Dude, you have blown my mind... "my mind has been blown".

The possibility that the third-spot, perhaps the most coveted spot in a baseball line-up could, in fact, be a give-a-way spot... wow. I mean, I could accept that the number nine spot could be rather easily interchanged with the number one spot as a "second leadoff" spot, but this is, as Ornette Coleman would say, something else.

I just wish we could see it applied somewhere

I fiddled around with the lineup analysis and ran it for the Mets.

In their optimal lineup, Kaz Matsui bats third.

I just cannot imagine that there is a baseball manager alive who would have the guts to bat someone as unpopular as Matsui in the coveted 3-hole.
Can you imagine what the beat writers would do to Willie Randolph if he told them that the Opening Day lineup had Matsui batting third? They'd run him out of town before the second game of the season.

The Lineup Analysis feature is fascinating, though, and lots of fun to play with.

Thanks for making it available to us.

Morong discovered an error in his original calculations on the #3 slot. Also, the #9 slot has inflated importance because he combined leagues (such that good production in #9 = AL team = more runs). He's got new #s, and if program is changed to reflect that I doubt it will still put bad hitters in #3 slot.

David's theory about spreading out easy outs is interesting, and I dont know if it's true. But this program can't really shed light on the issue because it isn't a true inning-by-inning simulation. It just takes each hitter's OBP and SLG, multiplies by fixed coefficients, and produces a R/G total. And since it's static rather than dynamic, it can't possibly work: if you move the traditional 3-4 hitters to hit 1st and 2nd to take advantage of their high OBP, the program should reduce the 'multiplier' for their OBP since they are now followed by inferior hitters. But the program doesn't do that.

Guy is right of course about the limits of the program, at least Ken Arneson's perl version which David rewrote.

Results which undermine the conventional wisdom are exciting, but a couple of studies by Tom Ruane bear on this discussion about possibly having a "leftover" or "below average" hitter in the #3 spot.

1) most recently an article at www.retrosheet.org, (following the research link) called "Evaluating tradional lineups" which addresses the question pretty directly. He used a markov chain analysis and 1993-2004 base-out transition data separated by leagues. For the American league, the best simulated lineups involved flip-flopping the current #3 and #4 hitters. For the National league the best lineups had a little more variation around the 3 spot; dropping the #2 hitter to the 3rd spot a few times, and bringing the #4 hitter forward most of the rest of the time. Basically all the best NL lineups involved shuffling the #2 through #5 spots among each other. So in his results, the #3 spot still gets filled by a "good" hitter

2) A much older study, on Situational Hitting, available on the baseball think factory site (http://www.baseballthinkfactory.org/btf/scholars/ruane/articles/situational_hitting.htm)
may shed some light on the "skills" best suited for hitting 3rd, and thus indirectly on what type of hitter is best used in that spot.

It contains a very interesting chart showing the percentages of base-out situations faced by each lineup position. [The leagues appear to be combined, and the data is from retrosheet for 1982-1983 and 1987]. The leadoff hitter, for example, being guaranteed a leadoff at bat once a game, had 40.4% of his plate appearances in a nobody on/nobody out situation. The #4 hitter was 2nd most likely to get a lead off type plate appearance. The #3 hitter was the most likely to bat with nobody on and two out, and the least likely to bat with nobody on and nobody out. The benefit of high on base percentage, particularly just reaching first, is minimized with this distribution of situations. The #3 hitter also batted an above average percentage of times with runners in scoring position, but about an average percentage of plate appearances in situations which include a runner at first.

In other words, the #3 spot had very limited leverage for ability to advance runners with extra base hits, but extra leverage for hitting singles. I'd infer from this distribution of situations that the ability to draw walks and hit doubles had somewhat less leverage in the #3 spot than for example the #4 spot, who had more opportunities to advance runners from first, and as noted, a higher rate of "leadoff" at bats. Consideration of these subtleties (beyond just overall OBP and slugging percentage) would probably increase optimization a little.

From the situational data, it seems like a spot suited for someone with at least average power and an above average batting average. By virtue of his position near the top of the order, the #3 hitter gets an above average number of plate appearances over the season, and his overall hitting skills are certainly leveraged. This conclusion could be undercut by ripple effects of situations passed to "down-lineup" batters, but Tom Ruane's markov chain analysis in "Evaluating Traditional Lineups" suggests this doesn't happen. The #3 hitter should be a good hitter.

as a followup to my earlier comment, Tom Ruane's recent article referred to an earler article by Mark Pankin. Now that I've looked at that article again I see that Mark's analysis includes regressions on component hitting skills. His results suggest that my specific speculation about the walk rate of the #3 hitter might be wrong.