My plan this year was to create my cheat sheet by aggregating the best algorithmic projections with the best market-based ones. The idea was to optimize the available information and then tweak it slightly to accommodate only my strongest personal preferences, i.e., move ~25 players up or down from the aggregate rankings. To that end, I used Derek Carty's THE BAT and Steamer's season-long projections as the algorithms, and the NFBC's ADP as the market. But it's more complicated than that. 

For starters, I wanted the rankings to reflect the NFBC 12-team online format. (I'll tweak it for the 15 team closer to the Main Event.) The NFBC is unique in that roughly half your entry fee goes toward an overall prize, i.e., you must play to win the overall, not just your individual league. That means you need to be strong in every category, and you cannot tank (or even be below average in) any. That requirement changes the way people play the game. For example, while you can start nine pitchers of any kind, virtually everyone starts 2-3 closers and 6-7 starters. Accordingly, when I converted THE BAT's and Steamer's projections into dollar values, I made the assumption that 2.5 closers and 6.5 starters were actually required. Otherwise, the results vastly undershoot the market rate for closers.

Still, the Bat didn't (as of last week) project saves (or free agents), so I projected saves myself for that set of rankings, and while Steamer does project saves, it does a poor job, in my opinion, arbitrarily giving players like Anthony Swarzak and Ryan Brasier 34 saves each while leaving Blake Treinen with only 23. I'll probably overwrite them in the next iteration. (For unprojected free agent non-closers like Bryce Harper and Manny Machado in THE BAT, I just used Steamer's numbers.)

I had even more trouble accounting for the scarcity of steals, however, as there's no one position you can create for them the way I did for closers, and rather than 30 players getting all the steals (as is the case for saves), the number of steal candidates is not fixed. Rather than arbitrarily overweighting the steals category, I left it alone.

To counter some of the weaknesses in the projections systems, I made a few other adjustments too. I added a per-game replacement value bump to every (non-catcher) hitter projected for fewer than 145 games and every starting pitcher projected for fewer than 28 starts. Steamer, which is much more conservative (Mike Trout gets 136 games, 36 HR) had a lot of players getting this bump, while THE BAT (Trout gets 152 games, 44 HR) had far fewer. This affected the level of replacement value for each, and required different baselines from which to calculate value over replacement in each category, which is the basis for their rankings. I also had to separate catchers with their own special baseline and re-incorporate them back into the hitting pool afterwards for both projection sets.

Finally, I had to make a few assumptions, based on prior research on the NFBC. First, that the resource split for hitting and pitching was 66/34, second that of the 34 percent spent on pitching, 28 percent of that (roughly nine percent overall) was spent on closers, and third that seven percent of the hitting budget (~4.6% overall) was spent on catchers.

But by far the biggest - and most arbitrary assumption I made, you could actually call it a guess - was to weigh the NFBC ranks at 60 percent, and Steamer and THE BAT at 20 percent each. That heavier weighting of the market's rankings somewhat mitigates the weak save projections by Steamer and the lack of scarcity adjustment for steals in an overall contest. I haven't tested this, but I think the market's ADP in the contest for which I'm setting the rankings has better overall information than algorithmic projections run through my formula, adjusted imperfectly to model the contest. That said, there's likely information in the industry's top projection systems that the market is missing - both love Chris Sale far more than the market, for example, and both have him and Jacob deGrom ahead of the 35-year old Max Scherzer, likely because pitchers, no matter how great, tend to decline with age. The algorithms are less subject to groupthink and particular biases to which the market can fall prey.

I am still going to tweak these results with my own strong leans - as I mentioned, I'll move roughly 25 total players about whom I have the strongest opinions up or down. But I have yet to do that for the rankings below. These are simply 20 percent THE BAT, 20 percent Steamer (as run through my formula) and 60 percent NFBC ADP over last 10 days or so, calibrated for the 12-team NFBC Online Championships. One last caveat, the NFBC site's ADP went 400 deep, so anyone who was not listed on all three lists (mostly relievers) were manually tweaked, leaning heavily toward the NFBC's rankings.

Here are the rankings: