NFL: Houston Texans at Baltimore Ravens
Evan Habeeb / USA TODAY Sports

I've long struggled with how to accurately utilize projections when drafting my Fantasy football teams, because they are a snapshot of just one potential outcome for a player. That outcome might be the most likely one — that's what projections strive to articulate — but because it's a single possibility, a lot of context is lost. 

To explain this a little more, I want to get into some math terms I probably haven't used since high school. Do you remember the difference between mode, median and mean? I could never keep them straight. Median is a fairly popular one, sure, and we usually use the word average instead of mean, given that they are the same thing. But mode is one that doesn't seem to come up as much.

But that's basically what a projection is — the mode of a range of possible outcomes.

Let's take an example. Consider a player who has only 10 possible outcomes for his season, and we'll grade all of those from 1 to 10, with 10 being an amazing season, and 1 being not worth a roster spot for Fantasy purposes. Let's say that player's possible outcomes look like this:

2, 3, 4, 4, 5, 5, 5, 6, 6, 7

This is essentially what would be called a normally distributed range, which means the most likely outcome is right in the middle of a fairly equal number of outcomes on both sides — those represent upside and downside. The mean, or average, of these outcomes is 4.7, while the median is 5. The mode refers to the value that occurs most frequently, so it, too, would be 5. This is a player whose projection is a pretty good indication of the most likely outcome and there is some upside and downside in each direction. 

But not all players are like this. Consider these 10 outcomes, which might be thought of as a handcuff running back:

1, 2, 3, 3, 3, 4, 5, 8, 9, 10

In most outcomes, that player doesn't really have much value. But there are three outcomes where that player is thrust into a very valuable situation. This would be called a bimodal distribution, which just means there are two modes, or two humps if it were to be charted. 

For this distribution, the mean (again, the average) is 4.8, so right in line with the average for the first distribution we looked at. But this player is very different than the first one we looked at. And for this player, an outcome around 5 is actually pretty unlikely. 

Most projections for this player with this type of distribution are actually likely be to around the 3 value, which is the most likely outcome here. That's the mode. But when we look at a single projection that represents one possible outcome for a player, especially in a case like this, we essentially ignore all of those outcomes from 8 to 10. It doesn't particularly matter whether that player has a 30% probability of being amazing or a 5% probability; similarly, it doesn't matter whether that upside ranges from 8 to 10 or maybe just from 6 to 8. We don't actually quantify in a projection whether the backup would be an absolute star or just an RB2 in the case of a starter injury. 

But when you're drafting, the differences between these two players matter very much. In fact, I would argue these are the distinctions that make or break Fantasy football championships. These peripheral outcomes are where league-winners are made, and league-winning upside outcomes have a disproportionate impact on who wins your Fantasy football league. 

We, of course, discuss these things — we do talk about the upside scenarios and how good a player might be. But we don't have a good way to quantify it, and it's easy on Draft Day to fall back on projections. There are tools out there that try to quantify a range of outcomes, to be sure, but many are built on comps or other historical data, which may not reflect the naturally changing atmosphere of the league as a whole or the specific player's circumstances or career arc.  To my knowledge, even advanced machine learning algorithms aren't going to be great at predicting the NFL, which is always changing in unpredictable ways. 

So as we go through our position preview weeks here at CBS Fantasy, I'm going to try to be a little more clear about my expectations. We'll start here with the tight end position, and I'm going to give a few comparisons at key decision points in drafts. The charts you'll see are subjective — they are inexact visuals I've created where I'm projecting a range of potential outcomes. Your results will and should vary, maybe on my whole opinion of a player, or maybe just on the shape of their curve i.e. their projected range of upside or downside. We've also published tiered versions of my baseline tight end projections that you can review for context. 

The way to read these curves is the probability under the curve at any given point is the probability of that outcome occurring. The higher the curve, the more likely I think that positional finish is in PPR scoring. Let's skip Travis Kelce and George Kittle at the top, and start at the TE3 decision point. 

Mark Andrews or Zach Ertz?

I said these are inexact, so try to think of them more as comparisons. Andrews has Type 1 diabetes, which raises some concerns amid the ongoing Coronavirus pandemic. It's worth noting the NFL left Type 1 diabetes off their list of high-risk conditions, clarifying Type 2 as high risk, and in an article in the Baltimore Sun in June, Dr. Jay Skyler of the Diabetes Research Institute was quoted as saying, "I don't think there's any difference for a physically fit person with a well-controlled blood sugar to be concerned about any excess risk of COVID. I don't think he's at any greater risk than a tight end who doesn't have diabetes. There's no evidence to say that."

I mention this because Andrews has typically been a part-time player at the NFL level, and a highly-productive one at that, and his upside outcome for Fantasy football hinges a bit on whether his role might expand. We don't know how much of that limit is due to his condition, but I noted earlier this offseason that Hayden Hurst's departure could mean additional routes for Andrews, given how frequently both played in the slot relative to the Ravens' third highly-used tight end, Nick Boyle. With Andrews' elite efficiency, any increase in routes run — assuming his health dictates — would mean an uptick in targets and the potential for a Kittle-like impact. It would give him a solid chance to challenge for TE1 overall. 

