Who is the most disagreeable Top 25 Under 25 voter this year?

This year’s Top 25 Under 25 has come to a close with the totally shocking conclusion that Auston Matthews is the best of the best. To refresh your memory, here’s the full list:

2019 PPP Top 25 Under 25 Final Rankings

There’s always controversy over how the rankers vote on the players on the list. After all, you have your opinion on things, and you’re sure it’s right, or it wouldn’t be your opinion. And so does everyone else. One thing we noticed in this year’s list is that there was a lot more chaos than usual in the lower half of the rankings. Jeremy Bracco was the first one to get all 17 rankers to put him on the list. That usually happens much lower down.

I’m going to dig in a little more on how we voted, and how much we differed. But first, this is  the summary of the votes you already know about, the ones for the 25 players on the list:

How We Ranked the Top 25

And now for the big list that might be easier to view on the web:

Our Individual Top 25 Lists

Variance in Votes

With 17 voters this year, there was a lot of room for differing opinions, and yet, by averaging out the votes, we ended up with a lot of agreement. There were a few people who really held strong opinions different to the group, however.

I took everyone’s votes and compared them to the final rankings. For this exercise, any non-rankings counted as a zero, and all the other votes were reversed in value, so every Auston Matthews vote was valued at 25, every Mitch Marner (except that one) counted as 24, etc. I then calculated the total variance each voter had from the final rankings, the mean of those differences (ignoring the zeros) and the total number of votes where they disagreed with the final ranking.

This chart is ordered by mean variance from lowest to highest. And what you should be able to see is that while the total variance from the final rankings goes from the lowest (Mel) at 42 to the highest (me) at 110, the number of votes that each voter had different to the final ranking is very similar. The number that varies ranges from 18 (Jensine and elseldo) to 26 (me). The distribution is very even, with seven voters varying on 23 or more votes, and I’m the closest thing to an outlier there.

The driver, therefore, of the big difference in total variance was the mean variance in votes. There is a weak positive correlation between number and mean, but mean variance is carrying the load here. Eleven voters had at least one ranking 10 places or more different from the final result, so there were a lot of strong differences. The other six voters, or less than half, were much closer to the final rankings on all of their votes. I don’t have the biggest variance; that’s a tie between elseldo and Brigstew at 13. I’m one of six voters who had at least one ranking 12 places different from the final result.

I win as the voter who varies the most because of the number of rankings I had different, but mostly by how much different many of them were. The mean variance ranges from 2.19 for Fulemin to 4.26 for Emily, so we aren’t actually talking about huge differences in opinion here. Even for me, 10 of my rankings for the official 25 are out by zero or one place.

Most of my really extreme differences are for players who either missed the list altogether or were ranked very low — even by me. With most of the disagreement concentrated in the lower half of the rankings, the top of the list has a lot of agreement. There is a section below Nylander where nearly everyone ranked the same five or six players in similar spots, but the order varied a lot.

If it was plausible to weight the data so that disagreements in the top 10 counted more and disagreements in the bottom half barely register, all our variance numbers would tighten up together. The trouble with that idea is there would be 17 different opinions on how to weight the list and where the big gaps are between values. It also gets a little silly to do complex analysis of the opinions of 17 people.

Just by restricting this exercise to the top 10, we get this total variance:

Suddenly, Kevin, who has an overall list in the middle of the pack for how close to the final results it is, is identical to me in terms of variance. Meanwhile, Brigstew becomes the biggest outlier, and elseldo becomes the most like the final list.

There were only 14 votes in the top ten that were three or four places away from the final ranking, and five of them were for Travis Dermott. This shouldn’t surprise anyone. Alexander Kerfoot and Timothy Liljegren also had a few each. In the bottom half of the list, there is total disagreement about everyone, and there’s few patterns to be found.

The more voters we have, the less one person’s strong opinions affect the final outcome, and in this list we have both the wisdom of the crowd, some foolish consistency, and a couple of real free thinkers. The whole point of the Top 25 is to get a diversity of opinion, not to find the mean opinion of all of the Leafs fans there are. It’s important to have enough voters that we get a good average, but there has to be room for the opinionated who really believe in some young prospect or the AHLer about to make the jump, even when no one else does. Every single voter had at least one vote that varied from the final ranking by at least five places.

So now you know who was the most opinionated and disagreeable. As if you had any doubts. But what we won’t know for a long time yet is who is right.