Forecasting 101: How to Project Offenses Knowing You Could Be Wrong
Imagine at the end of a long hike you were to finally get to the top and look out on a distant mountain range. The scenery looks breath-taking, almost as if the mountains were stacked on top of each other. It’s almost as if you could reach out and touch them…
There’s a point at Look Out Mountain on the North Georgia/Tennessee border I’ve been to multiple times where you can see seven states. It almost as if you could instantly jump to another…
The problem is there is a whole mountain range in between you. You check a map and realize that traversing the path is so much longer than you realized. In fact, the distance seems to be exponentially longer as you go with new challenges and obstacles you didn’t see up at the top. Walking down the path you quit leaving distraught.
This is forecasting.
An opportunity to foresee and project while recognizing the number of variables at play. For fantasy football, this is the name of the game and yet we often fail to develop the right language and perspectives to become not only more accurate but sharper in the way we pronounce outlooks.
Forecasting: You See… What Had Happened Was…
Perhaps you need a definition of forecasting that is a bit more nuanced than simply “predicting the future”. Forecasting can be misconstrued if we don’t first identify what it isn’t:
What It Isn’t
|What It Is|
|Finding One Projection||
Finding a Range of Outcomes
Assuming Best Case Scenario
|Exploring All the Scenarios|
|Knowing the Future||
Removing the Blinders
Convinced You Have it Right
|Knowing You Could Be Right|
|Pretending You Know||
Admitting What You Don’t Know
Before you think this is just too philosophical for you to move any further, let’s drill down on a couple of these points before actually analyzing a team.
We need to remove blinders and figure out what is noise.
The easiest place to figure this out is Twitter. And I don’t mean that in a kind way to Twitter. There an overemphasis on sharing opinions that any statistic was thrown your way become ammo or an argument.
There are a number of statistics we personally think are noise. Total fantasy points gives you an end-of-season number without the context of the roller coaster. You can see Tyler Lockett finished at WR9 last year but if you tried to start him every week, you were likely disappointed.
Catch rate is another stat that Ive recently been frustrated by on Twitter. We need these for projections-sake but, for predictions-sake, it honestly doesn’t tell us a whole lot about player’s future status. This statistic is not as “sticky” as others with some year-to-year fluctuations and not nearly correlative at all to fantasy production compared to the other counting stats over the last five years.
For instance, Amari Cooper has basically yo-yoed through his first four seasons seeing catch rates of 55.4%, 62.6%, 50%, sky-rocketing to 70.1% when he came to Dallas, and then back down to 66.4% in 2019. What did he do last year? Back up to 70.8 percent. As I mentioned in the Path to WR1 Season Primer for 2021, we use catch rate to convert targets to receptions but don’t use this percentage to project or distinguish who a WR1 could be by any means.
Yards After Contact is another one that is fun to track but it doesn’t give you any of the context needed. If someone like Derrick Henry or Nick Chubb breaks a long run, you can see why they rank so high throwing off’ ‘would-be’ tacklers. But it also can be skewed easily. An 80-yard run untouched and a 1 yard run into the line both are in the same data set but the player’s average doesn’t tell us anything.
This isn’t to say these statistics are trash but rather getting the context and removing the noise from the latest “hot stat” on Twitter. Being quick to listen and slow to speak is hard…
We are quick to make up our minds and too slow to change them.
This is true in almost anything in life. We tend to want to hear what we want to hear from the news sources, Twitter accounts, and friends that sound most like us. For players, we love what they could be and fall in love with potential or we close the door too early.
When it comes to player evaluation, I’ve strapped into the roller coaster that is D.J. Chark Jr. Follow this dynasty trajectory over the last few years:
Chark was drafted in the 2nd round of 2018 and I almost laughed at the pick by Jacksonville. He did nothing his rookie year and my confirmation bias had all but solved the case: this dude was a bust. I turned a blind eye (I mean most of us did he went undrafted in 2019) and he broke out in 2019. I was slow to believe in his ability and faded him as much as I could in DFS before finally coming around at the end of 2019. So what did I do in 2020? I jumped on board like everyone else here with the Fantasy Footballers and he fell flat again. I’ve changed my course to the tune of a fully laid out article: Five Reasons to Not Draft D.J. Chark Jr. in Any Format.
