Be it due to injuries, unpredictable play in 2019, or a lack of major league experience, certain players are tougher to project for 2020 than others. ESPN Fantasy's best minds when it comes to building projections, Derek Carty, Tristan H. Cockcroft and Todd Zola, got together to examine some of this year's most challenging players to forecast.
Cockcroft: If there's one player for 2020 who I think has a wide range of potential outcomes, it's Fernando Tatis Jr. Strangely enough, most projection systems have him playing around 145 games and amassing 630 plate appearances, despite his managing 86 and 380 in those categories between the majors and minors in 2019. I want to believe he'll stay healthy enough to play that much, and then there's the matter of his batting average -- .317, fueled by a completely unsustainable .410 BABIP. Tatis seems to be going at the end of the first or beginning of the second in NFBC-style leagues, but if the high end for batting average is .280, he'd have to go 25/25 at least to return on the investment. I like the guy a lot, but I think there's too much industry optimism.
Zola: The early NFBC drafters certainly are not following my rankings. I have Tatis ranked around 27. I agree with Tristan on health, he's young enough for it not to be an issue, but he has yet to display he can withstand the toils of a 162-game season. With respect to his BABIP, I broke it down by batted-ball components and he was fortunate on line drives and fly balls. Regressing those toward expected based on the component exit velocity landed at .340, still higher than league average, but his power/speed profile supports that level. However, unless he improves contact, Tatis' average should fall between .260 and .270, which is solid, but not first-round material. Keston Hiura is going a couple rounds later but is in a similar BABIP boat as Tatis, coming off a .402 campaign. Hiura should also maintain an elevated mark with a batting average in the neighborhood of Tatis, though Hiura doesn't have the raw speed of the Padres shortstop.
Carty: The industry is optimistic about a young, much-hyped player coming off a great performance in a short sample? You don't say, Tristan! This is exactly the type of player people want to buy into, and so they'll do whatever mental gymnastics they have to in order to justify it, even if literally zero projection systems support his ADP. Sure, he's a massively talented young player with loads of upside, but we have to remember that this is a player who had numbers bordering on mediocre in the minors, including below-average plate discipline that didn't improve in the majors, where the positive things he did do were done over a small sample. And even in that sample, he was more or less the luckiest hitter in baseball according to his expected Statcast numbers. He has plenty of value, with skills that are perfectly suited to fantasy, but he's the kind of player whose price is always inflated and whom I'm always out on.
Cockcroft: How about injured players, such as Yankees outfielders Aaron Judge and Giancarlo Stanton? Different players at different ages and career stages, and therefore different levels of injury risk, but tough to project nevertheless. From a projections standpoint, I tend to forecast exactly what I think the most likely outcome for the player, including guessing -- since that's all it is, really, a guess -- his playing time. So for a player like Judge, I'd probably project 115 games and 105 for Stanton (assuming a full 162-game schedule, in either case), apply the replacement-level player who fills the remaining open weeks, and run my values from there. But that's my approach, and it's going to keep either out of my top 100.
Carty: When I project players, I focus almost all of my attention on projecting their rate stats. My projection system, THE BAT, is housed at FanGraphs for season-long purposes, and FanGraphs has a team of guys who evaluate every team's playing time, depth chart and injury situations and projects playing time accordingly. The preseason is a busy time for me, and I feel like my time is better spent elsewhere, such as trying to improve the rate side of the projection, rather than trying to invent the playing-time wheel. So I simply trust the guys who are really digging into these things and scale THE BAT's rate production around that playing time.
Zola: As my colleagues suggest, projecting playing time for injury-prone players is a necessary evil. When I'm drafting, it's mostly a feel thing. You look at the remaining inventory, you look at your roster and decide, "Yeah, it's worth the risk." The problem is part of my job is setting a rank or expected earnings as a guide so I make my best guess and do my best to explain to the readers valuation is flawed. Tristan talked about it, you need to account for the production of the substitutes, adding a little to what you'll pay or jumping the player up the rankings.
Cockcroft: How are you both handling Astros players coming off the cheating scandal, and since there might be additional news related to other teams, are those adjustments being applied elsewhere? I'm not adjusting my projections much, but have done so by a couple of spots in the rankings for select players, Yordan Alvarez being one example in particular. His limited amount of big league experience concerns me from a league-adjustment standpoint.
