by Hannah Page

Preface

Many adjectives describe current Tennessee Titan, Taywan Taylor.

Hilltopper. Program record-setter. Play maker. Brandon Doughty's and Mike White’s go-to guy.

His name elicits memories of trick plays, deep catches, touchdowns, and 100+ yard games.

Everyone seemed to overlook and/or undervalue the two-star prospect from Pleasure Ridge Park HS in Louisville, Kentucky.

In a Billy Beane-esque fashion, then-WKU Football offensive coordinator and assistant head coach, Jeff Brohm, paid no mind to the rating system. After watching Taylor’s performance during a basketball game, Brohm encouraged then-coach Bobby Petrino to give Taylor a football scholarship.

Taylor’s story is the story of WKU Football’s (and most Group of Five programs') roster for years: lower-ranked recruits, as well as JUCO players, who were overlooked or perceived in a certain manner. Whether he recruited or inherited the player, Jeff Brohm and Company tapped into each individual player’s strengths and unequivocally elevated the program. By the end of 2016, the Brohm-era saw an overall record of 31-10. Brohm and his staff did the unimaginable with WKU given their short tenure in the FBS (and Conference USA); their departures were inevitable.

Despite success on the recruiting trail, WKU Football's offense has not been the same since 2016.

Some argue that the fall was bound to happen. Others point to the coaching change. A few fans believed, in time, the team would find their identity once again. None of those notions will be disputed in this piece.

Mike Sanford Jr., was dismissed on November 25, 2018 after two seasons, an overall record of 9-16, and a 2018 record of 3-9. After one day, Tyson Helton was named WKU Football's next head coach.

The prompt hiring of Helton - an ex-WKU Football offensive coordinator and quarterbacks coach under Brohm - was a clear and direct message to return to the "brand" of WKU Football that was competitive, exciting, explosive, and winning. Joining Helton, will be the 2017-2018 WKU Football defensive staff: a message of "if it ain't broke, don't fix it."

While I am not sure if Helton or Bryan Ellis (now-offensive coordinator and quarterbacks coach) are adherents of analytics, the 2014-2016 brand of WKU Football suggests an understanding and respect for decision making and trends that yield wins. During such time, WKU Football became an offensive juggernaut, a Top 25 team (2015), and back-to-back C-USA champions with a slew of overlooked, lower-ranked players.

There is an immense amount of data on any team. By way of analyzing such data, we can create a narrative that pieces together the past, present, and possible future of WKU Football.

Defining analytics

Sports analytics is defined as: "the management of structured historical data, the application of predictive analytic models that utilize that data, and the use of information systems to inform decision makers and enable them to help their organizations in gaining a competitive advantage on the field of play" (Alamar & Mehrotra, 2011).

Bill James - an American baseball writer, historian, and statistician - is regarded as one of the most influential figures in sabermetrics (baseball analytics). His research and works have gone on to influence analytics in all sports, including football (and this article). Bill Connelly of SBNation is another personage in the realm of analytics, specifically college football. His works and analyses take seemingly basic concepts (i.e., completion percentage) and elevates them to another dimension (i.e. marginal efficiency). His research and findings will be referenced throughout this piece.

However, baseball and football are two completely different sports. Where baseball is categorized as having sequential and distinct events, football is far more complex. First, there are 22 men moving at once: each man's performance tied to another (i.e., a quarterback's performance is tied to how well the offensive line protects him, the opponent's pass rush, if his receivers are creating separation, etc.). Second, to further complicate matters, it is nearly impossible to go back and identify every play-call. Third, which ties-in with play-calling, you have to identify the context of a situation (i.e., down, distance, time remaining, field position, margin, etc.). And finally, college football's sample size (number of games) is incredibly small compared to baseball.

Especially in college football, you will find that there is almost always an exception or an outlier to the norm.

Nonetheless, in American football, analytics help address issues such as evaluating talent, team and individual ability, per play success, and situational decision making.

Analytics. Advanced statistics. Data analysis. No matter how you say it, numbers and trends can help teams better understand their personnel from the individual all the way to the position group and cohesive unit level. Analytics answers the who, what, when, where, and why using a quantifiable approach.

Unfortunately, sports analytics [sometimes] receives a bad reputation from even the biggest names in the sports industry. Some, such as retired NBA player and NBA analyst Charles Barkley, are exclusively in favor of the "talent trumps all" mindset. (Golliver, 2015). The "talent" argument is a central theme in the book and film, Moneyball. Absolutely, any sport is more than just a stat sheet. There are talented but imperfect people playing an imperfect game. However, behind every season, every game, and every play, there is a stat. Statistics exist whether someone analyzes them or not.

