Friday, August 8, 2008

Cal Ripken, 1983

One of the things I'm looking to do long-term is to calculate WAR for every Oriole season ever. Considering that I'm in the (very) preliminary stages of this, it's going to be a long while (if ever) before I really get down to doing it. Part of the problem is my lack of skill in doing these things where database skills are imperative. The other part is simply deciding what exactly is *the* best way of doing it, meaning which data sources do I use.

As sort of a sanity check, I decided to start with 1983 Cal Ripken. This was, of course, his first MVP season. He had won the Rookie of the Year in '82, but this probably would have been the season where he truly broke out. It was also his first full season at shortstop.

First what I did was to figure his batting runs above average. There were a few ways to do this. I could just look at bb-ref and take their word for it at +40.2 runs. Fangraphs is another option for the lazy, and they check in with an estimate of +32.68. Call it +32.7?

The first one is nice because it adjusts for park. The second one is nice because it changes its run values based on base/out state rather than using a static estimator for the value of a single, double, etc.

I could also recalculate it on my own using a more advanced linear weights formula derived from Base Runs that uses custom weights for each team and year that models the actual run scoring of a particular team. Those weights, as with many more good things, can be found on tangotiger's site. The down side of this estimate - +37.39 - is that while it is more "exact", it is also more context specific - two players on different teams with the same batting events in the same run environment will have different values based on the hitters around them.

So in the end I decided to stick with the +40.2 value found on bb-ref. And not because that was most favorable to Cal.

Next is fielding. I'm using Sean Smith's TotalZone system because it's both the most thorough and the most transparent zone-based historical fielding system that I know exists. He rates Rip at +9 runs with an additional +1.6 on turning the double play, for a total of +10.6 defensively.

Now position. I usually use a position adjustment based on defense (so that we don't assume average 2B = average SS = average LF), but I don't have much confidence in what those numbers should be in 1983 when there was an even larger offensive gap at SS. Luckily Tom Ruane has me covered with his awesome chart which puts the average SS in the 1983 AL at -0.018 run/PA. Taking this and multiplying it by 726 PA gives Rip at staggering +13.1 for position.

RunsAboveAverage = offense+defense+position = 40.2 + 10.6 + 13.1 = +63.9

64 runs. Better than average. Holy Cow, are you kidding me?

Lets sprinkle in an extra +22.5 (a guess, but a good one) for replacement level to put Cal at +86.4 above replacement level. Now use 10.5 runs per win (a guess, but a good one) to convert from runs to win and Cal is at 8.23 WAR.

If I actually continue on with this project, I would think that his 1984 and 1991 seasons would be the only seasons from an Orioles hitter that could touch this one, but I could be wrong. This is a peak level Albert Pujols season. Maybe better. The only qualification is whether our defensive numbers are close to correct. Baseball Prospectus' FRAA has him at +24, and of course we know what his reputation would tell you his fielding was worth back as 22 year old kid. But even if we're wrong and he's -11, making him one of the lesser fielders in the league, he's still up above 6 WAR.

Baseball Prospectus has him at 13 WARP. Great season, but even this one didn't net the Orioles 13 wins.

No comments: