How We Made Fantasy Football Team Projections

One of the first steps in making good fantasy football predictions is having good projections for what each team will do in 2021.

May 12, 2023 - by Jason Lisk

At TeamRankings, you know that if we are going to do something, we are going to do it in an objective, data-driven way. No hot takes, and no guesses based solely on gut feelings or personal opinions. We are now bringing that same approach you have come to expect with our pool picks offerings to fantasy football rankings. And one of the first steps to developing objective fantasy rankings is to have solid team projections.

Why is this important?

If you don’t have good team projections, then your estimates on individual players could be way off. Even worse, if you aren’t balancing your fantasy projections for a particular team, you could be mis-pricing players and missing on values. Let’s say you think the quarterback will throw for 4,500 yards and 30 touchdowns. Well, if your predictions for all the receivers (including tight ends and receiving backs) add up to only 3,800 yards and 24 touchdowns, you are missing out on opportunities. After all, passing yards should exactly equal receiving yards. Completions should equal receptions.

Fantasy Football Team Projections Based on the Betting Market

We use betting point spread information and Vegas futures as part of our analysis with our pick’em and survivor picks. It makes sense to leverage betting market info for our fantasy football rankings as well.

So we designed a team stat projection model that uses the projected 2023 NFL win totals in the betting markets as a major factor. Teams entering 2023 should be projected to have similar stats to teams from the past who had similar expectations.

But teams also have different styles. The Los Angeles Chargers and Baltimore Ravens have a similar win total projection for 2023. But Baltimore is a team with a running quarterback while the Chargers pass at a higher volume with Justin Herbert at quarterback.

So while the expected win total is a big factor in setting the team projection, it’s not the only one. Things like points scored and allowed the previous season are considered. Other style factors, like passing stats and rushing stats, matter. Defensive stats are included. Quarterback age and whether he is new to the team, along with coach experience, are included. If you are curious to learn a little more about how we made the projections, and see all the factors used in helping to identify the most similar teams, that info is at the end of this post.

Let’s take the defending Super Bowl champion Kansas City Chiefs as an example. They are projected to win 11.5 games this season, the most in the NFL. After including all the other factors, the most similar historical teams, on which their projection is based, include several high-powered offenses with QBs of a similar age to Patrick Mahomes.

In fact, four of the five most similar are, well, the Chiefs of the last four years with Mahomes. Multiple Aaron Rodgers/Packers team from the previous decade show up, as do some of the recent Bills’ teams, and an Andrew Luck Colts team.

Fantasy Football Team Projections are Conservative, neither Best Case or Worst Case

Our team projections are based on weighted averages of how similar teams performed. Some teams will over-perform expectations, while other teams will have injuries, chemistry issues, or underperform for other reasons. We don’t want to project teams based on the most optimistic outlook, or pessimistic view.

Turning back to the Kansas City Chiefs example, we have them projected to average 27.6 points per game, score 53.8 offensive touchdowns, and throw 38.1 passing touchdowns. You might think that feels low, since Kansas City has scored over 28.0 points a game every year with Mahomes as the starter.

The similar teams used to develop the Chief’s projection had some teams that won a lot of games. But other seasons where teams were merely “good” on offense, including some where the starting quarterback missed games due to injury. Others had defensive collapses that impacted the overall performance (but led to more passing yards).

So our team projections for the Chiefs and all the other teams represent a middle-of-the-road prediction for that particular team. They do not represent either the best case or worst case scenario for teams. If you want to use our projections for teams to then mentally adjust teams that you are more or less optimistic about, you are free to do so.

Making Manual Adjustments to Fantasy Football Team Projections

While the fantasy football team projections from our similar teams model produced reasonable results most of the time, that’s not always true. Some teams simply don’t have a lot of truly comparable teams.

Do you know how many teams have rushed for over 3,000 yards and passed for fewer than 2,300, like Chicago did last season?

You have to go back to 1973 and the Buffalo Bills, the year O.J. Simpson famously ran for over 2,000 yards. And besides that, you have one season by the San Francisco 49ers in 1948, when they weren’t even a member of the NFL but were part of the All-American Football Conference.

The point here is that the Chicago Bears’ most similar teams aren’t going to be all that similar to them. Each might have been similar in some categories, but not overall. So there’s just lots of uncertainty about how a team like Chicago will do a year later, because we don’t have a whole lot of historical info to go on.