Talent and analytics do, in fact, go hand-in-hand. Separately, they are flawed; together, the two form a balancing act of athleticism and strategy. Given the right players for a program, a solid system, and calculated decision-making, coaches can augment and put their team in the best possible position.

Article Intentions

Delving into statistics can be a controversial subject. However, it is a far more honest illustration than assigning an opinion to a complex situation without empirical evidence.

This is not a "How to Coach 101" nor "The team should do this" piece. This article does not cherry pick information or stats that push any sort of personal or fan base agenda. Coaches are the experts. I am simply expounding upon existing data.

Our eyes often deceive. You might see a failed fourth down conversion and call it "poor decision making." However, what you did not see were the expected points added and win probability added IF the team had converted. You can watch the entire game, but there is a lot you do not see.

This article is for those who are interested in the intricacies of WKU Football beyond the stats you read on your phone at halftime or immediately after the game. Because this article is intended for the masses, I attempt to explain complex numbers and analyses in layman's terms (I, myself, am self-taught).

No bold takes. No grades. No non-expert advice.

As this article is composed from the comfort of a couch, its singular goal is to provide facts. No couch coaching.

Part 1: The Pythagorean Theorem & Predicting Wins

What you need to know: The Pythagorean Theorem is defined as a^2+b^2=c^2.

By using the principles of the Pythagorean Theorem, the number of wins in a given season can be predicted. From there, the evaluation of a particular season takes shape: answering the question of "did the WKU underperform, overperform, or meet the prediction?"

Twenty years of WKU Football data (1998-2018), a MAD (median absolute deviation) of 0.078, and an exponent of 2.5 yields the following table:

The greatest absolute errors came from the following seasons: 2001, 2003, 2006, 2007, 2009, 2010, and 2011.

Based on the other columns in the table, those seasons can be categorized as:

• Underperformed: 2001, 2003, 2007, 2009, 2010
• Overperformed: 2006, 2011

Just because a team met the prediction or underperformed does not mean it was an average or poor season. Likewise, overperforming does not necessarily make for a great season.

All factors considered:

• Top five seasons: 2000, 2002, 2004, 2015, 2016
• Bottom five seasons: 2008, 2009, 2010, 2017, 2018

Part 2: Multiple Regressions and What Makes a Team Win

What you need to know: A multiple regression is used to predict the value of a dependent variable based on the value of two or more other independent variables.

2a: The Eight Variables

What makes a team win? How does this translate to WKU Football?

For Part 2, eight independent variables were used to predict scoring margin: return touchdown differential, penalty differential, passing yards/attempt, rushing yards/attempt, turnovers, defense passing yards/attempt, defense rushing yards/attempt, and defensive turnovers.

2b. Pass Yards/Attempt

Result: Pass yards/attempt is [consistently] one of the most highly significant determinants of WKU Football (and Conference USA) scoring margins.

To verify the result, regressions were conducted for WKU Football (1998-2018), WKU Football in the FBS era (2008-2018), and Conference USA (2018). This is not a groundbreaking finding specific to WKU; pass yards/attempt is the most highly significant determinant of wins in the NFL, as well (Winston, 2009).

This trend does not make the other seven variables any less important. Think of this phenomenon like a class implementing weighted grades. All grades matter, but exams receive the most weight: having the most impact on your final class grade.

WKU Football in the FBS era (2008-2018)

After running a regression (and eliminating the intercept, which had a non-significant p-value) the output is as follows:

The R-squared of 0.95 tells us that the following equation explains 95% of the variation in WKU's scoring margin. The standard error of 60 means that our equation predicts the scoring margin for 95% of the seasons within 120 points.

predicted team scoring margin for season = 6.3(RET TD) - 0.2(PENDIF) + 74.1(PY/A) + 18.8(RY/A) - 2.9(TO) - 58.4(DPY/A) - 16.5(DRY/A) + 2.1(TO)

From this regression we learn (after adjusting for other independent variables):

• an extra TD on kick/punt returns is worth 6.3 points
• an extra yard in penalties costs 0.2 points
• an extra PY/A is worth 74.1 points
• an extra RY/A is worth 18.8 points
• an extra TO costs 2.9 points
• an extra DPY/A costs 58.4 points
• an extra DRY/A costs 16.5 points
• an extra forced TO produces 2.1 points

There is a 1/25 (4.0%) chance that pass yards/attempt does not affect the scoring margin given the eight variables. If non-significant variables are excluded, there is a 3/10,000 (0.03%) chance that pass yards/attempt does not affect the scoring margin.

