January 22, 2016

Internet Baseball Awards

The Jokester-Free AL Player of the Year

by Tom Tango

I love crowdsourcing projects. I love them because they take virtually no effort to set up, and yet we get a huge amount of information out of them. They work because the voters are the ones doing all the work, and whatever biases they may have get cancelled out, so that we're left with a reasonable view of the perceived truth. That's if everyone is playing fair. Sometimes, the voters try to game the system.

When I run my annual Fans Scouting Report, it became very apparent very quickly that ballot stuffing was a problem. The target of the fans, every year, was Derek Jeter. So, I knew I had to create a process to throw out ballots. Bill James 30 years ago turned the ranking system over to his readers in The Baseball Abstract (and in fact, it was that initiative that inspired my projects). And he also noted that fans would try to game the system by listing Don Mattingly high and Eddie Murray low (or vice versa). James said that if any of his readers tried to do that, he'd just throw out their ballots.

So, we come to the IBA. I love the IBA, for all the reasons I've already noted. Except for one: There's no throwing out of ballots. Normally, it doesn't matter because when you get hundreds of votes, a few dishonest votes won't affect anything. Until it does. And in the Trout v. Donaldson race, it actually did.

There's actually a fairly simple process to voting without resorting to the point systems being employed. Before we bring that up, let's discuss the reason that we use the points system. We're asking hundreds of voters to vote on dozens of people. This is in fact just like voting for a political leader, except in political elections we don't give them "place" votes. In effect, those candidates go through a "run off" system: After every vote, someone will drop off.

And then we vote again. And after another drop out, we vote again. And the cycle continues until someone gets 50%+ of the vote. This is a first-past-the-post system. The points system is a way to have this run off, without doing a runoff.

Since this is the internet, we don't have to ask people to go through multiple rounds of voting. We can simply ask everyone to "place" their preferred candidates, 1 through 10. And as people get dropped off, everyone below that person on the ballot moves up. If you vote for Pete Rose no. 1, he will get dropped off, and your no. 2 moves up to no. 1. This is called an instant runoff method.

Now, on to the main event: Trout v. Donaldson. There were 374 ballots cast, meaning the first-pass-the-post is the first one to get at least 187 votes. Here's what the initial totals look like:

• 182 Mike Trout
• 179 Josh Donaldson
• 5 Dallas Keuchel
• 1 (several players)

The first thing we do is throw out anyone who was listed, anywhere 1 through 10, on fewer than 187 ballots. Suppose, for example, that Donaldson was listed on 179 ballots, all at first place, and nowhere at all on the other 195 ballots. Well, he could never get first-past-the-post. Think of Donald Trump. Should Trump be allowed to "win" if he gets 179 first place votes out of 374 ballots, and not place at 2 through 10 at all in the other 195 ballots? I think most would agree he shouldn't get the win.

But, maybe I'm wrong. Anyway, let's proceed.

So, after throwing out all the players with fewer than 187 votes, we are left with these first place votes:

• 184 Trout
• 181 Donaldson
• 5 Keuchel
• 1 Machado
• 1 Cain

So, Trout picked up 2 votes, as did Donaldson. Notice however that we're down to 372 ballots. That's because two people put in joke ballots. When Salvador Perez and his first-place slot was removed, none of the remaining five were listed on his ballot. His ballot included all Royals players—yet somehow ignored Cain! The other voter had CC Sabathia first, and you don't want to know the rest of his ballot. (Note that Baseball Prospectus gave me the ballots with user names and user IDs redacted.) Because of the two jokesters, first-past-the-post is now set at 186+ ballots. The race is on.

We knock out Cain and Machado. As it turns out, Trout and Donaldson each pick up one vote.

• 185 Trout
• 182 Donaldson
• 5 Keuchel

Finally, where do the Keuchel votes go? Two went to Trout, three went to Donaldson. The final tally:

• 187 Trout
• 185 Donaldson

And based on a first-past-the-post system using the instant-run-off method, Trout wins the AL MVP according to Baseball Prospectus readers. Unofficially anyway.

Next time, I'll look at the three-way race in the NL Cy Young.