ADVERTISEMENT

## TwinsCentric: How much better do the Twins need to get?

- Blog Post by: John Bonnes
- November 13, 2012 - 9:09 AM

In the latest *Gleeman and the Geek* podcast, Aaron Gleeman and I argued, as we are apt to do. Among the questions raised was one that stuck with me: what is the goal of baseball’s regular season? Certainly, it is to make the playoffs, but beyond that, is there an advantage to posting a high win total?

The answer to that question influences the path one thinks the Twins should travel this offseason. The AL Central champion had only 88 wins last year, the lowest amount for any division. It is not unlikely that could happen again next year. It is not unreasonable to suggest that even coming off of a 66 win season, by piecing together even a mediocre rotation, the Twins could improve to a mid-80s win team.

But is that good enough? Or does a team need to win 90+ games to be taken seriously as a champion?

To be honest, I have no idea. Aaron and I have gone back and forth on it throughout the year and again on Sunday night. On the one hand, it makes sense that a better team (one with more wins) would be favored versus a worse team. (Vegas certainly thinks so.) Furthermore, over several games, that advantage would could be more pronounced.

On the other hand, it’s often said that playoffs are random. There are certainly enough counterexamples of underdogs who have held parades at the end of October, including this year.

It occurred to me today that this is something we can test, and it *may* provide a pretty definitive answer. Best of all, it isn’t that difficult to do. Here’s how….

*(Warning: high level stats discussion coming.)*

One sabrmetric tool used a LOT is called a correlation test. A correlation test compares two sequenced sets of data and sees what kind of relationships the two sets of data have. It is by using correlation tests that sabremetrics can definitively say that OBP or SLG is more important than BA, because it more closely correlates with the runs a team score. It is also by a correlation test the we know that xFIP is a slightly better predictor of future ERA than ERA is. We’re going to use it to compare wins in the regular season to series wins in the playoffs.

Here is a link to the data. It’ll consist of all the playoff teams from 1996 through 2012, along with their playoff series wins and also their regular season wins.* We’ll run a correlation test on those two sets of numbers, and the test will return a value somewhere between -1 and 1:

- The closer to 1, the more regular season wins translates to playoffs success. For instance, comparing temperatures in Celsius to temperatures in Fahrenheit would have a correlation of 1. Not only does one go up when one goes down, but it goes up or down proportionally the same.
- The closer to -1, then regular season wins would have a negative correlation to playoff series wins. For instance, comparing how much I cumulatively spend to my checking balance would have a correlation of -1. The higher the amount I spend, the more my checking balance goes down.
- The closer to 0, the more regular season wins and playoff series wins just aren’t related. If I were to compare the total wins of a team to the numbers of migratory monarch butterflies for each city, I would expect the number to be close to 0. The two sets of data mean nothing to each other.

So what do you think it will be? Take your guess, before I do the work. I’m guessing a fairly small correlation, somewhere around .25, which would be similar to the correlation that SABR folks use to conclude that pitchers can’t control if balls in play are hits.

*(Off to enter data and do the math….)*

Wow. The answer is actually quite a bit lower than that. *The answer is just .07*. Winning more games - being a 95 game winner versus an 85 game winner – affords a team almost no advantage in terms of advancing in the playoffs. If I wanted to drive home just how random this is (and I had a little more patience) I could compare the series wins to other ridiculous pieces of data for each team and find one that had a higher correlation. I’d venture to bet that one of these four items would have a higher correlation: team batting average, team errors, average height, or total letters in the names of all the players on 25-man roster. That’s how ridiculously low this correlation is.

To me, that means that success in MLB isn’t qualitative - it’s binary. Either a team makes the playoffs, and thus has a pretty even chance to win a championship, or it doesn’t. To give extra credit for wins is akin to giving extra credit for something like team batting average or how many ex-Twins they have – you might find it interesting, but that doesn’t mean it is important.

It also suggests that if you think the Twins can win the AL Central next year, then a complete overhaul might not be in order. A team does not need to be razed and rebuilt and win 95 games to position themselves to be a champion. Indeed, it earns them almost nothing at all. They just need to be good enough to get into the postseason, even if it’s in a poor division.

~~~

**Three geek notes about the data I used. First, I did all the teams since the wild card began. Second, I skipped 1995 because they didn’t play 162 games, and since I was using win total instead of win percentage, that would have produced skewed data. And finally, for 2012, I only used the two wild card teams who won their playoff wild card game.
*

© 2014 Star Tribune