Microsoft Word

Journal of Business Cases and Applications Volume 29

Statistical Analysis of 2020, Page 1

Statistical Analysis of 2020 NFL Mock Drafts

Cali Manning Davis

Troy University

Ed Pappanastos

Troy University

ABSTRACT

Even before a college football season has started, sports analysts are unveiling their

‘way-too-early’ predictions for the National Football League’s draft that is almost a year away.

As the college and professional football seasons progress and the date of the draft approaches,

analysts continue updating and publishing revised predictions. After both seasons have finished,

professional teams will interview and workout eligible college draft prospects to gather more

information. Sports analysts continue updating their predictions until the actual draft occurs. This

case study provides students with the opportunity to use descriptive statistics tools to compare

the accuracy of sports analysts’ prediction of the first round of the 2020 draft. The predictions of

12 different sports analysts are considered, with a total of 40 sets of first round predictions.

Keywords: Descriptive Statistics, Data Analysis, Forecast Accuracy, Correlation, Excel

journals. Please see the AABRI Copyright Policy at http://www.aabri.com/copyright.html

Journal of Business Cases and Applications Volume 29

Statistical Analysis of 2020, Page 2

INTRODUCTION

The National Football League (NFL consists of 32 different teams. Their season begins

in early September and ends in early February. At the conclusion of the season, each team looks

for ways to improve their team for the upcoming season regardless of their success in the

previous season. In some cases, that means firing the current head coach and coaching staff. In

other cases, it means releasing current players, picking up players who have been released from

other teams, or trading current players for players on other teams.

Another major method of improving a team is the NFL Draft, which is held at the end of

April. The NFL Draft is the means by which the professional football teams can select college

players they believe will improve their teams. There are seven rounds of the draft with 32

selections in each round. Using wins, losses, and tie-breaker rules, the 32 teams are ranked from

worst to best. The worst team from the previous season is given the first pick in the NFL Draft.

The second worst team is given the second pick in the NFL Draft, and so on. This draft order is

repeated for each of the seven rounds, assuming no teams trade draft positions. (The Rules of the

NFL Draft, n.d.

THE SCOUTING PROCESS

While the draft is held each year in April, many months and years go into scrutinizing

potential draftees. NFL teams employ staffs of scouts to conduct the evaluations of college

prospects. (Thelen, 2018 Typically, there is a director of college scouting who oversees a staff

of regional scouts. The scouting staff is charged with the responsibility of evaluating all aspects

of college prospects: physical/athletic skills, intellectual capacity, emotional stability, work

ethic, maturity, etc.

Scouts have a labor-intensive schedule that follows a yearly cycle. (What's the Life of an

NFL Scout Actually Like?, 2018) Much of their time during August is spent evaluating the

talent on their team and needs of their team. With the start of college football in September,

scouts find themselves watching lots of football film on players, traveling the country, visiting

college campuses, meeting with coaches, and watching players practice. As the college season

starts to end, NFL scouts travel to watch the best teams play against each other in college bowl

games in December and early January. After the season has concluded, several ‘all-star’ games

allow the best seniors the opportunity to have personal meetings with scouts and to display their

talents competing against other outstanding seniors. These games typically take place in mid to

late January.

The NFL hosts the NFL Draft Combine in late February. This combine invites a select

group of approximately 300 seniors to a central location for an opportunity to impress the scouts.

The multi-day event includes recording physical measurements of the athletes: height, weight,

hand, and arm size. Also included is a medical exam. Next, athletes’ performance is recorded for

several strength and agility activities, 40-yard dash, bench press, vertical jump, and several other

drills. The college athletes also participate in an on-field workout. (2020 Combine Tracker &

Results, 2020) In addition to the examination of physical traits and skills, players undergo

psychological testing and team interviews. (Schedule, n.d.)

After the NFL Draft Combine, many of the major college football programs host their

own ‘pro day.’ This is an opportunity for NFL coaches and scouts to visit players at their college

training sites. In addition, it is an opportunity for players not invited to the Combine to meet and

Journal of Business Cases and Applications Volume 29

Statistical Analysis of 2020, Page 3

workout for the scouts. Most of the pro day events take place in March. NFL teams also use

March and early April as an opportunity to host players at their team training facilities to have

personal meetings and individualized workouts.

