clock menu more-arrow no yes mobile

Filed under:

Another look at a Patrick Mahomes projection for the 2018 Chiefs

NFL: Kansas City Chiefs at Denver Broncos Isaiah J. Downing-USA TODAY Sports

For what it’s worth, this is not an attack on Mike Clay. I’ve been involved with data science long enough to understand how these things work, and it’s so hard to make predictions. I think most of Clay’s projection work is awesome, but I think he is very wrong about Mahomes and the Chiefs 2018 offense.

About a week ago NFL analyst Mike Clay published his 2018 projections for a number of NFL teams. You can see Clay’s prediction for the Chiefs here:

I’m sure every Chiefs fan jumped immediately to see how he projected Mahomes’ 2018 season, and I’m sure nearly every Chiefs fan was disappointed when they saw the following:

Mike Clay’s 2018 Mahomes Projection

Cmp Att Cmp% Yards TD INT
Cmp Att Cmp% Yards TD INT
336 536 63% 4041 21 15

I looked into the numbers even further and found Clay had the Chiefs offense finishing the season with only 32 touchdowns in 2018. That would have been good for bottom third of the NFL in 2017. I wasn’t a fan.

And I added a little more criticism for good measure...

Eventually someone brought Mike Clay into the conversation and we had a brief exchange. I was searching for context onto how he developed his projections.

Conversation with Mike

I’m just going to lay the whole conversation out here, and then I’ll have a brief reflection on the discussion.

(I’ll go into the “solve the unknown” bit a little later.)

I believe this last tweet is the most important part of the discussion in regards to figuring out how Mike Clay generated his numbers. It is my opinion the historical comps, league trend, scouting, and targets/scheme were manipulated in a way that hurt the Chiefs.

I cannot prove this beyond a shadow of a doubt, but I can explain what I believe given my experience and how I believe Mike Clay’s prediction is off.

Historical Comps

Over the years I’ve been involved with a number of sports science projects. One of the most difficult things to do while building projections for the NFL is to gather relevant data. I can also tell you relevant data is worth its weight in gold for building projections like these.

With Mahomes’ projection being so low, I believe Mike Clay used averages of all players over a certain period of time to generate historical comparisons. In this situation as an example, Johnny Manziel’s first two seasons in the NFL would be used in the averages built to help project Mahomes’ numbers.

The problem with this method is Manziel and Mahomes are not comparable players. Both quarterbacks have had wildly different situations to start their careers.

I am not certain this is what Mike Clay did; I did not get enough info from him. However, when so many teams are being analyzed, it is safer and quicker to use this method because it is a template which can be applied to every team. This is why I believe Manziel, and other dissimilar QB situations, were used to help build Mahomes’ projections. I would have likely done it this way myself if I had to evaluate 32 NFL teams and was hard up on time.

League Trends

Mike Clay mentions league trends as being very important to his projections. However, one such league trend I don’t believe he may be paying enough attention to is the relevance of young QBs in today’s NFL. Over the past 10 years, first and second year quarterbacks have been enjoying larger and larger production.

The following graphs show the progression of first and second year QB stats over the past 10 years. Please note a few things:

  • Only first/second year QBs who started in 15 or more of their games were included.
  • I did not include QBs who were thrust into a role due to an injury.
  • I also did not include Cleveland Browns QBs as they are an obvious outlier.

NFL QB stats have grown over the past 10 years, and young quarterbacks have grown as well - if not at a faster rate. With the NFL converting to more and more spread friendly systems, and coaches who are willing to incorporate college style plays into their game plans, young QBs are being put in a better position to succeed than ever and the data bears that out.

I am sure there are tons of other trends Mike Clay is using, but I believe the growth and prevalence of young signal callers in today’s NFL can’t be glossed over.


I’m not sure what Mike Clay’s scouts are seeing, but I can tell you Mahomes’ tape has been highly regarded by several scouts, and players.

I could keep adding to this list if I wanted. So I’m not sure if Mike Clay’s scouts disagree with the likes of Seth, Brian Baldinger, Voch Lombardi, etc. Of course it’s entirely possible the scouts who are high on Mahomes could be wrong, but from what I’m seeing the scouts who dislike Mahomes are the minority.

