Why the Twins Might Win the Central

[Sssuperdave]

I was originally going to post this in response to the "Twins will not make the wild card" thread, but it got long enough that I figured it was worth a new thread.

 

Vegas, as well as various projections and experts, have the Twins finishing 10-15 wins behind Cleveland. How could the Twins outperform these expectations? Well, luck is an obvious possibility - things like lack of injuries, hit sequencing, certain players outperforming, etc. But, what if the Twins have some strengths that aren't adequately accounted for by these projections and experts?

 

Many of these win projections are based on statistical models. I work a lot with statistics in my day job, so I spend a lot of time thinking about where many statistical models break down. A couple of common areas include:

1. Extrapolation. What I mean by extrapolation is predicting outcomes that are outside of, or on the extremes of, the observed sample a model is built from. Most statistical methods will do a great job of fitting data in the middle of a distribution, but not such a great job at the extremes. Consider OBP - these statistical models probably have some sort of formula equating OBP to runs scored. It's probably very accurate and reliable for OBP between .300 and .400, but not so much when Barry Bonds has a .609 OBP in 2004. The models will give you a result, but it's probably not correct as there weren't any .600+ OBP years in the sample the models were built on. The models might over-predict or under-predict runs scored, but they are unlikely to be correct.
2. Interactions. When you look at the influence of variables on an outcome, it's easiest to look at things one variable at a time and potentially miss when variables influence each other.  To use an example from insurance, say single people get in car accidents 20% more frequently than married people.  If single men get in accidents 30% more frequently than married men, rather than the 20% you would expect based on marital status alone, and single women get in accidents only 10% more frequently than married women, rather than 20%, you have an interaction between gender and marital status.

With that in mind, here are two areas where perhaps the statistical methods underappreciate the Twins:

1. Speed. This is potentially an extrapolation issue. According to this, the Twins are the fastest team in the league. Buxton in particular has elite speed. I wouldn't be surprised if the statistical models and methods underestimate the impact of elite speed on wins. Consider this Billy Hamilton example: https://www.fangraphs.com/blogs/lets-watch-billy-hamilton-make-a-run-happen/. I certainly don't know this for sure, but I wouldn't be surprised if the models aren't accurately measuring the impact of that kind of speed.
2. Interaction between outfield defense and fly-ball pitching. The Twins have elite outfield defense, and have also put together a starting rotation filled with above-average-to-extreme fly ball tendencies. League average fly ball percentage is 35%, which was exceeded in 2017 by all of the Twins starters except for Gibson:

• Odorizzi - 47.3%
• Hughes - 45.5%
• Santana - 42.5%
• Berrios - 40.0%
• Lynn - 36.2%
• Gibson - 26.1%

Last year Santana outperformed his FIP by 1.2 runs based on ERA. Some of that was probably luck, but what if much of it was due to outfield defense combined with all the fly balls? Odorizzi is even more extreme than Santana, so perhaps he stands to benefit even more? That said, Odorizzi already dramatically outperformed his FIP last year (4.14 ERA vs 5.43 FIP), so perhaps the potential additional lift is limited.

 

Could these two things be worth the 10+ expected wins the Twins need to overtake Cleveland? Probably not, but hey - it could happen!