Further, some situations have unique changes, including massive changes to personnel, that require adjusting the automated projections with some specific knowledge from this year.

Teams with Quarterback Changes

Another big source of manual adjustments are for teams with significant changes in style at quarterback this offseason. We have several changes this year, with the biggest, stylistically, probably being in Indianapolis, where the team has drafted Anthony Richardson fourth overall, after cycling through several veteran QBs in recent years.

Our team similarity model did include inputs regarding quarterback age (both the previous season and the projected starter for the upcoming one) as well as whether the team had a different starting quarterback. But the fact that,say, a team is changing from a veteran pocket passer to a mobile quarterback who throws much less is not picked up in the model. So we did further research in order to make adjustments.

Adjusting the Overall Projections to Reflect Recent Seasons

Those were just some of the key teams where we determined that manual adjustments were necessary because the results of a similar teams model using last year’s stats were not sufficient. We made smaller adjustments in some other cases to specific stat categories based on players returning from injury or key personnel changes.

After all that, we also took a higher level view of the fantasy football team projections as a whole. Some stats were adjusted uniformly across all teams to better reflect recent NFL trends. Some rate stats have changed significantly in the last decade. For example, interception rates have gone down quite a bit compared to even 10 or 15 years ago. So the interception numbers projected for the teams were high across the board based on what we have seen in recent years. We lowered the interception expectation uniformly for all teams.

The final result is that our projections, when you look across all 32 teams, are in line with what we have seen in recent seasons. We have calibrated the overall numbers to produce realistic average numbers across the league for number of passing touchdowns, passing yards, and rushing touchdowns, along with all the other stats.

Details Of The Fantasy Football Team Projection Model

To close, for those who are curious, here is a little more detail about the model that we used to create the baseline projections. (We then manually adjusted these baseline projections as described above.)

We used a similarity-score-based model. What that means is that for each NFL team, we took their 2022 stats, 2023 season win totals, and some other factors, and compared them to every other team since 2003. We assigned each factor a specific weight. We then multiplied that weight by the difference between the current team and the historical team for that factor. Then we added all those values together to get the total similarity score for the historical team.

We then used those similarity scores to rank the historical teams from most to least similar. Each team’s baseline 2023 projection is a weighted average of the performance of the 25 most similar teams in the following season. For example, since the Arizona Cardinals were similar to the 2015 49ers, and the preseason win total for the 2016 49ers was 5.5 wins (in a 16-game schedule), the stats for the 2016 49ers (in their one year with Chip Kelly as head coach) contribute to the projection for the 2023 Cardinals.

Finally, here are the statistical categories we included in the similar teams model.

Team Stats

• Year (teams further back in time were considered less similar)
• Points For
• Points Allowed
• Point Margin
• Wins
• Preseason Win Total Previous Year
• Preseason Win Total Current Year

Offensive Stats

• Total Yards
• Gross Passing Yards
• Sack Yards Lost on Offense
• Rush Yards
• Rush Yards per Carry
• Pass to Rush Play Ratio
• Completion Percentage
• Pass TD to Rush TD Ratio
• Interceptions Thrown

Defensive/Special Teams Stats

• Total Yards Allowed
• Pass Yards Allowed
• Rush Yards Allowed
• Interceptions
• Fumble Recoveries on Defense
• Defensive Sacks
• Field Goals Made
• Field Goals Attempted

Team Characteristics

• Primary QB Age previous year
• Projected QB Age upcoming year
• QB Status (rookie, new veteran starter, returning starter, on roster/partial starter, main starter injured) previous year
• QB Status (rookie, new veteran starter, returning starter, on roster/partial starter, main starter returning from injury) for projected starter in upcoming year
• Coach Status (new or returning) previous season
• Coach Status (new or returning) upcoming season
• AV-Adjusted Offensive Age the previous season (new for 2021)
• Coach Age previous season and upcoming season
• Coach Years of Experience with Team, previous season and upcoming season

The inclusion of these Team Characteristic factors helped to further identify teams that qualify as more similar. Take the Denver Broncos with new head coach Sean Payton and a veteran in Russell Wilson at QB. By including some similarity for teams that had quarterbacks of a similar age and a new head coach, the list of similar teams “feels” more appropriate for assessing Broncos. All of the similar teams had new head coaches, with seven of the top eight having a new head coach above the age 0f 50, including teams coached by Mike Shanahan, Jon Gruden, and Peter Carroll.