Targets and Scheme

During some of the conversations with Mike Clay, a bright Chiefs fan inquired as to why Sammy Watkins’ projected yards were so low, but his targets were so high. Mike Clay responded:

Here we have the ‘historical projection’ term again. Look, I love stats. I do. But sometimes the tape is infinitely more valuable than any number can offer. I believe this is a perfect example of this. Does anyone remember Seth’s film review of Sammy Watkins?

Seth went on to state:

In the four games I watched, Watkins repeatedly got open down the field and either:

1. Goff missed him entirely, or

2. Goff threw an uncatchable pass, negating the work Watkins had done.

Now, none of us can know if Mahomes will be incredibly accurate down the field. But the simple fact of the matter is if Watkins’ past target to reception rate is used, it has to be understood Goff was not doing a good job getting him the ball.

I believe Mike Clay is wrong to use a statistical projection in this manner because the tape shows Watkins was doing his job, while the QB was not. The target to reception rate is not entirely Watkins’ fault, and it’s not a stretch to believe Watkins may perform better with a QB who is known (in college) for having an accurate deep ball.

It’s also important to note, if Clay was using historical projections, he should have also taken into account the Chiefs QB over the past five seasons. Alex Smith was never known as a QB who would throw for a lot of touchdowns and yards, so wouldn’t it be safe to assume perhaps the statistical history of Tyreek Hill and Travis Kelce were also undervalued?

My Projection

What fun would this article be if I didn’t give a counter projection to Mike Clay’s projection? It would probably suck.

So since I’m a man of the people, I am going to make a projection myself, and here’s how:

  • I’m going to use the ‘knowns’ (Kelce/Hunt/Hill/Watkins/etc) instead of the ‘unknown’ (Mahomes)
  • I’m going to average relevant data.

I thought long and hard about how to approach my projection for Mahomes. I knew I would eventually need to make one, and what better time is the present?

Making projections is more of an art than a science, IMO. I’ve been fortunate enough to be able to predict the Chiefs record for the past three out of four years. It’s also important to know that sometimes projections can be muddled from paralysis by analysis.

Paralysis by analysis is when a subject is over analyzed. Often times too much irrelevant data is included and the truth is hidden behind too much noise.

For me, I believe this projection needs to be built in as simple and elegant a way as possible. (If you’re curious, my inspiration for thinking in this manner came from Claude Shannon’s approach to any problem.)

Solving for X

I believe Mike Clay’s projection focused too much on the unknown variable; Mahomes. With the Chiefs offensive cast, I just cannot see them having a TD output that lines up with the bottom third of the NFL.

Probably all of us have taken algebra at some point in times in our lives. One of the main goals of algebra is to solve for an unknown variable ‘x’. I can think of no better way to describe Mahomes; an unknown.

However, the Chiefs have several known players on the offensive side of the ball who have prior years worth of stats to pull from. So instead of generating projections for Hill/Kelce/Watkins/etc based on an unknown QB - it is much more intuitive to simply generate Mahomes’ projection using the known players.

3 Year Averages of Chiefs Offensive Starters

Player Rec Yards TDs
Player Rec Yards TDs
Kelce 1013 5.7
Hill 888 6.5
Watkins* 820 6.3
Hunt 455 3
Conley* 365 0.3
Harris 140 1
Rest of '17 475 4
TOTAL 4156 26.8

* = Two year average taken into account due to injury in one of three years.
Rest of ‘17 = Total of rest of Chiefs 2017 production. Albert Wilson’s numbers were excluded since Watkins is essentially his replacement.

So, using what we know about the 2018 Chiefs, it appears the starting QB would throw for 4,156 yards and 27 touchdowns. I bet you like that a little more than the other projection, amirite?

But I’m not done. I’m going to merge this line with what I would describe as relevant QB data.

