The Sabermetric Transformation

At some point, it happens...for some reason, we ask the question: are we evaluating this right?

It happpened to me a bit differently, and it was almost 4 years ago that it happened. But I'll give some background.

I'm 18 years old now, a college freshman. I started watching baseball in 1994 (at the tender age of 7). So you can imagine how depressing it was to have my first season cancelled by something I barely understood and only could decipher with the help of my dad's disappointment and WFAN.

But there was nothing I looked forward to more than the opener in 1995. It was April 27, I think, and the Mets were playing the Rockies. I listened to most of that game, as I listened to Gary Cohen and Bob Murphy paint the ageless Brett Butler's leadoff at bat...it was a nice thing. And I was hooked.

The next few years weren't "die-hard." I was still a young guy then, but by 1996, I'd deciphered a lot of what was happening. The Yankees weren't going to beat the Braves that year! The Braves were really good....oh wait. Got that one wrong, there.

I was good at math in those days (while I'm still OK at math, I hate it, unless it has to do with baseball), so I loved reading the box scores. Numbers telling the results. Batting average was king for hitters. Steals were great, too.

The next step in my evolution was a text-based simulator: Baseball Mogul. My friend bought it, and, luckily, you didn't need the CD to run it. So I played it very frequently (usually in spurts). I best remember the game for it's usage of great names like "Bytch Honry" instead of Butch Henry.

I got a newer version of Mogul two years later, and that one, I took very seriously. Two stats ended up catching my attention in that game: "Production" (which is known more commonly as OPS) and "Team Defense" (which is either DER or 1-BABIP). My friend Craig (who is as into this stuff as much as or more than I am) and I would play that frequently as well, sitting at the computer for hours. One of the interesting things was a fictional league from the 1940s and 50s that we did. I was the Giants and played my games at the Polo Grounds....and I lost a lot of games b/c of bad defensive outfielders in the corners. I replaced them, my Defense score went up, and my pitching improved. I found that interesting.

The turning point, though, was certainly high school. As an eighth grader in 2000, I boldly predicted that the Mets would go to the World Series. They did, for reasons much different than the ones I thought would get them there. But that was cool, and I really thought I'd figured the game out. I remember not being too happy with who won the MVP in 2000 (I think I wanted Piazza to get it, but Kent won. This persuaded me to do the next part of this....).

Come spring, the teachers were describing our options for summer reading. The list of books we had was OK, and I was pretty settled on reading April Morning and 3 other books for sophomore year. But the other option was the "Independent Summer Learning Experience," where we'd pick a topic that interested us, do some research, and then write 10 journal entries describing the process. I spent a week or two trying to come up with a topic to no avail, but towards the end of Biology, I came up with one. I presented it to the teacher, and he said it was good.

"To devise a statistic to determine the league's most deserving MVP winner."

I was young and stupid with this stuff in those days (now I'm just older and stupid), but I had no idea about the field of sabermetrics being so involved and scientific. I just thought that there were numbers and no one ever bothered to adjust them. I was quite wrong.

So there were two main places that really swayed me, and it wasn't reading or finding Bill James that did it (that came after these guys):

Stephen Tomlinson's Blue Jays Abstract
Introducing XR

Tomlinson's work was like reading Darwin's The Descent of Man as a creationist. This guy was projecting the number of wins the Blue Jays would get and was doing a really good job of it. How? Could I do this with Mets stats?

But his little primer for GMs and his article on Paul Johnson's ERP were most influential.

His warnings:

Attention general managers: always check the runs produced formula as part of your evaluation of a ball player.
Attention general managers: always, always, always check the park factor before evaluating yours or other teams' talent.
Attention general managers: always check what you have in your minor league system, or what you can acquire from someone else's system, before investing in "proven" players.
Attention general managers: always check on which side of 27 a player's age is when evaluating the player.
Attention general managers: compute the estimated number of wins a player will add to your team before spending millions.

Wait, 25-29 is the peak? I thought a player's prime was 28-32! Park-adjusted numbers exist?

John Olerud was the selling point in all of this. Tomlinson outlines why the Jays should have kept Olerud, whose production was undervalued in comparison to Joe Carter b/c of RBIs (I did kinda figure out the RBIs thing on my own; players with no runners on base aren't going to get RBIs). And I remembered how solid Olerud was for the Mets (and how he was one of my favorite Mets) in 1997-1999.

ERP was the other eye-opener: ((3*1B)+(5*2B)+(7*3B)+(9*HR)+(2*BB)+(SB)-(Outs*.61))*.16 = ERP. Something that simple could estimate a player's value? At the bottom of that article was the XR formula. So I googled it and came to the second article on that list.