DRAFT STRATEGY

The NFL Draft is an opportunity for NFL teams to improve their rosters. Scouting allows

the teams to learn more about prospective players and helps the teams formulate strategies for

selecting the players that offer the best opportunity to improve their team.

Regardless of a team’s draft position, there are at least three approaches to selecting

players. The most common approach teams use in selecting players is to identify deficiencies

their team has and to select players who have skills that that will eliminate deficiencies. While

not used very often, another approach is to simply select the best player available at the time of

their draft pick.

A third strategy that teams employ is trading draft picks. Obviously, teams with the

earliest draft positions have the opportunity to select the best college players. Therefore, early

draft picks are very valuable. Oftentimes, teams with worst draft positions will attempt to trade

with teams in better positions to select more desired players. For example, suppose that Team A

has the 25

pick in the first round and there is a player they consider special. They anticipate this

player going early in the draft. To ensure that Team A is able to select that player, they might

offer to trade their first round and second round picks to Team B, who has the 3

pick in the

first round. While Team B has given up the 3

pick, they have replaced it with two picks, and

Team A is now in a much better position to pick the desired player. Trades such as these are

more common in the first round because that is where the most valuable players are selected. As

a result of the final records for the 2019 season and previous draft trades, the ‘planned’ 2020

draft order is shown as indicated in Table 1 (Appendix. Because of previous trades, not all 32

teams currently have first round picks. In fact, the Dolphins have three of the 32 picks in the first

round.

THE UNIQUENESS OF THE 2020 NFL DRAFT

During a typical year, scouts have the numerous opportunities previously mentioned to

gather data on potential players. Those methods include:

• Reviewing game film

• Observing on-campus practices during the college season

• Watching all-star practices and games

• The NFL Draft Combine

• College pro days

• Player visits/workouts at team facilities.

Two of the most-valued opportunities are the two events after the NFL Draft Combine – college

pro days and player visits/workouts at team facilities. These two types of events, which occur

approximately during the 30 days prior to the draft, are important opportunities for teams to

really get to know players and develop relationships with them. During this period of

information gathering, teams often find themselves in a fluid situation regarding their

Journal of Business Cases and Applications Volume 29

Statistical Analysis of 2020, Page 4

planned draft picks. As new information is gathered, some players become more desired while

other players become less desired.

Because of the 2020 coronavirus pandemic, almost all events after the combine had to be

cancelled. (West, 2020) Teams have attempted to look for other means to gather data (Benson,

2020); however, nothing matches the value of the face-to-face meetings. Not having the

opportunity for the additional meetings with players, teams might be more inclined to remain

committed to the post-combine rankings of players, which might result in less trades among

teams.

MOCK DRAFTS

Not only is the NFL Draft important for the teams, it is a big business. In typical year,

one without a pandemic, the draft is held over three days at the home city of one of its 32 teams.

The 2017 draft was held in Philadelphia, the home city of the Philadelphia Eagles. It is estimated

that the event had almost an $95 million economic impact on the city of Philadelphia. The 2018

draft, hosted in Dallas, reported an economic impact to city of Dallas of $125 million. Last

year’s 2019 draft was a tremendous success in Nashville, with approximately 600,000 people

attending the three-day event and an estimated economic impact of $224 million. (Axson, 2019)

In addition to the event itself, creating mock drafts to predict the results of the actual NFL

draft has become cottage industry. Anyone with a computer and a web site can create their own

mock draft while sitting in their home. A Google search for ‘NFL Mock Draft 2020’ identifies

over 65 million results. Major, well-respected sports media organizations, such as CBS Sports,

ESPN, and Sports Illustrated, are among the first few pages of results. Mixed in with these

organizations is a multitude of smaller organizations.

