I really thought I was on to something, discovering replacement value by looking at the performance of actual replacements in the NFBC 12-team Online Championship format, but I've hit a wall, and I think I know why.
The problem is simply that replacement value is so high it essentially turns the 12-team format into a six-team league. Only 84 hitters have a dollar value above zero, and only 59 have a value above $1. Accordingly, Mike Trout is worth $93 and Giancarlo Stanton more than $97. Trevor Story, Javy Baez, Travis Shaw and Whit Merrifield all have negative values. This can't be right, can it?
It might if replacement value were as ridiculously high in 2018 as it was in 2017, and the above players played exactly to their 50th percentile projections per Steamer. But their projections are mere averages, so some "sub-replacement" players will have far more value than replacement (which is likely to be lower), and others - who get hurt or can't repeat their strong 2017s - will be dropped or benched.
Moreover, if replacement value were to persist in being this high long term, owners would jettison starters much more quickly for the immense bounty available on waivers, and undrafted players would occupy a larger portion of the league's total at-bats. But if that happened, the owners would necessarily be deeper into the undrafted pool, i.e., using lesser players and lowering the replacement value to begin with. What replacement value wound up being at season's end isn't necessarily reflective of the assumptions under which owners were operating.
Incidentally, a Twitter follower ran the numbers for his 12-team NFBC league in 2016 and found replacement value to be significantly lower than last year (keep in mind he normalized for 500 ABs):
@Chris_Liss ran the replacement level analysis like you did for my league for past 2 yrs. 2017's waiver hitters much better than 2016. Jump in OBP, SLG, RBI pic.twitter.com/HvhYFGG9jU
— Matt Olson (@Mols10) February 23, 2018
Bottom line, it's much easier to model projected reality with projected stats than it is to take actual results and mix them with 50th-percentile projections. I still think this exercise was instructive, if not as specifically useful as I had hoped. The takeaway for me is the vast amount of free-hitting loot in the 12-team format, and the lack of decent freely-available pitching, assuming baseball's new HR-friendly era persists.
One quick way to illustrate is to look at last year's HR leaders: Stanton, Aaron Judge, Khris Davis, Joey Gallo, Nelson Cruz, Cody Bellinger, Justin Smoak, Mike Moustakas, Edwin Encarnacion and Logan Morrison. Only four of the 10 were even on the preseason radar. Contrast that with the 13 pitchers to register 200 Ks: Chris Sale, Max Scherzer, Corey Kluber, Chris Archer, Jacob deGrom, Luis Severino, Carlos Carrasco, Robbie Ray, Carlos Martinez, Zack Greinke, Jeff Samardzija, Stephen Strasburg and Clayton Kershaw. Only Severino was a big surprise, while Greinke and Samardzija were mild ones.
Accordingly, I'll want to get at least two aces early in most formats before backing off the shakier pitching in favor of hitters with upside for their draft slots. If the volatile hitters cash in, great. If they don't, I can feel comfortable knowing there's backup on the waiver wire.