Thursday, August 28, 2003

OBP, Baby You Know Me

"We're going to put all our resources into offense, especially on-base percentage, and pitching." - JP Ricciardi in the amazing interview done by Batter's Box. Ricciardi Interview

Back when Moneyball came out, I jumped on the OBP bandwagon like everyone else. I started giving out the Paul DePodesta Medal of Honor to any player who had success in every at bat and did not get thrown out on the bases over the course of an entire game. My fantasy baseball team, now 50-15, took on a decidedly OBP flavor, and I created All-Star teams based on (3 times OBP)+ SLG. I was all in.

Then I came up with the BAP thing (if you're new to this site, read the BAP For Beginners link on the right). It incorporates a large amount of OBP and SLG, and I figured the best way to measure the success of BAP was to compare it to those two very dependable statistics. I set it up that I would measure the BAP, OBP, and SLG scores for every close game, choosing close games because they would be the hardest for any statistic to accurately read. For this contest, close games were defined as any one-run games, and any games where the loser had as many or more hits than the winner.

BAP worked, which didn't surprised me. The shocking news was that OBP didn't. Here are the final results for that first contest:

BAP 41-1
SLG 36-6
OBP 28-12-2

I couldn't figure this out. OBP made so much sense. If you never made an out, you would go on scoring forever. I understood the argument that a SLG percentage of 1.000 for an inning could be a homer and three outs, producing only one run, but an OBP of 1.000 for an inning could be four walks, leaving you with one run, the bases loaded, and no outs.

Due to popular demand, I ran the test again using OPS. This time I covered random series, and these are the current standings:

BAP 39-6
OPS 36-8-1

I haven't kept the separate statistics for this, but I have seen SLG repeatedly bail out OBP in the analysis of one game.

To me, this is huge. If a stat can't be used reliably to measure the winner of a single game, how can it be used to predict pennant winners? After all, we don't play the season in an instant: it is a long, drawn out collection of games. Shouldn't a team be modeling itself after the statistic that gives it the best chance of winning individual games?

I thought about the 1.000 model for OPS and SLG again. That works for an inning, but what about a player? A player that goes 1 for 4 with a homer has an OBP of .250, a SLG of 1.000, has produced at least one run on his own, and may have driven home others. A player that goes 2 for 2 with two walks . . . who knows? Maybe the runner got stranded on third. Maybe he got wiped out on a doubleplay, or a fielder's choice. Maybe he singled, and a runner got thrown at home on the play. Maybe he got a single with two outs and no one on, and was left there.

There's the flaw. In close games, especially low scoring games, it's the stringing together of bases that often decides the winner. Sometimes, that's four hits in a row. But often, it's a team with only four hits for the game getting a walk and homer in the same inning, and winning 2-1 when their pitcher scatters ten hits. Scenarios like this come up over and over.

Here's another way to look at it. Let's take a look at two of my favorite fantasy players, Bobby Abreu and Richie Sexson:

OBP - Abreu .410, Sexson .369
SLG - Abreu .489, Sexson .515
OPS - Abreu .899, Sexson .884
BA + SecA (my newest stat) - Abreu .692, Sexson .679

Almost everyone would argue that Abreu is the more complete and preferable offensive player. But then there's this other part:

Team Runs - Phi 629, Mil 591
Runs Scored - Abreu 82, Sexson 78
Percentage of Team Run Total - Abreu .130, Sexson .132

Team RBI - Phi 599, Mil 569
RBI - Abreu 87, Sexson 96
Percentage of Team RBI Total - Abreu .145, Sexson .169

Runs Scored Percentage and RBI Percentage (with the percentage being a player's percentage of a team's totals) is not useful when one team has scored a lot more than the other, but in this case it's significant. Sexson has almost as many runs scored as Abreu and more RBI, on a team that is scoring less. Sexson has batted in only the three and four spots in the order this year, Abreu mostly the same. They have almost the same number of plate appearances.

So why is this happening? Statistical fluke? I hate that answer. Is Sexson a better baserunner? Hard to believe. Is Sexson hitting more in the clutch? Possibly. Or is OBP a little overrated? Have we swung too far the other way?

I don't have the answers to all of these, but I do like posing the questions. So here they are again, and I'd love to hear from all of you any theories you may have:

1. Why is OBP worse than SLG at predicting the winner of single games?

2. Shouldn't a team be building itself around a model that is the most reliable predictor of single games?

3. Why is Richie Sexson driving in and scoring more runs total than Bobby Abreu when Abreu has a better offensive team around him?

4. Why is Richie Sexson driving in and scoring a higher percentage of his team's runs than Bobby Abreu?

5. Could OBP now be underrated by the general populace, and overrated by statheads?

I'll have more BAP scores tomorrow, and lots of football stuff over the weekend. For all you newcomers, it's good to have you.

Dave's Email


Post a Comment

Links to this post:

Create a Link

<< Home