Monday, September 28, 2009

Partial Postseason Prognostication

With only a week remaining in the MLB regular season and only two races still undecided - the AL Central and NL Wild Card - we're seeing more and more articles seeking historical trends that predict postseason success. Last week, we linked to Lisa Swan's article about strong Septembers and Flip Flop Fly Ball's chart displaying the correlation (or lack thereof) between the having the best record during the regular season and winning the World Series.

Well late last week and over the weekend a few more of these pieces popped up and here are three of them (all via BBTF):

Similar to the chart at FFFB, Tom Verducci sorted the last 9 World Series winners by their regular season record and found an almost perfectly even distribution of League Championship pennants and World Series Victories. One championship came from each of the top 7 seeds and two came from teams seeded 8 or higher, such as the 2006 Cardinals who had the 13th best record in the league that year but still won the WS.

It's a pretty rough measure for a couple of reasons. The first, which Verducci points out himself, is that the teams don't play balanced schedules so teams get as much credit for the record in the NL Central as they do in the AL East, and so forth. Secondly, in 2006 for instance, four of the best 5 records came out of the American League. As a result there was an artificially higher chance of a low seed winning a pennant and a World Series (which happened with the Cardinals) since three of them were on one side of the bracket.


Next up, Dave Cameron wrote a piece for Wall Street Journal's website which showed that over the last 7 years, the teams that won the World Series have had a relatively small amount of meaningless games leading up to the postseason.

Dave quantifies "meaningless" games as ones that were played after their team has locked up a postseason berth either by winning their division or securing the Wild Card, but doesn't count ones that factor in the race for homefield advantage or are played against potential postseason opponents. Those certainly are played with a higher level of effort and urgency that the ones the Yankees are about to play against the Royals, one would assume. As Rob Neyer points out, 7 years is also a pretty small sample size.


Lastly, we have a piece by Lincoln Mitchell in The Faster Times which would seem to fly directly in the face of the half-baked theory proposed by Jeff Pearlman that we took issue with last week. Mitchell attempts to identify the team with the best two starters each year and concludes that since 1995, that club has won the World Series only twice.

There are problems with this methodology as well. ERA+ is not a bad measure but it certainly is not conclusive in determining the team with the best two starters when the postseason rolled around. Furthermore, if we wanted to do more of a complete assessment, we would rank each team's top two pitchers, or better yet give them a score on a scale of, say, 1-10 which would more accurately display the difference between each duo. In some years the top two hurlers for the best team might get a "9" and the next best team only receives a 6, but a straight ranking wouldn't convey that. Once the scores were determined then we could look at the results and see how they correlated with postseason success. Then we might have a better picture of how having two dominant starters predicts playoff series victories.


Do you notice a theme developing here? These methodologies are all pretty incomplete. A huge part of the problem is that baseball is littered with variables, there is no control group and you can't just create situations and run them over and over. Of course it's much easier to point out the deficiencies of these findings than take the time to complete the analysis.

All of these writers are admittedly taking a "quick and dirty" type of approach to their analysis and that's completely understandable. At one point, I started a post that was similar to Lisa Swan's about finishing the regular season on a high note (but I wanted to look over the last 10, 15 and 20 games as opposed to the final month) but as soon as I figured out how many tedious calculations I was going to have to do and how much time it would take me to do even a couple years worth of analysis, I scrapped it.

Another factor that makes it difficult to go more in-depth is that, ideally, you want to have a point when you are done with your research. These pieces were written for online outlets where time is always of the essence. You don't want to spend a bunch of time and at the end of the post have to say, "Through all of this research I determined that there is no noticeable correlation in the data."

And this brings me back to a piece that we looked at back at the end of August by Jay Jaffe of Baseball Prospectus. Jaffe brought up the research conducted by Nate Silver and Dayn Perry that lead to the creation of BP's Secret Sauce and applied it to this year's playoff picture. The difference here is that the Silver and Perry went into their research testing all sorts of variables and determined that strong defense, a pitching staff with a high K/9 and strong closer all had high correlation to postseason success. The Secret Sauce is the closest thing we have to a scientific study of what teams are "built for the playoffs" and it's still fairly incomplete.

In all likelihood the Yankees will take the long ALDS, so we still have over nine days before the first postseason pitch is thrown. We're to see more attempts at unlocking the postseason code and they'll probably be pretty interesting. But I highly doubt any of them are going to take the time to shift our thinking with a truly in-depth study of what October baseball really comes down to. Let's just try to enjoy the fact that the Yanks can coast home, throw out their B and C line ups and take things as they come.

No comments:

Post a Comment