WKU Football (1998 - 2018)

A regression on twenty years of WKU Football will be different than the FBS era since WKU - like a vast majority of teams - has expanded their use of the pass. There are ebbs and flows - schematically and national trends - that cause one season to be completely different from the next.

After running a regression (and eliminating the intercept, which had a non-significant p-value) the output is as follows.

The R-squared of 0.88 tells us that the following equation explains 88% of the variation in WKU's scoring margin. The standard error of 65 means that our equation predicts the scoring margin for 95% of the seasons within 130 points.

• an extra TD on kick/punt returns is worth 2.2 points
• an extra yard in penalties costs 0.4 points
• an extra PY/A is worth 35.2 points
• an extra RY/A is worth 21.6 points
• an extra TO costs 2.5 points
• an extra DPY/A costs 17.4 points
• an extra DRY/A costs 35.9 points
• an extra forced TO produces 0.003 points

There is a 17/2,000 (0.85%) chance that pass yards/attempt does not affect the scoring margin given the eight variables.

Remember the five best seasons referenced in Part 1? It should be of no surprise that the following WKU Football seasons saw the highest passing yards per attempt:

• 2000 - 10.3 PY/A
• 2002 - 10.0 PY/A
• 2004 - 9.2 PY/A
• 2015 - 9.4 PY/A
• 2016 - 10.0 PY/A

This is not a knock on current quarterbacks Steven Duncan or Davis Shanley. Rather, it seems that the PY/A was a product of the system in place.

2c. Penalties

Based on the same regression, penalties tend to have the largest p-value and the smallest coefficient. There is a greater chance that penalties do not affect the scoring margin.

Do not fall under the impression that penalties (or any variable with a large p-value) do not matter. The explanation is simple.

Penalties generally affect plays and drives, not the outcome (yes, there are exceptions). An average or above average offense will compensate for penalties. For example, a false start will push the offense back 5 yards. However, the subsequent play(s) might consist of a 9 yard pass and 4 yard rush: making up for the false start. The same can be applied to defensive penalties. An average or above average defense will compensate for the penalties. For example, a personal foul will move the opponent up 15 yards; afterward, the defense only allows a 4 yard rush and, then, breaks up two passes.

WKU's worst penalty differential during their FBS era falls within the 2015 and 2016 seasons (23-5 record, Top 25, and back-to-back conference championships). During those seasons, both sides of the ball were able to overcome any significant penalty setback.

Part 3: Success Rates

What you need to know: Success Rate (SR) is a tool used to measure the efficiency (success) of each play. A first down is considered successful if 50.0% of the needed yards are gained. A second down is successful if 70.0% of the needed yards are gained. Third and fourth downs are successful if 100.0% of the needed yards are gained. Below is an example of some down and distance situations.

AS A FRAME OF REFERENCE, THE AVERAGE SUCCESS RATE FOR FBS TEAMS FROM 2005-10 WAS 41.6%. -BILL CONNELLY, SBNATION FOOTBALL STUDY HALL

Defining "success" is subjective even with SR. If a team gained 9 yards on 3-11, it is technically an unsuccessful play. But those 9 yards have value if it puts the team into field goal range. In this situation, consideration should be made for, both, the SR and the EPA. Regardless, this section will still focus on non-adjusted numbers.

From 2014 - 2017, WKU's offensive success rate was above the FBS average for that particular season. WKU's SR from 2014 - 2016 was Top 5 throughout the FBS: #5, #1, and #2, respectively. Between 2016 and 2017, the Hilltoppers SR decreased by a difference of -7.6 or 14.4%. In 2018, WKU had a SR of 40.0%: below the FBS average of 42.23%

WKU's defensive SR has remained steady for four seasons. From 2015 - 2018, the defense had a SR of 41.325%: ever-so-slightly better than the FBS average of 41.435%. In 2017 and 2018, WKU's defensive SR went above the FBS average; however, much of this can be attributed to the offense's setbacks.

