Fun Tidbits - Poking Around With 538's 2016 Elo Ratings

Rather than attempting another large-scale, code heavy analysis (see last week's post) this week, I decided to do some simpler little projects just with the Elo ratings I have scraped from FiveThirtyEight. Are they rigorous? No. Are they a little fun to think about? I hope so!

In an effort to keep with the spirit of this blog, I included the methods used so others can learn. Feel free to skim to the results if that's all you want.

I did an unpublished version of some of these with NFL ratings last winter, and felt that some might translate over into baseball.

The spreadsheet I used is located here on Google drive for reference.

Biggest MLB Upsets in 2016 (so far) --- According to Elo

This is a simple analysis. The absolute value in Elo ratings was taken:

=ABS(Home Elo - Away Elo)

And then a simple If/then filter was applied to runs scored to determine if the team with the lower rating won.

=IF(OR(AND(Runs Scored<Runs Allowed,Home Elo>Away Elo),AND(Runs Allowed>Runs Scored, Away Elo<Home Elo)),1,0)

Or, in English:

• If one of these conditions are true
• A. Home team scored less points than away team AND home team was rated higher
• B. Away team scored less points than home team AND home team was rated lower
• THEN return a 1
• Else, return a 0

Then these results were sorted by the above condition, and then by the absolute difference of Elo ratings.

I highlighted the Cubs just to make a point. As mentioned above, I did the same thing last winter for NFL Elo ratings, and the results made much clearer sense. I think this highlights a major difference between the two sports. The sheer number of games played in baseball lends itself to being less of a surprise that the Cubs were a part of 9/10 of the top 10 upsets this season. Whereas the Patriots losing to the Dolphins last winter was a much bigger deal.

2016's "Boringist" Games

A similar analysis was done, but instead sorted by the higher rating team winning and a difference in runs using the formula:

=ABS(Runs - Runs Allowed)

This would give a list of "boring games", where the higher rated team won and by a lot of runs. One might think that the stands were rather empty for these.

2016's "Closest" Games

Here's the flipside, where I sorted on games with the lowest Elo difference, and that were only won by one run.

Best,

Bryce