Last week, I ran the Steamer projections through my dollar-value formula and generated hitting and pitching prices, respectively. Today, with the benefit of some early NFBC auction prices (the average of three auctions), we can compare the algorithmically-generated values with the market-based ones.

Before we get to the table, a couple caveats are in order: (1) My values are flawed in that they don't reflect the market's perception of closers, i.e., I have them priced much lower than they'd go in any auction or draft. There are a few reasons for this, but one is my formula doesn't take category balance into consideration. For example, if a player were projected for 5,000 stolen bases and zero homers, the formula would tell you to max bid on him because he's so valuable. But you'd win steals by a uselessly huge margin and finish last everywhere else if you did that. As such, it probably overcounts raw production and undercounts the contributions of 30 players who are uniquely situated to give you a particular category you cannot acquire anywhere else; (2) Because I have closers lower, my starting pitchers will occupy a greater portion of the overall pitching budget and therefore be priced slightly higher; (3) My formula is based on Steamer's static projections, whereas the NFBC players are able to price in other variables, e.g., a pitcher projected for 150 innings due to missed starts will have other pitchers occupying his slot when he's out; and (4) the distribution of Steamer's necessarily conservative projected stats will not match the more extreme year-end, real-life production of actual players. While some NFBC buyers are working off projections, others are probably paying for prospective breakouts and avoiding potential collapses to mirror real-life outcomes.

For those reasons, the dollar values aren't a fair 1:1 comparison, but the disparities among starting pitchers between Steamer and NFBC auction buyers are still instructive, for they show either (1) exceptions to the algorithmic rules; or (2) the market's blind spots.

Below I've sorted the players by difference between Steamer prices (based on my formula) and NFBC ones:

Sonny Gray is the biggest outlier at -$16.1 as a starting pitcher. You can see Gray is predicted for significant regression in ERA (3.67) and WHIP (1.29.) Why would that be? Because Gray isn't a big strikeout pitcher, and he's a ground baller. With more balls in play, you give up more hits, and that's especially so when those balls are predominantly on the ground. Throw in improving but hardly masterful control, and Gray profiles as a middle-of-the-rotation starter. Put differently, Steamer is unconvinced Gray's two seasons of sub-.285 BABIP are sustainable from a pitcher that makes a living on balls-in-play on the ground. If his BABIP regresses to .310 or whatever we'd expect from a GB pitcher, Steamer is likely to be correct.

The NFBC owners (through three auctions) disagree. Gray has never had a BABIP above .293 in the majors, an ERA above 3.08 or a WHIP above 1.19. He's thrown nearly 500 innings, and instead of regressing in 2015, he got better.

So who's right? Steamer's probably right generally, but the NFBC owners are betting there's something different about Gray in particular. What might that be besides the unlikely results?

For starters, according to Fangraphs, Gray yielded the fourth lowest percentage of line drives. While ground balls more often result in hits than fly balls, line drives are by far the most damaging form of contact to one's BABIP. Moreover, Gray also had the 11th lowest percentage of hard-hit balls in the league. (He was only 45th in soft contact percentage and third overall in medium contact.) The question is whether Gray's ability to avoid hard contact (rather than generate weak contact) is a repeatable skill. If it is, Gray will likely prove to be an exception, and the NFBC auction buyers will get their money's worth. If not, Steamer's prediction is likely to carry the day. Of course, if Gray loses velocity, learns a new pitch or sees substantial changes in the defense behind him, we're looking a different situation altogether.

Steamer is also lukewarm on Gerrit Cole, giving him a 3.17 ERA and 1.16 WHIP. We could go through the indicators to figure out why that might be and then the counterargument, but in the interest of economy, I'll leave that to you. The point is he's another player the algorithm does not like as much as the market. Other laggers: Cole Hamels, Dallas Keuchel and Adam Wainwright.

On the flip side, Steamer thinks Michael Pineda, Yu Darvish, Masahiro Tanaka and Chris Sale were underpriced. Pineda had a BABIP of .343 and wound up with a 4.37 ERA and 1.23 WHIP despite a 156:21 K:BB ratio in 160 innings. And he gave up 21 homers despite a 1.86:1 GB/FB ratio. For 2016, Steamer projects a 3.38 ERA and 1.12 WHIP for him, more in line with his peripherals than his results. Health concerns, diminished velocity from his prospect days, a tough division and home park are probably why the NFBC players were more cautious.