.

So for each player (with over 25 PA’s and with PA’s at both a low and high INB) we group the results of all plays within a given INB range and average them. I set up buckets of 0.5 INB beers above and below each player’s average level.
So far example: Brian has an overall average wOBA of .653 and an average INB of 1.25. When we look at his plays where he was between 0.5 and 1 beers higher than this average INB (That would be 1.75 to 2.25 beers) he averaged a wOBA of 0.675 or 0.022 higher than his average. And during that bucket of play his average INB was 1.96 or 0.71 above his average. This ends up as a point on the graph as 0.71 INB and 0.022 wOBA.

Result: So where does this leave us? Behold!

That’s right. We’ve found a relationship. Not much mind you, still only an R2 of 0.06 but it’s there. And it’s not what conventions would have us assume. The shocking reality is that as one drinks more that player’s offensive ability does not decrease. Instead it rises!

You actually get better as you drink!

As for why the effect is more prominent in wOBA versus bWPA/LI. This can likely be attributed not to the batter but to the defense. If they also drink and are impaired by it, this could lead to better outcomes for the batter on the same types of batted balls. But who’s to say fielders and pitchers aren’t also enhanced by the beers they drink?

The biggest take away should be: Don’t let fear of a performance decline stop you from throwing back another cold one. Keep on chugging and don’t stop!

Cheers!

Author: The Coach

#8 The Coach, founding Regent of Tapey Beercone.

Leave a Reply