Some very valid questions from readers prompted me to take a second glance at my second version of GRIT. As I mentioned before, GRIT is a bit of a work in progress that I pick at it whenever I get some time. With the holidays granting me some freedom from responsibility, I decided to do a bit of a revision that consists of two major changes.

• The Elimination of Eligibility Requirements
• The Discontinuation of Opportunity Normalization

Now, the following article discusses these changes to earlier versions of GRIT and why they needed to happen. To reduce confusion, I’m going to differentiate between these earlier versions of GRIT, by appending version numbers (e.g. GRIT1, GRIT2, and GRIT3).

1. The Elimination of Eligibility Requirements

Under the GRIT1 formulation, I used 81 games and 100 ABs as the cut-off points for an eligible season (technically, player-season-stintÑmore on that in a bit). These requirements were used for two reasons:

A. To prevent Excel from shitting bricks.

Excel really, really, really hates running rather complex formulas across tens of thousands of rows of data. I’ve since moved everything in to Access, so keeping these thresholds in place is unnecessary. Nonetheless, I thoughtlessly carried over the eligibility requirements to GRIT2 last week.

B. To prevent pitchers from biasing the standard deviations and averages of the components.

After a little experimentation, I discovered this second concern really turned out to be unfounded. I’ll spare you an in-depth treatment of why this is so. Just take my word on it.

Right? Right.

In light of this, GRIT3 scores are now determined for every player-season-stint so long as they featured at least one (1) AB.

What’s a player-season-stint, you ask?

A player-season-stint is the period of time any given player spends during any given season with any given team (i.e. Felipe Lopez’s stint with the Nationals in 2006). This is different from a player-season, or the sum of every player-season-stint for a given player during a given season (i.e. Felipe Lopez’s 2006 season).

2. The Discontinuation of Opportunity Normalization

In the earlier formulations, I normalized plate appearances. After a bit of consideration, I realized that there’s no good reason to do this. However, the new addition of a large number of very brief player-season-stints will negatively skew the distribution of plate appearances by season stint. This, alone, is probably a good enough reason to not normalize the Opportunity values. As a result, Opportunity is now calculated by dividing Plate Appearances by 100.

The effect of these two changes makes GRIT3, bar none, the best metric for all of your grit quantification needs. They also enable some serious statistical analysis. Specifically, it allows us to answer an extremely important question.

Where’s my Lil’ Davey Eckstein? Why can’t I find him?

Many readers looked through the Top 50 Grittiest Players of 2008 and were left pondering this glaring omission. Now, a lazy gritistician might take the easy way and tell you to simply check all the customary places where David Eckstein gets misplaced (e.g. between couch cushions, behind your headboard, the back of your junk drawer, etc.) but not me. As Eckstein was the impetus for GRIT1, any grit quantification metric that doesn’t include a full accounting of his intangible contributions is certainly useless. No, no, no. Now is not the time for half-assery. I got to the bottom of the issue and came up with a solution.

With 324 at bats in 94 games during 2008, Eckstein’s player-season ostensibly met the eligibility requirements. However, Eckstein divided his time between the Blue Jays and the Diamondbacks. This gave him two separate player-season-stints for 2008, both of which fell short of meeting the 81G/100AB requirement.

Eckstein’s omission in the Top 50 of 2008 table was thus due to GRIT2’s use of player-season-stints instead of player-seasons.

So, Eckstein never even received a 2008 GRIT2 score. By removing the eligibility requirements, Eckstein’s 2008 player-season now has a GRIT2 score of 7.427. This is rather balmy compared to his career averages, but still puts him at 27th for 2008. Had he received his pre-2008 average of 609 plate appearances, he would have put up a 12.029, a solid 9th place finish.

About a year ago, I posted an article to Flotsam Media that introduced GRIT to the world of sabermetrics. GRIT had a nice night and said it was going to call the world of sabermetrics back in a few days, but it never did because it’s a good for nothing loser with a one-track mind. Flotsam is now defunct, so I’ve posted the article here for posterity’s sake. I also spent a few days in my laptop science lair working on an update. That’s what this post is about.

Just as a refresher, here’s the idea behind GRIT laid out in the original article.

Gritty players are those who are determined to win or succeed at baseball, but due to a lack of natural skill or talent, are forced to do so through the least efficient means possible. This inefficient play results in excessive dirt on their uniform.

If this hypothesis is true and dirt, determination and talent can be reasonably quantified, then it should follow that each of these three components can be plugged in to a formula to determine a given player’s grittiness (GRIT).

GRIT’s a bit of a work in progress, but I hope to continue to refine it and serve up lots of pseudo-statistics about your favorite baseball players–unless your favorite player happens to be Cesar Izturis. In that case, you’re welcome to leave.

CHANGE LOG

There are three major changes since last year’s article.

1. The Acronym

Originally, GRIT was comprised of some simple criteria that gritty players seemed to have in common. Thus, it made sense way back then that the acronym stood for ÒGeneral Requirements of Intangible TalentÓ, but not anymore. ÒGeneral Rating of Intangible TalentÓ is probably the most logical name, so I’m going with that. Sweet!

2. The Dirt Formula

I capitulated to public opinion and overhauled the Dirt component pretty substantially. Dirt changed from:

DIRT = HBP – IBB + ((CS +1)*CS))/(SB + CS + 1))

to

DIRT = HBP – IBB – (HR/2) + ((SB*-.3)+(CS*.6))

What’s the rationale for the addition of (HR/2)? You can pick one or both of two explanations.

• Home runs are a way of avoiding dirt since you get to trot around the bases like a pony. Ponies aren’t gritty. Burros? Gritty. Ponies? Not gritty.
• Because Don Baylor was ranked really high in last year’s rankings. I’m not dumb enough to call Don Baylor a pony. That said, he wasn’t gritty.

And the stolen base stuff? Well, the new statistic is a fairly rough estimate for the number of runs that a player costs his team with his (lack of) base-stealing ability. This link explains the logic behind valuating stolen bases. I didn’t feel like manually pulling the data for each individual year from BP’s website, so I used some unscientific estimates.

3. Normalization Methodology

Another change is the way the component values are normalized. I’ve switched to using averages and standard deviations grouped by year. This has the benefit of keeping the GRIT score of a given player-season consistent because it’s only dependent upon other players’ component scores from the same year.

As an example, under the old methodology, Hank Aaron’s TALENT score from 1965 could change if Ryan Howard put up an extremely high TALENT score in 2008. While the effect would be negligible, the new method ensures consistency from year to year. This modification resulted in slight changes to the GRIT scores for all player-seasons and, consequently, there has been a slight reordering of the all-time GRIT leaders.