Carty: I'd love to have a perfect answer here, but there just isn't one. You look at a player such as Marwin Gonzalez and how his 2017 season sticks out, and you have to think an adjustment is in order. But then you look at the numbers on the whole, and at studies people like Rob Arthur has done, and it shows that the net impact is fairly small or even nonexistent, that the times the Astros got the signs wrong were so detrimental as to wipe out the positive effects of the times the signals were accurate. Evidence is so incomplete and so contradictory we would really just be guessing what to do. And we don't know if any other teams were doing similar stuff and didn't get caught. Then there's the narrative issue to consider, if being hounded by this stuff on the road all year will take a toll and impact performance separate from whatever benefits might have been gained in the past. And then how do you handle a player like Jose Altuve, whom everyone is mad at and thinks was heavily involved in this but for whom the evidence -- at least what evidence we have -- shows maybe he wasn't actually cheating? Oh, and then what about potentially added injury potential if these guys are getting thrown at all year? It's a problem with a mess of things to consider and without a clear solution for any of them, at least without taking massive amounts of time to try to properly quantify that, can't be perfectly quantified no matter and may be not even have a large impact anyway.
Zola: The Astros are going to be so intriguing to watch. In a recent roundtable where we asked the Tout Wars participants this same question, Ron Shandler said it best. This isn't a valuation question, but rather a matter of how much risk you're willing to absorb. All the factors Derek and Tristan mentioned are impossible to quantify, but they add risk. I will say I feel the talk of perpetual hit batsmen is overblown, but I do suspect Dusty Baker is going to give the position players more days off than they're used to. I'm nicking the skills ever so slightly while lowering playing time expectations 2-3 games. At the end of the day, the rank/projected earnings isn't as important as team construction. If you're looking for batting average help with some power and speed from a second baseman, Altuve's raw rank is moot; he's the pick. As an aside, other than the obvious reasons, it's a shame this is a thing. Statcast data and how it's misleading in Minute Maid Park with the Crawford Boxes is such an intriguing topic, drowned out by the banging of trash cans.
Cockcroft: By the way, Derek, THE BAT has Madison Bumgarner projected for a 4.59 ERA and a negative dollar value in NFBC-style leagues for 2020, the lowest valuation of any of the major projection systems out there. So, Derek, why so low on Bumgarner?
Carty: Because he's not actually good anymore, and this is a pitcher whose underlying talent was always overrated to begin with. He played his entire career in one of the absolute best pitchers' parks in all of baseball. He threw to Buster Posey for years while Posey was one of the best pitch-framers in the game. He played with mostly positive defenses behind him. And now his skills have declined to the point where he's merely an above-average pitcher, and he's changing teams.
Despite league-wide strikeouts going up, since his offseason 2017 injury, Bumgarner's strikeouts have been way down. He went from four straight seasons of a context-adjusted strikeout rate 23%-33% above league average to only 8% above in 2017, 5% below in 2018 and 4% above in 2019. In other words, his strikeouts have basically been average. His surface numbers have looked good because of a deflated BABIP and inflated left-on-base percentage, but peripherally he has deserved an ERA over 4.00 in each of the past three seasons. Now he leaves the great context of the Giants and goes to still-good-but-definitely-significantly-worse context in Arizona. Between the declining skills and worse context, Bumgarner is one of the absolute worst picks you can make in your fantasy drafts this year.
Zola: I don't think Bumgarner is one of the absolute worst picks, at least in the right context. Another flaw with valuation is all of a pitcher's starts are dumped into our little black valuation boxes. While Bumgarner obviously isn't close to what he was, he should be a viable spot starter for most home games and some road affairs (Giants, maybe Padres). If the discount on Bumgarner drops him to at least an SP5, I'll take that and use him judiciously.
Cockcroft: What is your approach to projections in the new ballpark in Texas, as well in Miami and San Francisco, where the Marlins and Giants moved the fences in this season? I examined this in more detail in my annual Park Factors column recently but would love to get either of your thoughts.
Carty: THE BAT accounts for physical attributes of ballparks and has a method for estimating park effects using just these attributes. In the case of Marlins and Oracle Parks, I look at the previous fences, the new fences and the difference between them. From there, it's easy to calculate what the difference should be based on the new dimensions. In both cases, it turns out not to be too extreme. They're both still top-3 pitchers' parks (although Citi Field now projects as more pitcher-friendly than both).