Why did I share that journey through Chark-ville? To show you every one of us can fall victim to the same course of action.
Learn how to not create hard-line, dogmatic statements.
In one of the best reads of the off-season for me personally, Superforecasting: The Art of Science & Prediction discusses (among a number of topics) the difference between hedgehogs and foxes. Hedgehogs tend to be “big idea” people that use similar criteria over and over again to apply to problems. Foxes, on the other hand, see issues through multiple lenses and tend to gather together multiple perspectives. The tendency is to create a dichotomy and separate the two approaches as one being better than the other. But this is actually a spectrum in itself and a chance to not make a dogmatic statement about forecasting.
Foxes aggregate perspectives which really is the beneficial approach. However, if there are big ideas that have carried you through your analysis over the years, factor that in as well. When people ask me if I’m an analytics “spreadsheet bro” or grinding the tape as a “film guy”, I usually respond with “Yes!”. You don’t have to pick and choose or create a false dichotomy in forecasting.
Over the years, I’ve given some extensive time (and spreadsheets) to the topic of vacated targets. In Vacated Targets & Predicting the Future in Fantasy Football, I looked at every team since 2014 and their market shares per position. A glaring trend emerged: teams with massive amounts of vacated targets tend to target the RB position more the following year. I emphasize that one word in the statement because this isn’t the case in every single circumstance.
While I’ve struck gold hitting on RBs like Leonard Fournette, J.D. McKissic, and Nyheim Hines as PPR threats, the Ravens annually are one of those teams that show up in the study. Their WRs recently was a revolving door of middling options (John Brown, Michael Crabtree(!), Willie Snead) leaving room in the offense for someone to emerge. However, Lamar Jackson routinely has not targeted the RB heavily at all. Chalk it up to his rushing prowess or whatever you want but my vacated targets premise is true… most of the time but not every time. I’ve carried a hedgehog mindset into forecasting while hopefully emerging with the tools of a fox to also aggregate other lines of thinking.
How Do We Project a Team?
There are two “levels” of analysis when discussing how to project a team: how does this team function and how do the players function within that team’s environment. The mistake is moving ahead and looking only at player evaluations and how we think players will finish.
We spent a lot of the beginning of the off-season discussing schedules to the point of looking at every single game on the new 18- week menu. Here is some of the content in case you missed it:
- Analyzing Vegas Win Totals & What We Missed
- Schedule Adjusted Forecasts
- Insights to the NFL Schedule Release
- AFC Win Totals
- NFC Win Totals
We care about the schedule and wins because this takes our analysis from the clouds to a concrete projection based on actual games. For instance, the 49ers play a “4th place schedule” this year thanks to finishing 4th in their division due to a massive amount of injuries in 2020. That cake schedule is ranked the easiest by Warren Sharp and gives us a starting place when we view their players. It’s not final and it’s not perfect but when you look in their division, their is a wide discrepancy in added opponents this year:
They still play the same standard NFC West matchups as out-of-division matchups against the AFC South and NFC North. However, those two games on the schedule are drastically different for San Francisco compared to Los Angeles who seem like they get the worst hand here against two “second place teams” from 2020: the Super Bowl champion Buccaneers and the Ravens.
Strength of schedule shouldn’t be a formula of adding last season’s wins but instead looking at Vegas’ projected win totals and the team implied win totals that are available. We use DraftKings Sportsbook because the lines of the games are easy to access. The 49ers are projected for 10.5 wins
However, beyond the schedule and what Vegas thinks, the play-calling and coaching tendencies are so vital to find what could happen on the football field. The Baltimore Ravens are a great example because of how extreme some of their tendencies are:
- Last two seasons: 32nd and 32nd in pass rate
- Highest run rate in league in neutral situations
- Pace: Dead last last year, 2nd slowest in 2019
- Ravens games totaled 4th fewest plays per game
This gives us a starting place for their tendencies BUT… what if Lamar were to throw another 50 times in 2021? How would that change the offense? Instead of locking them into being 32nd in pass attempts, would their efficiency as an offense still be elite?
