Yesterday was one of my favorite days of the baseball offseason, as it was PECOTA day. The publication of the full Baseball Prospectus projections for the upcoming season is marked on my calendar because it always occurs near the beginning of spring training and is arguably the most accurate of the various projection systems. In perusing the site’s 2018 expected results, a theme emerged: BP likes the Yankees more than other contemporary projections.
While Steamer and ZIPS jointly forecasted 91 wins for the Yankees in a dogfight for the AL East title, and USA Today’s projections pegged the team for 93, PECOTA sees the Yankees as a 96 win team running away with the division. The Boston Red Sox are ticketed for a rather pedestrian 87 win season, and the rest of the AL East trails behind.
96 wins, according to PECOTA, sets the Yankees third in the AL, behind Houston (99) and Cleveland (97). All three teams are seen more generously than Fangraphs’ combined Depth Charts projections, but the Astros are a single game better and Cleveland three games better, neither benefiting from the jump PECOTA gives the Yankees.
So what is it Baseball Prospectus likes so much about the Yankees? It must be that PECOTA sees more upside in individual Yankee players, right? Take a look at the difference between the Yankees’ projected starting lineup and pitching staffs’ Fangraphs and BP WAR:
Strange. Overall, PECOTA projects that the most important Yankees will actually be worse than their Fangraphs projections. The projected bench players - Ronald Torreyes, Clint Frazier, Tyler Austin, Tyler Wade and Austin Romine - add less than a half-win in positive differential over their Depth Chart projections. Despite being bullish on the Yankees as a team, PECOTA doesn’t really love the individual outputs of each player.
In this case, then, the sum must be greater than the parts. Maybe PECOTA is bearish on the AL East as a whole, relative to the Depth Charts projections:
Overall, the Yankees’ AL East opponents are projected to be about nine games worse by PECOTA. Obviously, the lion’s share of those extra wins will stay within the division, giving some room for the Yankees to scoop up their surplus victories PECOTA projects. In other words, the Yankees are picking up the scraps from a weaker division, not unlike what we can expect from Cleveland or the Washington Nationals.
That doesn’t explain the drop off in individual projections, but I think the methodology behind the various projection systems does. PECOTA’s algorithms don’t generated expected values for teams, just players. The raw data spat out has to be massaged by BP’s staff to come up with reasonable team projections. For example, PECOTA projected a bunch of terrible players making up three of the five AL East teams, and the BP staff follows that up by assuming the Yankees will “pick up scraps”.
The key difference between PECOTA and other systems is in the former’s emphasis on player comps over trends. PECOTA works best by sorting players into buckets and then analyzing how each bucket performed moving forward, whereas other systems will focus more on upward or downward career trends. The fact that some of the Yankees’ biggest names - Stanton, for example - have so few true historical comps drags down their respective projections.
All projection systems are just expected value generators. The games still need to be played and almost nobody will equal their PECOTA exactly. The one thing ALL systems can agree on though, is Yankee fans are in for a fascinating 2018.