Many of the web sites offer mock drafts for free, while other sites require a monthly fee

to access them. The CBS Sports web site is one of the many sites that are free. In addition to

displaying several advertisements, the CBS Sports web site provides the current mock drafts for

seven of their sports analysts. (NFL Mock Drafts 2020, 2020) Two of the most well-known

‘draft experts,’ Todd McShay and Mel Kiper, reveal their mock drafts as exclusive content on

ESPN’s premium service, ESPN+, that requires a monthly fee for access. (Ota, 2020) In addition

to tremendous amount of online mock draft content, entire television shows are usually devoted

to predicting draft results during the two-month period prior to the draft.

Sports analysts use a variety of resources for creating their mock drafts. The primary

resource they use is their own analysis of each teams’ weaknesses. Whereas teams that have a

young, talented quarterback are much less likely to draft a quarterback in the first round, teams

that lack a pass rush are much more inclined to draft defensive linemen. Combined with their

analysis of the teams’ weaknesses, is their personal analysis of college prospects. If an analyst

decides a team needs a quarterback, then the analyst uses his knowledge to identify the ‘best’

college quarterback for that team.

Another resource that influences mock drafts is the philosophy of various teams. Teams

that are more focused on offense may have certain tendencies regarding how they draft. For

example, a team that is pass-oriented might be more inclined to select receivers.

A major resource that influences mock drafts is the analyst’s relationship with teams.

Most analysts have personal relationships with scouting staffs across the NFL. Through these

personal relationships, analysts gather information that helps them gain insight into the team’s

perception of potential draftees and into the team’s draft plans. As the draft gets closer, these

Journal of Business Cases and Applications Volume 29

Statistical Analysis of 2020, Page 5

insights help the analysts revise their mock drafts. In the 90 days prior to the draft, an analyst

may release four to five different versions of their mock draft. Ideally, with releases closer to

the draft being more accurate.

CASE ASSIGNMENT

The data file NFL Mock Draft 200 Data.xlsx contains 40 different mock draft versions

from 11 different analysts for a total of 1,280 rows of data. There are at least three different

releases for each of the analysts, except for Mike Florio who prides himself on releasing ONLY

one mock draft. For each release, the following data is provided:

Date: the date the release was made

Analyst: the analyst making the release

Pick: the pick in the first round (1 to 32)

Team: the team predicted to be making the pick

Player: the player predicted to be picked

Position: the position of the predicted player

College: the college of the predicted player

Also included in the data file is the actual first round results.

Two options for a student case assignment include: (1 have the students identify

questions of interest and then perform the subsequent analysis necessary to answer the questions,

and (2 provide the students with a specific set of questions that need answering. Option 1 offers

the benefit of giving the students the opportunity to develop critical thinking skills. It requires

students to examine a set of data and identify the valuable information that it might provide.

From an instructional point of view, Option 2 is preferred because it creates a standard set of

work for the instructor to evaluate. Below are a set of questions that can be provided should

Option 2 be selected.

1. Is there a difference in prediction accuracy among the analysts?

2. Are predictions made closer to the draft more accurate?

3. Are some analysts ‘biased’ toward some colleges?

4. Were there any players who were consistently predicted to be first round picks but

were not picked?

5. Were there any players picked in the first round who were not predicted to go in the

round?

6. Are analysts more accurate at predicting the first 10 picks than predicting the last 10

picks?

The answers to the first six questions are provided below.

Journal of Business Cases and Applications Volume 29

Statistical Analysis of 2020, Page 6

ANALYSIS

Question 1. Is there a difference in prediction accuracy among the analysts?

The first step in comparing prediction accuracy is to measure accuracy. Three different

measures of accuracy are considered:

1. Team/Pick Accuracy: measured as the percentage of picks for which the correct team was

predicted.

2. Player/Pick Accuracy: measured as the percentage of picks for which the correct player was

predicted.

3. Player Accuracy: measured as the percentage of players, regardless of team and pick, that

were accurately predicted as being first round picks.

Because analysts are expected to gather more relevant draft information as the draft

approaches, it seems reasonable to compare their final mock drafts. The accuracy measures for

each analyst are shown as indicated in Table 2 (Appendix). It should be noted that by simply

predicting the teams in the ‘planned’ 2020 draft order would result in a Team/Pick Accuracy of

81% (26 of the 32 first round picks were made by the ‘planned’ team). While six of the 12 of the

analysts achieved a Team/Pick Accuracy of 81%, none of them exceeded it. A couple of analysts

were only slightly greater than 50% accurate.