XR was like ERP, but it worked for me more...I dunno why. Furtado made a convincing case, and I still frequently use XR. So the components of XR were definitely part of my little MVP formula, for which I was beginning to realize how much deeper I'd have to go for all of this. I wasn't at the cutting edge; this was developed.

My two other things I thought I needed to add to the XR for my project:

- winning percentage
- BA with runners in scoring position.

Then I found David Grabiner's Clutch Hitting Study, and he's saying that clutch hitters don't exist! And he can prove it! This was all quite a bit for me to handle as a traditionalist.

Stolen bases were risky and not worth very much. Walks were worth almost as much as hits, in the XR formula. Sacrifice bunts are only worth .04 runs? It was all very different.

After much thought, I opted not to put in the BARISP into my formula, so I just had a winning percentage adjustment. (It was something like adding 30*winning percentage to the XR formula; I don't even remember now.) I still believe that BARISP is important to determine value because hits that are with runners on are worth more in a game-context (we see this now with the Win Expectancy approach). But in terms of player evaluation, it's really not worth much.

The project got a 96. I felt bad for the female English teacher who graded it. I think that it was all so foreign to her that she just gave me a 96 to say, "well, I can't give you anything higher because this could possibly be bad, but I can't penalize you for going in depth."

I was now starting to apply this line of thinking to my everyday baseball-watching. I'd watch players in disdain at times when the sac bunt would be dropped (a single is worth 12 times that, was my thought). I still hadn't embraced it all.

My next attack was pitching. It occurred to me that pitchers' ERAs can be skewed by one horrendous game, when, in reality, giving up 30 runs and giving up 9 runs in the first inning are pretty close in terms of negative value. If you could break down a pitcher's pitching performance into 1 number based on the innings and the runs allowed, you'd be grading them more fairly and rewarding the pitchers who gave you the best chance to win the game. (I called the stat "Earned Win Percentage." Basically, if a pitcher did this, the average team would win _____% of the time.) It ended up just being a stat to see which start was better from a pitcher, which was flawed inherently.

Example: 9 innings, 0 runs got a score of 100.00%. 8 innings, 0 runs got 98.25%, because the only start that could beat it was the 9,0. This progressed in similar fashion on two lists: the innings-favoring list and the runs-favoring list.

I wrote an article for Baseball Primer, because I really thought I had something. In retrospect, it was dumb. It wasn't well-researched because I didn't have the time or the information available (to compile numbers, I literally combed through Sportsline boxscores, putting them into Excel). Several times after this, I tried to make a similar model based on probability, but the system always broke down while I was getting my numbers (negatives would come into play, which hurt it a lot). I do think that the idea still has some potential, and if a pitcher's stats seem weird, I look at the game logs to see if it's consistent or b/c of one really bad game.

One of the things I always liked to do was to calculate the numbers myself. I didn't just want to read the statsheets, I wanted to make an Excel spreadsheet that did the calculations myself. That was always my problem with Michael Wolverton's SNWL. While the stat is great, it's too complicated for me to program and too detailed for me to memorize.

As time progressed and I read more stuff, I gradually accepted a lot of the tenets of sabermetrics. "Wasting outs" is not good. OPS is a quick and dirty stat but it counts singles twice. The "three true outcomes" are the most important skills in pitching (I kind of disagree with this; I think that G/F ratio is a bit more important than homers, but that's not all that original, either).

Perhaps it was because watching the New York Giants has been so painful of late, but I find myself counting the days until Spring Training in November. That's why I have this blog. I won't profess to be particularly insightful; occassionally, I'll come up with something, but it's rare. But I want to do something with it, more than just watch. I host a radio show on 91.1 WZBT, Gettysburg, on Sunday mornings here, about baseball (The aforementioned Craig is actually willing to wake up and call over the phone to talk over the air). It was my childhood goal to be an announcer or a sports writer, but I've largely abandoned that and made it into a hobby (I now want to teach history to pay the bills). So I write here these days and I read other people's stuff. I remember a few years ago, another friend of mine almost set up a sports site to do some form of coverage, called FLSports, but it never got off the ground. Which was alright; there's always other ventures, like this one here.

So yeah, I love baseball. I play Out of the Park baseball now, which is like "Baseball Mogul EXTREME" with all sorts of other factors. If I lived before the age of computers, I'd be a child of Strat-o-Matic, I'm sure. I love baseball history, too, although I don't know as much about it as other forms of history, which is sadly disappointing, at times. I wrote a friggin' college essay on wishing I were at the October 3, 1951 ballgame at the Polo Grounds. Yeah, I'm a nerd.

I'm happy to be here, though, and even as I sit at my desk at 12:36 and finish this up, having listened to another Met loss, it's good. The game is alive and well; steroids haven't killed it. The next crisis won't kill it either. And I'll be watching. Or listening.