Search THT:

Or you can search by:

August 12, 2010

Scouts, statisticians and wizards

I was about to leave a comment at The Book Blog in a thread titled When to go from the eyes to the numbers when I realized it was probably worth delving into here as well. It opened a philosophical door in my mind that occasionally opens and closes. It seemed a good time to take advantage of the open door and get my jumbled, innermost thoughts peer reviewed (or at least committed to paper for future reference)!

In post No. 8 of the thread I linked above, Nathan asked:

I guess the real question is, can anybody, just by watching 20 games, tell the difference between a .275 and a .300 hitter? I’m referring to batting average here which I know is lame, but the point is that the difference over 80 at-bats (roughly 20 games) is 2 hits! Can an observer notice 1 hit every 10 games?

Now, the fact that two hits separate .275 from .300 indicates that stats aren’t great in that small of a sample, but really, what can a scout see, can he see that this player is hit every 10 games good?

My response is No. 9 and you can view it for yourself if you like. In fact, I highly recommend taking a look at the whole thread. I'll hit the highlights a little later. Really, it's unfathomable to me how any person can divine the difference between a .275 and .300 major league hitter when he's in high school, rookie ball, or even Double-A. The rate of attrition among minor leaguers could mean that perhaps they can't, at least not with any real degree of accuracy . The problem I see with figuring it out is that there are so many different inputs that make up a good hitter. Coordination, reflexes, strength, eyesight, reaction time, mental toughness, intuition, temperament, focus, etc. all have some bearing on whether one player is one hit per 10 games better than another.

Numbers, with sufficient sample size, of course, allow us to proxy the net product of all the myriad inputs that make a player a player. But the sample size is the limiting factor. Beyond that, Cliff Lees abound in the baseball world, players whose skill sets undergo such massive changes that the previous data becomes nearly worthless. And anyone who watches the game knows that other difficult to explain phenomena occur. Raul Ibanez and Adam LaRoche come to mind.

It seems to me that scouts and statisticians are asking similar but ultimately different questions. The scout's job is to learn the player, to become familiar with his mechanics, his strengths, his flaws, how he handles himself under pressure, how he spends his time off the field, how he relates to his family and loved ones. By doing this, the scout tries to paint as detailed a picture of the player as humanly possible so he can convey to his employers how much that player is worth. He judges the quality of the player's skills. Knowing the quality of those skills and knowing which ones can be improved, he can estimate where a player is at now and where his ceiling is.

Statisticians do something else entirely. We ignore the majority of the inputs and focus on the measurable output. When we look at numbers (or at least when I look at numbers—maybe I'm being presumptuous in using "we"), we're trying to quantify a player's skill in a succinct and tidy manner. We don't care if Chris Coste has a godawful swing, that it's of poor quality and so very unlikely to stick at the major league level. We care that he produced a .316 and .326 wOBA in '07 and '08 respectively. And we care that he was a catcher, making him above average for his position. We don't care that Milton Bradley's temperament is at best questionable. We just care that he can mash the crap out of a baseball when healthy. (Yes, I'm simplifying.)

I must admit, that philosophical door in my mind that I mentioned earlier rarely stays open long and now it is closing quickly. I hope I got my views across clearly enough for some good dialogue. Oftentimes I've heard the work of saberists referred to as statiswizardry (which can be intended to compliment or disparage). Ultimately, I think it's the scouts who do the magic by divining the quality of a player's individual skills.

And I think that partially explains why some casual fans are resistant to openly accepting saberist ideas. All we have is charts and graphs and output from R and Stata. It's all very convincing and useful stuff to statistically oriented minds, but the scouts have something more popular with the masses: Magic.
Posted by: Brad Johnson

July 31, 2010

Measures of Qualls-ity

Arizona has shipped Chad Qualls to the Rays, eliciting responses ranging from anger to mockery to happiness.

Qualls is having an off year, his worst showing of the past four campaigns. It's just not as bad as his bloated ERA indicates. Using pitch-by-pitch linear weights to derive "ERA", both rvERAe (based on batted ball types) and rvERAa (based on actual hits and outs) have jumped for Qualls in 2010. This is along with an apparent downward trend in both whiff and ground ball rates.



Not a stud, not a dud. You may even expect Qualls to regress towards his career/recent levels—improve, in other words.
Posted by: Harry Pavlidis


Click here for more THT Notes.