There is a significant amount of variability with regard to Player/Pick Accuracy. Almost half of

the analysts accurately predicted over 20% of the players for each pick. However, a quarter of the

analysts accurately predicted slightly more than 10% of players. The first two picks were believed

to be fairly certain. In fact, 10 of the 12 analysts perfectly picked the first two picks. As such, a

Player/Pick Accuracy of slightly more than 10% means that the analyst was only able to pick an

additional two or three players in the correct picks.

For the last measure of accuracy, Player Accuracy, the analysts were extremely consistent.

Seven of the 12 analysts accurately predicted between 75% and 78% of the first-round players. All

the analysts predicted between 72% and 84% of the players in the first round. Clearly, there was

agreement in who the best college players were.

Two possible ways to assess the overall accuracy of the analysts is to compare accuracy

rankings of analysts and to compare the average of the analysts’ accuracy measures. The analyst

accuracy rankings, individual rankings for each measure and an average ranking, are shown as

indicated in Table 3 (Appendix). Where analysts were tied, each analyst was given the rank for

which they were tied. For example, if the top four analysts had Team/Pick Accuracy measures of

81%, 81%, 78%, and 78%, both analysts with 81% would be given ranks of 1 and both analysts

with a 78% would be given ranks of 3. There is clearly a correlation between the two measures as

both averages result in essentially the same order -- Will Brinson is the exception. Brinson is

penalized using Average Ranking because his rank for Team/Pick is 7 when there are six ties for

first place and his Team/Pick Accuracy is only marginally less than the top six analysts. It is

interesting to note that top four ranked analysts are the only ones to have two rankings of 1 or 2.

Similarly, the bottom three analysts are the only ones with at two rankings of 10 or more. The two

most-respected analysts, Todd McShay and Mel Kiper, both ranked in the middle of the group.

Instructor Note: In order to complete this analysis, the following was required in Excel: a

data table for actual first round results, a data table for all athletes predicted in any of the 40 mock

Journal of Business Cases and Applications Volume 29

Statistical Analysis of 2020, Page 7

drafts, =IF statements with embedded =vlookup(.,.,) or =lookup(.,.,) formulas. The worksheet

Prediction Accuracy by Analysts worksheet demonstrates the how these Excel tools are used. It can

be completely removed from the data file or modified for student use.

Question 2. Are predictions made closer to the draft more accurate?

As teams gather more data on players, the pool of potential players they are considering

drafting begins to decrease. Using their relationships with the scouting staffs, analysts get a clearer

picture of teams’ draft plans, including potential trades with other teams. Hence, it is reasonable to

assume that mock drafts would be more accurate as the actual draft date approaches.

In order to do the necessary analysis, it is helpful to create another variable to record the

number of days each mock draft is made before the actual draft. As the number of days decreases,

one would expect that all three accuracy measures would increase. The correlations between the

four different variables are shown as indicated in Table 4 (Appendix).

The correlation coefficients indicate that there are weak, positive relationships between

‘Days Before Draft’ and Team/Pick Accuracy and Player/Pick Accuracy. This is contrary to what

is expected in that the coefficients are both positive, suggesting that mock drafts closer to the

actual draft do a worse job in predicting the order of teams drafting and the order of players being

drafted.

However, there is a strong, negative correlation between ‘Days Before Draft’ and Player

Accuracy. So, while more recent drafts do a poor job of predicting the exact order of teams and

players, they are much better able to select the actual 32 players selected in the first round.

Instructor Note: The new variable, ‘Days Before Draft,’ can be found by subtracting the

date of the release of the mock draft from the actual draft date, April 23, 2020. Using Excel’s serial

representation of dates makes this calculation easy. The worksheet Prediction Accuracy vs. Days

worksheet demonstrates the how these Excel tools are used. It can be completely removed from the

data file or modified for student use.

Question 3. Are some analysts ‘biased’ toward some colleges?

Over the last decade, several college football teams have been extraordinarily successful.