Elite upside appears to be behind Ertz, on the other hand. The Eagles use two-tight end sets more than anyone, but in 2019 that came to only about 56% of total plays, per Sharp Football Stats. That left more than 40% of plays where they were only using one tight end, and with Dallas Goedert's emergence last year, it seems reasonable to expect Goedert could mix in more in 2020 in those one-tight end sets than we've seen in the past. That's a big deal for Ertz, who isn't the most efficient player with the volume he gets. He was one target shy of Kelce for most among tight ends in large part because he also ran the second-most routes at the position, just eight shy of Kelce. 

But Ertz probably also doesn't get enough credit for his floor. The Eagles have improved their receiver group, to be sure, but it's very hard to make a case for someone else to lead this team in targets. Ertz's target totals of the past two years — 156 and 135 — may now be upside outcomes for 2020, but he still looks to have a stable floor well into triple digits. I prefer Andrews of the two, but Ertz is a fine pick at his fifth-round ADP. 

Upside vs. stability

Darren Waller and Tyler Higbee are currently the fifth and sixth tight ends off the board, and they are good proxies for other names that come off the board shortly after them. Players like Evan Engram (TE7 ADP) and Rob Gronkowski (TE11) would plot out similarly to Higbee, with flatter curves and plenty of downside, though I'd argue both have higher healthy floors, and at least Engram's ceiling is comparable to Higbee's (I have Engram ranked as my TE5). 

Meanwhile, players like Hayden Hurst (TE8 ADP), Hunter Henry (TE9) and Jared Cook (TE10) have ranges more similarly-shaped to Waller's, though centered closer to their own ADPs. Cook's target ranges are relatively low from floor to ceiling, and his upside is largely defined by touchdown potential, making him a fade for me. 

Higbee is the ultimate boom-or-bust play in 2020 as a player who had 884 total receiving yards (15.2 per game) in 58 career games before Week 13, 2019, then closed the season doing something no tight end had ever done over five games. In that stretch, he had 522 yards (104 per game). But Gerald Everett only played four snaps in that stretch, something some believe indicates Higbee had clearly surpassed him, given Everett was activated for the final two games. It could also be that the Rams were rolling and Everett wasn't fully healthy.

There's also the schedule he faced, as four of the five games were played against the top-three TE matchups in terms of PPR points allowed. You read that right — Higbee played the single best matchup, Arizona, twice in that stretch, and also played Seattle and Dallas, who were second and third. There are plenty of reasons to be wary of the sustainability of Higbee's five-game stretch, but the upside is undeniable. I'm lower on that upside than some. 

Waller is essentially the discount Ertz, a player who due to team improvements is unlikely to see the same number of targets as last year. Remember, Waller stepped up as the clear No. 1 after Antonio Brown left the team in training camp, leaving the Raiders with a huge unexpected opening atop their pass-catching crew. But Waller also projects to lead the Raiders in targets again in a huge percentage of potential outcomes, and his role as an outlet option seems secure given Derek Carr's well-established dink-and-dunk tendencies. My baseline projection for Waller has him at 94 targets, a drop of more than 20 from 2019, but still puts him at my TE6 in PPR, and he seems very likely to finish in that range. 

Later options

Gesicki and Hockenson both represent favorite breakout candidates for me, as both were plus prospects who were drafted highly by their respective teams. Hockenson has a little more outright upside, both because he was a better prospect who went early in Round 1 and because he was more efficient as a rookie than Gesicki has been in either season to date. If Hockenson breaks out, it could certainly be a league-winning type Year 2 ascension. 

Gesicki, though, ran the fourth-most routes of any tight end last year, and also saw the fourth-most air yards. There's some risk of ruin — new offensive coordinator Chan Gailey doesn't have a strong track record with tight ends, and the coaching change might mean more time inline for Gesicki after he was overwhelmingly a slot option in 2019; there's also a concern he might just not be very good. However, the top-five volume he received last year certainly helps his median projection look a little better than some of the other later-round options. 

Then there's Hooper, whose new circumstances can't be seen as a positive for his Fantasy value. Hooper leaves an Atlanta team that led the NFL in pass attempts in 2019 for Cleveland, where he'll face plenty of competition for targets in what is likely to be a much less pass-friendly offense. His upside appears capped, and while a healthy season should put him in the back end of the TE1 group, he's a fallback plan on Draft Day, not a target for me. 

There are a ton more options later that I haven't plotted here, and some notable ones include Dallas Goedert (ADP TE15) and Blake Jarwin (TE19), whose floors I think are higher than one might think on first glance given their competition. Noah Fant (TE16), Jonnu Smith (TE17), and Jack Doyle (TE18) qualify as players for whom I don't quite see the targets.