In the case of Texas, where it's an entirely new park, I can still do this exercise, but error bars are much wider. As we saw with the previous Globe Life, there are things beyond just physical characteristics that can make a park play a certain way. It was an average park physically but a top-3 hitters' park anyway -- and that's before accounting for the hot weather. The new Globe Life projects as a top-10 hitters' park physically (more favorable than the old park), but it likely won't have the jet stream and whatever other effects were going on in the old park (although, things such as ball-mudding, groundskeeping, etc., could carry over, so you never know). It will also have a roof, which based on rough estimates of closure thresholds and historical weather, I estimate to be closed roughly 70% of the time during a normal April-September season, leading to average temperatures for the season less than five degrees hotter than league average. In total, I'm expecting it to play hitter-friendly, but nowhere near to the degree the old park did.
Zola: I'm not ashamed to admit Derek's THE BAT does a better job quantifying changes in park effects than anything I do. I'm also not afraid to say I consider his work when deciding what to do with my park factors. As was shown with the humidor, science is a great tool in this area and while it seems like a lifetime ago, I have an advanced degree in science.
I slightly increased the home run index for Oracle Park in San Francisco and Marlins Park, which should be called The Aquarium. The change benefited lefty swingers more, but I didn't ignore right-handed batters since the home run index for righties also accounts for opposite-field power. Maybe this is a mistake, but I didn't change the run index to the same extent; I only tweaked it a little, which could get lost in round off for some pitchers. The reason is that homers and runs don't always correlate. Some parks give up homers but suppress runs (Yankee Stadium, Guaranteed Rate Park) while others are favorable for runs but hinder homers (Fenway Park, Wrigley Field). The dichotomy results from what generally happens to the fly balls staying in the yard. With the fences closer in San Francisco and Miami, it might enable the right fielder to play shallower, thereby getting to more softly hit fly balls.
While I applied the change in park factor globally, I also dug in on a few batters from the Marlins and Giants based on spray charts. Brian Anderson has solid opposite-field power and could benefit from the shorter fences in South Beach, while many of Brandon Belt's lofted balls are directed toward where the fences are closer in San Francisco. I didn't further adjust either projection, but I did make a note of it and used it to break a tie if they were one of the choices I was considering in a draft.
Science can also be applied to the new Globe Life Field. Derek's manner of handling dimensions is more accurate, but I see the slight changes in dimensions to be a wash. However, the roof being closed lowers the temperature and reduces humidity. There are several online studies -- which are no doubt sourced by THE BAT -- explaining how the flight of a fly ball is influenced by atmospheric conditions. I got out my old Texas Instruments TI-40 calculator and came up with the adjustment which is in sync with Derek's findings. The factors may not be right, but I'm confident the new park will play better for pitchers than the old one. There were some seasons Arlington was better for homers than Coors Field. My sense is not many are making an adjustment to Texas pitchers, so unless the air conditioning is blowing out, Mike Minor, Lance Lynn, Corey Kluber and even Jordan Lyles and Kyle Gibson can be drafted at a discount.
Cockcroft: I tend to find rookies the toughest to project, because while I'd like to rely entirely upon MLEs (Minor League Equivalencies), they're often unreliable and with the change in the baseball last season, I'm less apt to trust them. How are you handling players like Luis Robert, Nick Madrigal and Nate Pearson, none of whom has any major league experience?
DC: Projections for minor leaguers will always have wider error bars than major leaguers, because there's an imperfect (but necessary) translation process. We have to take the minor league stats, run some math, and say, "This is what these stats would have looked like if he played against tougher competition in the major leagues." We also don't know if certain players are working on certain aspects of their game in the minors at the expense of their stats, but which actually make them better players when they go full-out in the majors.
That said, MLEs absolutely add to our accuracy and are a sort of necessary evil when it comes to projecting players. Most fantasy players just buy into the hype of rookies each year and are willing to spend a higher pick than any projection system thinks is smart because they want to hit the upside and find next hot new thing. This is usually a terrible strategy. I am completely out on Luis Robert this year, whose ADP is just stupid. If you take Robert at his ADP, you're probably really bad at fantasy baseball and will lose this year. Maybe he turns into the next Acuna or Tatis, but the odds are always against that, and I think they're more against it for Robert.
Not only do elite prospects like Acuna and Tatis fail plenty of times (see: Byron Buxton, Yoan Moncada, Dansby Swanson, and on and on and on), but most scouts don't actually even put Robert in that class anyway. He just happens to be the best we have this year. But he's a guy who hit zero (yes, zero) home runs in 2018, and while he showed good power in 2019, he still has weak plate discipline and weak MLEs. Hard, hard pass.