3a. 3rd Down SR

SR on 1st and 2nd downs give teams a bit of leeway. However, 3rd downs conversions are key in keeping a drive alive. If a 3rd down is not converted, a team is forced to either kick a field goal, punt, or go for it on 4th down. What is WKU doing when it matters "most?"

3b. 3rd Down SR per Quarter

Context is important. Take, for instance, 2016. WKU had a great season (11-3, C-USA Champions, and Boca Raton Bowl Champions), yet their 3rd down SR (garbage time included) took sharp declines quarter-to-quarter. This can be attributed to the fact that WKU usually held a significant lead, gave non-starters playing time, and then ran the clock during the second half.

3c. 3rd Down SR Leaders

Quarterback (passes only)

Running Back (runs only)

Receivers (receptions only)

Two interesting tendencies were illuminated by pinpointing an individual player, their success rate, and targets/carries.

1. Taywan Taylor was rarely targeted on 3rd downs. Jared Dangerfield, Willie McNeal, and Nicholas Norris each had more targets. Taylor was used as a deep threat on previous downs.
2. In 2018, there was a pattern of players receiving equal, or close to equal, targets or carries on 3rd downs.

3d. 3rd Down SR Allowed (Defense)

The 2015 and 2018 seasons saw the lowest (best) success rates.

The 2018 season saw a five-season high on 3rd down (standard down) SR allowed but, also, a five-season low on 3rd down (passing down) SR allowed.

Part 4: Explosiveness

What you need to know: Explosiveness or explosive plays are exactly as the name indicates: big plays. In terms of distance, an explosive play in the NFL is either a 20+ yard pass or 10+ yard run. An explosive play in college football is, typically, a 16+ yard pass or 12+ yard run.

I WANT TO GET BACK TO THAT BRAND. I WANT, EVERY TIME WE THROW THAT FOOTBALL DOWN THERE, EVERYBODY TAKES A DEEP BREATH AND GOES [GASPS] 'OH MY GOD. THERE THEY GO AGAIN.' -TYSON HELTON, WKU FOOTBALL HEAD COACH

It seems - aside from winning - what has been sorely missed of the Hilltoppers is their aptness for explosive, aggressive, and highlight-worthy plays.

...the type of explosive and aggressive plays that had the Patriots' Bill Belichick naming a Super Bowl LI play after WKU.

We call it the Hilltopper play...the fake kneel down. -Bill Belichick

If pure explosiveness is isolated and non-adjusted- disregarding solely successful explosive plays or other adjustments - how much do explosive plays (16+ yard pass, 12+ yard rush) matter in terms of game outcome? Or is it just entertainment and puts fans in seats?

Simply put, explosive plays [usually] increase win probability as well as increases expected points.

4a. Offensive Example - WKU vs. LA Tech (C-USA Championship), 2016

With 10:04 minutes left in the 3rd quarter, WKU was down 3 points (38-41). To this point, the Tops had a win probability of 60.8% and were expected 0.793 points on the play given the time left, down, distance, and field position. Ace Wales had just rushed for 1 yard on 1st & 10.

2nd and 9. Quarterback Mike White hands it off on the jet sweep to Nicholas Norris. The play picks up 26 yards and puts the Tops at the LA Tech 39 yard line. WKU's win probability is now 69.9% and is expected 2.983 points post-play.

Because of the explosive play, WKU was in LA Tech territory and, eight plays later, Skyler Simcox was able to make the 28 yard field goal. WKU and LA Tech were, now, tied 41-41.

4b. Defensive Example - WKU vs. Southern Miss. (C-USA Championship), 2015

Likewise, it is critical that the defense limit explosive plays. In the event that the opponent is able to execute an explosive play, the defense must amp-up its efforts to kill the opponent's momentum on the subsequent play. For example, during the 2015 C-USA Championship game, Southern Mississippi's Nick Mullens completed a 25 yard pass to Michael Thomas: putting the Golden Eagles at WKU's 23 yard line and with a 77.4% win probability. However, WKU's defense was able to hold Southern Miss. to a 3 yard rush on 1st and 10. Southern Miss., then, committed a holding penalty: pushing the offense back 10 yards. On the final play of the drive, Southern Miss. posted a 3 yard loss and WKU's defense was able to force a fumble on 2nd & 17. WKU scored on their next drive: putting them up 31-28 and holding Southern Miss. to 28 points the remainder of the game.

Part 5: S&P+ (SR + IsoPPP)

*This is not Standard & Poor's

What you need to know: S&P+ ratings consider efficiency (success rate), explosiveness (IsoPPP), and factors related to field position and finishing drives.

