A NOTE ON ARMCHAIR PROJECTIONS
A few caveats about Armchair projections are in order.
While I am confident in my work, I make no claim to the most analytically sound projections. Hat tips to Baseball HQ, Steamer, PECOTA at Baseball Prospectus, Derek Carty's The BAT and The BAT X, ZIPS by Dan Szymborski, and ATC Projections by Ariel Cohen. The more complex algorithmic projections take far more into account than I do, and they do an excellent job at cutting out the noise and turning complex datasets into projected statistical outcomes.
For Armchair Projections, the starting point is a tool of my own creation that takes my chosen inputs from multiple statistical samples and turns them into a projection. I then consider additional factors. Did something in the player's profile change? Did the player play hurt? Did the pitcher change his pitch mix mid-season? Any number of factors might lead me to adjust the initial formulaic projection.
There are several key areas in which an informed human mind can fill in the gaps missed by the computers. The number one area where algorithmic projections fall short is playing time. They miss injuries, likely prospect call-ups, and opportunities they would know about if they could read and process the news. Players who put in the work by adjusting projections based on division of innings and plate appearances can gain an edge on players who do not, and that is something I work hard on before I put my own projections out each season.