Much of the success of these teams is due to the outstanding players on the teams. Given this, it is

reasonable to assume that players on these teams would be some of the most coveted players in the

NFL Draft, and some of the most popular players in the mock drafts. However, it is possible that

because of the teams’ success and the attention they have received, analysts are biased and tend to

rank their players higher than they deserved to be ranked.

One way to assess this is to look at the average number of players an analyst predicts for

each college. To simplify the analysis, only colleges who averaged more than one pick over the 40

mock drafts was considered. The results are shown as indicated in Table 5 (Appendix).

For almost all college teams, there is consistency among the analysts in the average number

of first round draft choices. However, there is significant variability surrounding Alabama and

LSU. This should be expected as these two college teams were expected to account for almost a

third of the first round draft selections. Alabama is the only college team the analysts forecasted

with more picks that were selected. On average, the analysts forecasted that 5.5 Alabama players

Journal of Business Cases and Applications Volume 29

Statistical Analysis of 2020, Page 8

would be selected in the first round, but only four were drafted. Three of the analysts predicted

Alabama would have at least six first round draft choices, an over-estimate of at least 50%.

Question 4. Were there any players who were consistently predicted to be first round picks

but were not picked?

The number of mock drafts made by each analyst consistently fell between 3-5 mock drafts

(minimum: Mike Florio with 1 mock draft, median=3 mock drafts, maximum: R.J. White with 5

mock drafts).

The total number of mock drafts in the data is 40. This information is important since the

meaning of “consistently” will vary by analysis. “Consistently,” for this question can be

determined using the notion that 75% of the picks included the player picked in the first round.

Therefore, the players that had at least 30 or more of the 40 mock drafts identifying them as being

picked in the first round would be the players expected to be first round picks. (30/40=0.75 or

75%)

The players consistently predicted to be first round picks are shown as indicated in Table 6

(Appendix) – organized by Mock Draft Appearances and the alphabetical order (last name). Also

included is a column showing if the player was picked or not (binary variable).

Two players appeared in 75% or more of the mock drafts but were not selected in the first

round.

• Yetur Gross-Matos (Round 2 to the Carolina Panthers)

• Josh Jones (Round 3 to the Arizona Cardinals)

Question 5. Were there any players picked in the first round who were not predicted to go in

the round?

There are 11 players that were picked in the first round who were not predicted to go in the

first round. These players are shown as indicated in Table 7 (Appendix) – organized by Actual

Pick (lowest to highest). Also included is a column showing the number of times the player

appeared in the mock drafts.

Question 6: Are analysts more accurate a predicting the first 10 picks than predicting the last

10 picks?

Two binary data field can be added to the data file to show Top Ten and Last Ten status.

• Top 10?: If player was drafted in picks 1-10 then Top 10?=yes, otherwise=no.

• Last 10?: If player was drafted in picks 23-32 then Last 10?=yes, otherwise=no.

Top Ten – The players picked in the top ten of the 2020 NFL Draft appeared 400 total times in

the mock draft predictions. The players appeared in the top ten of the mock draft predictions 315

times. This makes the accuracy rating of predicting the top ten players = 315/400 = 0.788 (π1) or

78.8%. The summary of top ten players and mock draft predictions are shown as indicated in Table 8

(Appendix). Players are sorted by draft pick.

Last Ten – The players picked in the last ten of the 2020 NFL Draft appeared 177 total times in

Journal of Business Cases and Applications Volume 29

Statistical Analysis of 2020, Page 9

the mock draft predictions. The players appeared in the last ten of the mock draft predictions 123

times. This makes the accuracy rating of predicting the last ten players = 123/177 = 0.695 (π

) or

69.5%. The summary of last ten players and mock draft predictions are shown as indicated in Table 9

(Appendix). Players are sorted by draft pick. A test of proportions can be used to test if the population

proportion of mock drafts that correctly predict the Top 10 is great than the population proportion of

mock drafts that correctly pick the Last 10. That is, the following test can be conducted.

Use a test of two proportions with:

: π

≤ π

: π

> π

Using a significance level of α = 0.05 (z-value = 2.3; p-value = 0.759), the null hypothesis

would not be rejected. By failing to reject the null hypothesis, we cannot conclude that analysts have a

greater ability to accurately predict the first 10 picks than the last 10 picks.