Relevant Data

When finding relevant data, it’s really important to be able to see similarities and weed out the noise from what’s really important. When looking for relevant data to project Mahomes’ stat line, I looked for mixtures of the following:

  • A team with recent success
  • Had a year to sit and learn system
  • A solid head coach
  • A talented offense
  • Connections to Andy Reid
  • High praise for QB talent

I believe these are the core items surrounding Mahomes’ situation in Kansas City, and only QBs who were in similar circumstances could be used to project Mahomes’ outcome in his first full year as a starter.

I ended up with the following list:

  • 1992 Brett Favre
  • 2000 Daunte Culpepper
  • 2002 Michael Vick
  • 2002 Chad Pennington
  • 2004 Carson Palmer
  • 2006 Philip Rivers
  • 2007 Jason Campbell
  • 2007 Jay Cutler
  • 2010 Michael Vick
  • 2012 Colin Kaepernick

I’m sure several of you could make variations to this list of quarterbacks, but I tried to be as thorough as possible. I even took to the Twitter streets to see if I could gleam any more information from people (and there were many good responses.)

So let’s take a look at the raw numbers for each of these QBs. I’ll split the tables up so they’re more easy to view on your phones.

Similar QBs to Mahomes - Yards (raw data)

Year QB GS Yards
Year QB GS Yards
1992 Favre 13 3227
2000 Culpepper 16 3937
2002 Vick 15 2936
2002 Pennington 12 3120
2004 Palmer 13 2897
2006 Rivers 16 3388
2007 Campbell 13 2700
2007 Cutler 16 3497
2010 Vick 12 3018
2012 Kaepernick 7 1814

Similar QBs to Mahomes - TDs & INTs (raw data)

1992 Favre 13 18 13
2000 Culpepper 16 33 16
2002 Vick 15 16 8
2002 Pennington 12 22 6
2004 Palmer 13 18 18
2006 Rivers 16 22 9
2007 Campbell 13 12 11
2007 Cutler 16 20 14
2010 Vick 12 21 6
2012 Kaepernick 7 10 3

Uh oh, we have a problem though ... How can we project Mahomes’ numbers using quarterbacks from so many different eras? It’s not like 1992 QB stats were the same as 2017 QB stats.

We’ll take care of this by extrapolating the older QB stats to be relevant for the year 2017. This means Brett Favre’s 1992 numbers would look much more glamorous in 2017, than in 1992 - but Favre’s position compared to other QBs would be unchanged.

Essentially look at it as a way to make Favre’s 1992 season speak the same language as the 2017 season.

We also have the issue of QBs who did not play a full season. For the sake of this article we’ll also extrapolate each QB to have a full season of 16 games.

After doing a lot of work, we have the following adjusted data to use to try and project Mahomes’ 2018 season.

QBs Similar to Mahomes - Adjusted to Full Season and 2017 Production

Year QB Yards TDs INTs
Year QB Yards TDs INTs
1992 Favre 4749 28 12
2000 Culpepper 4270 37 13
2002 Vick 3311 18 7
2002 Pennington 4398 31 6
2004 Palmer 3799 22 18
2006 Rivers 3711 25 7
2007 Campbell 3479 15 11
2007 Cutler 3661 21 11
2010 Vick 4074 28 7
2012 Kaepernick 4022 22 6

Most importantly, the averages come out to be the following:

3,947 yards, 25 TDs, and 10 INTs

We’ll combine this with the projection from above, and come to a final projection.

Final Projection

Remember, with the ‘solve for x’ method we received the following numbers for Mahomes 2018 season:

4,156 yards, 27 TDs, 15 INTs

I’m also going to assume Mike Clay’s work regarding the INT numbers are better than the INT numbers I could come up with for this section (there are none), so that’s where the 15 INTs come from.

Once we average out the numbers from the two projections we have the following:

4,052 yards, 26 TDs, 13 INTs

That’s my 2018 projection for Mahomes, and my response to Mike Clay’s projection.

There is a huge part of me that believes Mahomes season can be comparable to Daunte Culpepper’s massive 2000 season, but there is too much other information to consider to make quite that jump.

So how do you feel? Which projection do you believe is more accurate?

Sign up for the newsletter Sign up for the Arrowhead Pride Daily Roundup newsletter!

A daily roundup of all your Kansas City Chiefs news from Arrowhead Pride