It’s fun to take scenarios like that and apply them especially to teams that we usually say “I know what they are”. If you look back at Baltimore’s gamelogs from 2020, there were a lot of blowouts (sorry Bengals fans). What if the margin in their games is a little bit closer?
Ok now that you have some semblance of a starting point with the team, you can now move to how the players move within that system. I could go on and on about this aspect but honestly most of fantasy football content is about player evaluations. We (including me!) love crafting pieces with a blend of narrative and statistical analysis mixed in. This is what makes the off-season so fun!
But here are a couple of overarching principles to keep in mind as you do player evaluations:
1. The Past is Our Best Window to the Future… But It’s Limited
I’ve mentioned this in previous articles but last year’s stats are descriptive of what happened but not necessarily prescriptive of this year. I mentioned this in Best Ball Win Rates & What They Tell Us but the stats from last year are fun to talk about and throw around on social media. But we are playing with a whole new system and chaos called the NFL season… in 2021. Keep that in mind before we end up copy and pasting the past. Even elite players can bust (see DeAndre Hopkins in 2016).
2. Naming Our Priors
What is already there? What thoughts (both good and bad) are lurking underneath the surface in our player evaluations? Priors don’t only have to be viewed as a negative thing.
In the past, if an offense had three solid WRs, I used to fade them. I’d say there are too many mouths to feed… until I dove into that question: Do Offenses Have Too Many Mouths to Feed? Now, I’m able to say… maybe. But investing in an offense with three capable WRs usually means the offense is good. Crazy, right. And in Best Ball, if one of those WRs is injured or takes a step back, you might’ve invested in one of the other two wide receivers who had a deflated ADP because of competition but now can be a difference maker. The Dallas, Cowboys, Cincinnati Bengals, and Pittsburgh Steelers are perfect examples of those kind of 3WR priors this year.
3. Find a range of outcomes for each player.
The simplest way to do this is find a median projection and then use a percentage to dictate what the low (25-35%) and high-end outcomes (65-75%) could look like. However, this method can be flawed when you consider the median projection for a player is at best a guestimate in itself.
I’ll close with the player who the industry whiffed on the most in 2020: Stefon Diggs.
He was going at the beginning of the 6th round in redraft leagues after going to Buffalo in the off-season. The Bills were a supposedly “run-first” offense with a QB who only had been completing less than 58 percent of his passes. “Putrid!”, we cried.
This was a death sentence for Diggs’ fantasy value and we all collectively bought into this assumption. (Except for our own Kacey Kasem who destroyed our Writers League and many other after she pounded the table for him last year)
We loved Diggs’ talent and ability to get open but couldn’t foresee a scenario where Diggs and Josh Allen collided into fantasy goodness. But it was there all along. There was a range of outcomes (even if it was small) where Diggs went ham as the alpha in this offense. The Bills decidedly dropkicked everyone in the face of their forecasts and instead reinvented that offense.
What a difference a year makes… Buffalo had three total snaps of four or more WRs on the field in 2019. As mentioned in Tracking 11-Personnel & Other WR Usage form 2020, , they paced the league in those sets with 16 percent of their plays. OC Brian Daboll gets a ton of credit for reinventing Josh Allen as a high completion percentage passer but the production Buffalo received from their WR3 & 4 (Cole Beasley and Gabriel Davis) is probably just as important. In Pass-Catching Trends and What They Mean from 2020, I detailed how the growing trend of WR3s seeing 75+ targets for competent passing offenses will continue. The Bills changed it up.. imagine that.
Let Diggs and the rest of the Bills be a reminder to search for a range of outcomes when forecasting and knowing you aren’t looking for one point of final data. You are playing with a mountain range of obstacles to overcome and traverse and that journey you’re making… everyone else is too.