Zola: I don't have much to add to what Tristan and Derek said in terms of MLEs. They are necessary evils like Derek said, with some survivor bias contributing to the unreliability Tristan mentioned. I would, however, like to point out something about ADP (average draft position). In general, I agree with Derek's point about Robert and the overly aggressive drafting. That said, the primary ADP early in the drafting season emanates from the National Fantasy Baseball Championship and their draft and hold leagues. This contest crowns an overall champion where the winner bests a field of more than 1,000, so there's a train of thought suggesting upside wins. Again, not saying I agree with this principle (I don't), but it adds a little perspective to the aggressive drafting of high upside players.
Cockcroft: How heavily do you weight "luck" factors, such as BABIP, home run/fly ball percentage or strand rate in your projections? Has xwOBA (expected wOBA) taken on any significance in your process? Mitch Keller is a good example: I'm a big fan of his this season due to his elite strikeout rate in limited big league action in 2019, but when it came to the actual projection, I had a hard time estimating a 4.25 (or better) ERA. I'm ranking him mostly off of feel.
Carty: I don't necessarily view things as black-and-white, lucky-or-unlucky. That's the easiest way for the general public to understand them, and the way to do back-of-the-envelope math and to talk about these things generally, but in terms of full-on projections, we can be more precise. We can calculate the variance in things like BABIP. We can say that it takes the equivalent of eight years of data to stabilize, that batted ball profile and pitch mix gives us some extra information, and then regress the appropriate amount based on each player's sample size. That's a simplified explanation of how THE BAT handles it. It gives Keller a .318 BABIP, which is actually a bit lower than some other systems (ZiPS: .328, ATC: .325), but BABIP is closely linked to left-on-base percentage, and THE BAT is easily the most pessimistic there at 69.8% -- the only system under 70%. Most other systems are about league average, which to me doesn't make sense given that we all expect an inflated BABIP. THE BAT's 4.58 projected ERA is the worst, and I suspect this is one of the reasons why. It's also lowest on his strikeout rate, probably due to him being a fairly low-K pitcher in the minors and the way the MLE shakes out.
Zola: While good and bad luck is common in the fantasy baseball vernacular, I prefer to view it as events out of the players control. A batter can hit the ball, what transpires is most often out of his control, affecting both the hitter and hurler delivering the pitch. My job as prognosticator is first to determine the most likely skills level of players and then translate that into the most likely outcome. While it's myopic to say what happened last season is moot with respect to outcomes, we have the entire database of MLB history to call upon and help discern what should have occurred. Statcast is a recent tool, further refining probable expectations.
With respect to projections, I call in the statistical definition of regression. Unfortunately, regression has become synonymous with "play worse." The truth is, regression can go in either direction, helping or hurting the player. Progression is not positive regression; it's a made-up word meant to connote favorable regression. Sorry for the digression, but regression is one of the most misunderstood concepts in all of fantasy sports.
As discussed, I first project a player's skills and the most likely outcomes based on history, with a lot of the calculus thanks to Statcast. That said, even though we're getting better at distinguishing skill from happenstance, it's not a perfect science. As such, some of the player's outcomes could manifest from a latent skill. Ergo, my final projection regresses the player's historical outcomes toward expected. The extent of regression differs per metric and I reserve the right to subjectively massage it per player.
Cockcroft: How heavily do you weigh the aging process in projections? Nelson Cruz has been rock-solid for quite some time, but he'll turn 40 years old in July. Derek, you had the low-end batting average projection for Cruz (.268), so do you think it's wiser to downgrade him assuming he'll decline in 2020?
Carty: I've always loved Cruz, and I still think he's a great buy this year. THE BAT was actually one of the higher systems the past few years on him, if I recall correctly. I'm surprised it doesn't like him quite as much this year, and I suspect aging curves have something to do with it. Aging curves can break down at the extremes because our samples are so small and we're kind of extrapolating based on larger samples at lower ages. There's also been some evidence that players are aging differently in the post-steroid era, and that's something I haven't had a chance to examine yet. It's on my list, though, and it's possible that THE BAT is too pessimistic on Cruz as a result.
Zola: Derek touched on the main failing of aging curves; survival bias resulting in small samples with a lot of noise. Still, it's better to factor in aging than to ignore it. With players like Cruz, I may opt to override the global effect. It's one of those things where you're right to do it until you aren't. As someone paid to do projections, I have to come up with numbers. As a drafter, I don't care about the exact projection. Players such as Cruz always get an organic discount. It doesn't matter if I override a 5% aging decline, the unadjusted and adjusted projections are still more optimistic than the market, presenting a buying opportunity.