The Red Sox are the two-time defending American League East champions, and now they've just added the best free-agent slugger of the winter, J.D. Martinez. Armed with a first-time manager in Alex Cora, they enter the spring expecting to be the division favorites, which is partially what they are by the FanGraphs projections that have them tied at the top of the East.
The Yankees made it to the seventh game of the American League Championship Series last year, and they added (via trade) the best-known slugger in the game, Giancarlo Stanton. Armed with a first-time manager in Aaron Boone, they enter the spring expecting to be the division favorites, which is exactly what they are by the Baseball Prospectus projections, by nine games, though that will drop slightly when Martinez is added.
The only correct takeaway here is that these are two extremely talented teams that ought to be in a tightly contested battle all season, ending with one of them winning the division and the other hosting a Wild Card game. But we're talking about a gap of nearly 10 wins between two of the most respected projection systems around, which is a lot. What's fueling this?
It's important to remember that projection systems aren't predictions and aren't intended to be, and they can't predict every unexpected breakout or injury or in-season trade acquisition. Nor should we really want every projection system to say the exact same thing, because that wouldn't be interesting, but you'd expect them to say mostly the same thing. The good teams are likely to be good, and the rebuilding teams probably aren't. That's why there's a pretty decent correlation between preseason projections and end-of-season win totals, and you don't see this kind of disagreement in any of the other five divisions.

In both systems, and most likely in most fan predictions, Cleveland and Houston are slated to take the other two AL divisions, and once again, the Dodgers, Cubs and Nationals look to be the class of the NL. The biggest disagreement between the two sites is four games, coming in the NL Central (where one sees the Cubs as four games better than the Cardinals and one saying eight games) and AL West (12 games or 16 games, take your pick).
Four games isn't a lot, when looking ahead. Seven or eight or nine games is. Where do they disagree? We looked at the 74 players who have a depth-chart entry at both sites, and we found the biggest outliers. For many players, they do agree. FanGraphs projects Aaron Judge for 4.1 WAR, while BP says 3.9, which is basically identical. Both sites have Tommy Kahnle at 1.1 WAR, and Eduardo Nunez at 0.9 WAR. (Two WAR, as a reminder, is league-average, and four is star-level.)
There are a few differences, of course; there has to be for this gap. Baseball Prospectus likes Dellin Betances (+2.2) far more than FanGraphs does (+1.2), as well as Chris Sale (+6.2 vs. +5.4). FanGraphs has Steven Wright as being below average but pitchable (+0.8), while BP has him as being a disaster (-1.1).
But we're looking for the biggest discrepancies, and the five below are the largest. Which side do you take?
Jackie Bradley Jr., +2.4 wins (BP 1.0, FG 3.2)
The biggest disagreement comes with Boston's center fielder, and it's probably not hard to see why. As if it wasn't hard enough that his biggest value comes on defense (fielding being traditionally the hardest to quantify), Bradley has been wildly inconsistent at the plate. He was terrible in 2014, above average in 2015 and 2016, and below average in 2017. Compound that by a tale of two halves in 2017, with a strong first half (.280/.363/.490, 122 wRC+) followed by a miserable second half (.204/.277/.302, 51 wRC+).
So, you'd expect some disagreement, but this is extreme. They both expect a slight rebound at the plate, which is fair (though FanGraphs is more optimistic), but the difference is that BP viewed his defense as a net negative. The right answer here is probably "in the middle," but given that Statcast™ showed Bradley to be one of the best outfielders in baseball last year, we'll take the higher end of this disagreement.
Giancarlo Stanton, +2.4 wins (BP 3.9, FG 6.3)
BP, quite reasonably, takes the more conservative view here, because although Stanton did have a monster year, it was also the first time since 2014 he'd taken 500 plate appearances. We're talking about a difference of 100 points of projected slugging here, from BP's .558 to FG's .665. Considering that even last year he slugged "just" .631, and that was a career high, the lower end of this feels more appropriate -- though again, "the middle" is probably right.

Xander Bogaerts, +2.2 wins (BP 1.4, FG 3.4)
Entering play on July 6, Bogaerts was hitting .308/.363/.455 (114 wRC+), which is more or less the exact line he'd had in 2015 and 2016. Then he was hit by a pitch, played through the injury and hit just .232/.321/.340 (74 wRC+) the remainder of the season. If he does that all year, then the 1.4 projection is probably right. But we saw two excellent seasons in 2015 and 2016, and we were going to see another one before the injury. Bogaerts hasn't quite become Francisco Lindor or Carlos Correa, yet an above-average season seems more than realistic. Take the high number.
David Price, +1.5 wins (BP 2.1, FG 3.6)
You know the Price story by now: A good Boston debut followed by an injury-plagued 2017 that ended with some strong postseason bullpen work. Price claims he's feeling good and ready to excel, and he'll need to be; both sites have him putting up under 2 wins in 2018. That makes the hesitancy here warranted, though he's never failed to perform when he's on the mound. We'll cautiously take the high number, knowing he could return to the disabled list at any time.
Dustin Pedroia, +1.4 wins (BP 1.4, FG 2.5)
The two sites don't disagree on his performance all that much, which makes sense because he's been around for so long. One sees a .351 OBP, the other a .358. One sees a .391 slugging, the other a .413.
The difference is in playing time, as BP has a conservative 423 plate-appearance estimate, and FG has him down for 595. Given the uncertain timing of Pedroia's return from knee surgery, it's probably fair to take the under on this one.
The answer here isn't that either site is "right" or "wrong," of course. Projection systems do pretty well for themselves, but they can't tell the future. What you can say for sure is that of the six division races, this is going to be the one that's the most difficult to call. The experts won't be able to come to consensus and neither, apparently, will the computers. Just like every year, the Red Sox and Yankees are the class of the East. It's too close to call. Again.