The Case for Human Intelligence

There are a hundred projection systems out there, and all of them are great. ZiPS, Steamer, Marcel—I could go on and on…but none of them will be right. They’ll all be accurate; these are good systems, designed by smart people, but in the end, the old trope about the game being played on grass, not paper, is true.

The fact is, we can’t possibly know everything about a player, let alone have enough inputs in a projection system, to account for all the variables. What we try to do in projecting players is project the results that take into account as many of the variables as possible, but I believe that is where there might be an edge.

An edge? I think every fantasy player (or gamblers, and let’s be honest, the Venn diagram on those two has a big overlap) suddenly pricked up his or her ears. While projection systems probably get most of it right—90 percent seem high to anyone?—there are those decimal point jumps that we get from season to season, from iteration to iteration, or even from generation to generation.

But none of them account for the kind of things like what happened last year with Mookie Betts.

Betts went from top prospect to MVP candidate due to an increase in power, average and on-base percentage over even high expectations. For 2016, ZiPS projected Betts to post a triple-slash of .297/.355/.474, with 18 homers. Steamer wasn’t far from that, pegging him for .300/.361/.474, also with 18 homers. The consistency in his projections is pretty astounding, indicating that Betts was “easy” to project, or that there was a widely accepted range for this projection.

What happened, as we know, was very different. Betts put up a .318/.363/.534 line with 31 homers. While his OBP was about right, the rest was off enough that it’s clear the systems all missed something.

Chili Davis, the Red Sox hitting coach, thinks he knows what that was. Davis helped Betts and several other Red Sox hitters switch to a new type of bat called the Axe Bat. (It should be no surprise that Axe is now selling a Betts-inspired model and using Betts as an endorser.) The bat has a different kind of handle that has been shown to change hitting biomechanics, which can be a major positive for at least some players’ swings.

I’m not here to debate the handle, but it’s a good example of my point. Betts made the change late in the 2015 season, one that is knowable, but was clearly unknown to the projection systems. If you weren’t paying close attention, you might not have heard about this until Axe made the signing announcement in July. Again, there was a league-wide increase in homers, so I’m not debating that part of this.

But what if we’d known that Betts’ switch would really open up his power and bat control? That seems to be something we could know, perhaps very early in spring training. With all the major projections systems within a small range, we have perhaps one fact that could have changed it by 60 points of slugging and 75 percent of homers.

In the intelligence community, this would be called a conflict between signals intelligence and human intelligence. The projection systems were right, given their inputs, but something as simple as a beat writer sharing a story could have given a significant advantage to someone who could process this properly.

That’s not to say beat writers or national columnists should be out there hunting for this kind of information. It might have been an interesting story, but processing it is difficult and asymmetrical. If Andrew Benintendi shows up in the spring swinging a new bat, should we suddenly bump up his power projections?

What this shows us is that there is likely a large gap in information, one that someone could be exploited for gains by someone both in obtaining the information and correctly analyzing the information.

Allow me another wild example: let’s say Corey Albertson decided to take some of his winnings and set up a news site. (Who’s Corey Albertson you ask? He’s a former FanDuel and DraftKings champion with estimated winnings well over $5 million in his career. He could afford to set up a news site. I single him out here because he’s thoughtful and successful, and has put himself out there as a representative of players. I don’t think he has any nefarious intentions, and he’s a downright likable guy.) He could send beat writers to every team, looking for precisely this kind of information. Easier, he could drop a couple hundred bucks into the pockets of existing beat writers, clubbies, or team officials around the league.

We don’t need to have a fantasy Arnold Rothstein to scare everyone, but the fact is that it could happen. There’s a clear information gap that exists outside the sabermetric bubble—and one that not only is exploitable but instructive to all fantasy players.

There’s information of value—Betts’ bat is just one example—and information that’s likely of no value. “Best shape of his life” and such phrases are going to show up in stories this spring, just like every other. New pitch? New grip? Great relationship with the catcher? New video intern who has a knack for finding tells? LASIK in the offseason? Took up surfing? Gave up meat? These are all examples of stories we’ve seen in the recent past. The trick is to collate this information and figure out what is important.

The only way to do that is more information. While it often seems we’re swimming in data, we’re not always close enough to the shore of context to do much. It’s time to move away from pure data and algorithm and get the advantage of human analysis.

References & Resources

• Jeff Passan, Yahoo! Sports, “Why the Axe Bat, Dustin Pedroia may help make the round handle obsolete”
• Ted Berg, USA Today For The Win, “Curious axe-handled bats catching on in Red Sox’ clubhouse”
• Jen McCaffrey, Mass Live, “Dustin Pedroia, Mookie Betts using Axe Bat to prevent injury, improve swing and offense”
• Ben Rains, Sports Illustrated, “Axe Bats signs Mookie Betts to sponsorship deal”
• Keith Wagstaff, NBC News, “Fantasy Moneyball: How Algorithms Are Minting Big Winners”