Baseball Musings: Run Estimation, II

February 08, 2005

Run Estimation, II

I'm feeling better today and was able to think through the problem a little more clearly. Thanks for all the comments on the last post.

I did a simple calculation using the latest Bill James runs created formula for teams, found in The Bill James Handbook 2005. I first imagined a team that had 6000 plate appearances and got 1500 singles as a result (no walks, extra-base hits, etc). The RC formula would estimate that the team would score 422 runs. I then added 44 singles, giving the new team 6044 plate appearances. The run estimate for this team was 444 runs, or 22 more. So estimates from 18 to 30 do appear to be in the correct range.

I would offer one caveat, however. Because Eckstein is getting 44 fewer outs doesn't mean those outs aren't being picked up by other fielders. Some have pointed out that there is noise in the popup data; pretty much every popup is caught by a fielder. There is noise in the groundball data as well. There are going to grounders where both the third baseman and the shortstop have non-zero probabilities of fielding the ball. If the third baseman, due to superior range, cuts in front of the shortstop often he'll take outs away from the middleman. In most cases these should even out, but in some situations the shortstop will appear not to get to balls that indeed are outs.

Posted by David Pinto at 09:35 AM | Defense | TrackBack (0)

Comments

A team which hit all singles would be an extreme to say the least-- as a quick and dirty methodology its okay, but is RC really accurate to 11 runs in such a usage? Or 30 for that matter?

Posted by: john swinney at February 8, 2005 11:45 AM

John has a valid point. A single on a team that hits all singles (let's call them the Martiners) is worth less than a single on a team that leads the league in slugging (let's call them the R. Sox, no, that's too obvious - the Red S). Perhaps test it on an average team.

OTOH, Eckstein's 3B for much of the year was Chone Figgins, #4 on the 3B list. So Mr. Eckstein may have had some balls "stolen" by Mr. Figgins.

Posted by: Jeff at February 8, 2005 01:35 PM

Sorry David, but I'm not with you, as I noted in the other thread.

Eck *is* in your model getting to 44 fewer outs that someone else is not getting to. If you were to throw out infield pops, you will get virtually the same results. (Maybe Eck won't be -44, but someone else will. The distribution will remain pretty much the same.) So, it's not a "shared zone" issue.

You want to add in 44 singles, that's fine... but you've got to take 44 outs away. Once you do that, you realize that the opponent now is going to get about 66 more PA.

I think my Spike/Ozzie example should make it clear that the run value to use is .80 runs.

Posted by: tangotiger at February 8, 2005 02:59 PM

i'm no mathematician, but +/- 30 runs seems reasonable to me if you just think it through.

in the majority of cases, one extra single randomly inserted within a 27-out game is going to cost a defense nothing; it's just another meaningless baserunner. but in a fraction of cases, that extra single is going to knock in a run. in a smaller fraction of cases, it's going to knock in two runs. and in a very small fraction of cases, that extra single is going to open the floodgates. we see it all the time ----- a defense fails to make a play it should have, an inning goes on longer than it should have, an offense gets an extra opportunity to cash in with men an base. over the course of the season, a handful of those extra singles are going to cost a LOT ---- maybe four or five runs. maybe a ballgame.

what the AVERAGE extra single costs is almost beside the point. i think most of them cost nothing, and a good minority only cost a little. it's those few killers you have to worry about

Posted by: l boros at February 8, 2005 03:56 PM