CONCLUSIONS

Not only are the NFL Mock Drafts a tremendous source of entertainment, but they also

provide a great opportunity for students to practice critical thinking and their data analytics skills.

One interesting finding from this case is with respect to the analysts’ accuracy. While analysts can

develop more accurate predictions as to who will be selected in the first round as the draft date

approaches, their predictions for team/player draft results do not improve over time.

Additional questions that can be considered include:

1. How consistent are individual analyst’s forecasts?

2. How much agreement exists between the different analysts?

3. Are some positions more likely to be drafted in the first round?

4. Are some positions more likely to be selected in the early part of the round and other

positions more likely to be selected during the later part of the round?

5. Are there more trades in the early part of the round than in the later?

6. How does Mike Florio’s ‘one mock draft’ compare to the final mock drafts of the other

analysts?

Journal of Business Cases and Applications Volume 29

Statistical Analysis of 2020, Page 10

REFERENCES

“2020 Combine Tracker & Results.” 2020 NFL Combine Results & Tracker | NFL.com, (n.d.) h

ttp://www.nfl.com/combine/tracker.

Axson, Scooby. “Draft Viewership Sets Records, NFL Says.” Sports Illustrated, 29 Apr. 2019,

h ttps://www.footballoutsiders.com/stat-analysis/2020/value-mock-drafts.

Belson, Ken. “Can't Scout Players in Person? The N.F.L. Turns to a Brooklyn Start-Up.” The

New York Times, The New York Times, 16 Apr. 2020,

h ttps://www.nytimes.com/2020/04/16/sports/football/nfl-draft-scouting-technology.html.

“NFL Mock Drafts 2020 - Football Draft Projections.” CBSSports.com,

h ttps://www.cbssports.com/nfl/draft/mock-draft/.

Ota, Kevin. “Exclusive NFL Draft Content from ESPN+: In-Depth, Expert Analysis and Insight

Available Only to ESPN+ Subscribers .” ESPN Press Room U.S., 23 Apr. 2020,

h ttps://espnpressroom.com/us/press-releases/2020/04/exclusive-nfl-draft-content-from-e

spn-in-depth-expert-analysis-and-insight-available-only-to-espn-subscribers/.

“Schedule.” NFL Combine, (n.d.) h ttp://www.nflcombine.net/players/schedule/.

“The Rules of the NFL Draft.” NFL Football Operations, (n.d.) h ttps://operations.nfl.com/the- p

layers/the-nfl-draft/the-rules-of-the-draft/.

Thelen, Paul. “What Exactly Does Each Member of an NFL Team's Front Office Do?” Bleacher

Report, Bleacher Report, 3 Oct. 2017, https://bleacherreport.com/articles/1666258-what- e

xactly-does-each-member-of-a-teams-front-office-do.

“What's the Life of an NFL Scout Actually Like?” Inside The Pylon, 25 June 2018,

h ttp://insidethepylon.com/nfl/long-form-editorial/profiles/2018/06/26/whats-the-life-of- a

n-nfl-scout-actually-like/.

West, Jenna. “NFL Cancels Pro Days Due to Coronavirus.” Sports Illustrated, 13 Mar. 2020,

h ttps://www.si.com/nfl/2020/03/13/nfl-cancels-pro-days-prospect-visits-coaches-travel- c

oronavirus.

Wyche, Steve. “How Coronavirus Pandemic Is Affecting HBCU-NFL Draft Pipeline.” NFL.com,

National Football League, 16 Apr. 2020,

w ww.nfl.com/news/story/0ap3000001109651/article/how-coronavirus-pandemic-is-

a ffecting-hbcunfl-draft-pipeline.

Journal of Business Cases and Applications Volume 29

Statistical Analysis of 2020, Page 11

APPENDIX

Journal of Business Cases and Applications Volume 29

Statistical Analysis of 2020, Page 12

Journal of Business Cases and Applications Volume 29

Statistical Analysis of 2020, Page 13

Journal of Business Cases and Applications Volume 29

Statistical Analysis of 2020, Page 14

Journal of Business Cases and Applications Volume 29

Statistical Analysis of 2020, Page 15