Part 6: Decision Making

6a. 2nd-and-short

In a recent Twitter thread, Sean J. Taylor - a social scientist and statistician - conducted a litmus test on NFL coaches and their 2nd and short decision making.

Why this particular down and distance?

A coach's proclivity in certain situations, like 2nd-and-short, illustrates his optimization strategy. Essentially, is the coach trying to score or extend the drive (get the first down)? Is he risk-averse or risk-taking? Does he mostly favor the run (first down) or does he choose the pass despite the short yardage (more yards that primes the team to score at a faster clip)? There is a time and place for each. Nothing is ever quite as black and white as we would like it to be.

6b. 4th Downs

4th downs are, often, synonymous with the phrase "decision making." A team has three options on 4th down: kick a field goal, punt, or go for it. Based on what the coach elects to do given a particular situation and the subsequent outcome, fans are likely to characterize the coach as having good or poor decision making skills.

What many spectators may not process during such a short timespan is the present play selection (punt, field goal, missed field goal, 4th down converted, 4th down not converted), expected points per play selection for both teams, and probability.

Expected points are the "average potential points a team can expect given a certain situation" (Burke, 2014). However, EP can look vastly different between various circumstances.

Without diving too deep into the expected points model and calculator, we can still establish a confidence "level" or percentage that was needed to go for it. The higher the percentage, the more risky (>50.0%)

The goal of Part 6 is not to identify the coach with the best 4th down decisions or declare whether or not a decision was "good," rather, it is to illustrate the complexities of such down. Even if spectators or commentators do not applaud the decision, a coach has a reason for whatever decision. Each team, season, and game is different, and decisions are made based on more than expected points. For example, a typical probability for a 42 yard field goal might not be the typical probability for a specific team's kicker: prompting a coach to go for it rather than kick given a certain down, distance, yards to go, score, and time remaining.

Maine (2018) - 4th and 1 at the WKU 15

Coach's Decision: Go for it

Result: Failed; WKU turned the ball over on downs. Maine would score on that drive.

Kicking a field goal was out of the question. WKU had the option to punt or go for it. Per probability calculations (which are not perfect and require further tweaking), Sanford needed roughly a >90% confidence level in order to go for it. The risk of WKU turning over on downs and Maine's expected points starting at WKU's 15 yard line far outweighed the expected points if WKU was able to convert.

Interpreting the decision: Although Alex Rinella had averaged 37.5 yards per punt to that point, WKU was at their own 15 yard line. Maine would likely have decent field position even if WKU chose to punt. Perhaps the WKU defense could have stopped Maine given more breathing room. The game had been tied 21-21 for six-straight drives, and Sanford was trying to keep the drive alive and allow momentum to swing WKU's way.

Old Dominion (2017) - 4th and 3 at the ODU 15

Coach's Decision: Go for It

Result: Failed

WKU's options were to either kick the field goal or go for it. Sanford needed roughly a 36%-40% confidence level in order to go for it.

Interpreting the decision: This was not a huge risk considering the field position, and it was still early in the game.

Vanderbilt (2016) - 4th and 2 at the WKU 43

Coach's Decision: Punt

Result: Jake Collins punts for 34 yards, out-of-bounds at the Vandy 23.

WKU's options were to either punt or go for it. Brohm needed roughly >90% confidence level in order to go for it.

Interpreting the decision: WKU was up 7 with 1:25 left in the 2nd Quarter. A failed 4th down would put the ball back into Vanderbilt's hands at WKU's 43 yard line. The Commodores would, then, have enough time for a couple of plays and/or kick a field goal. Heat could have been a contributing factor, as well. At that point, it was around 90 degrees fahrenheit, and Brohm may have wanted to preserve the offense's energy for the 2nd half.

FAU (2014) - 4th and 2 at the FAU 24

Coach's Decision: Kick FG

Result: Garrett Schwettman makes FG; WKU up 17-7

WKU's options were to either kick the field goal or go for it. Brohm needed roughly a 38% confidence level in order to go for it. Garrett Schwettman made the 41 yard FG.

Interpreting the decision: With 2:55 left in the 1st Quarter, Brohm may have been seeking quick, easy points. There was still plenty of game left, and even a field goal would make it a two possession game.

Part 7: Evaluating 2018 Position Groups

7a. Quarterbacks

Part 3a does not include sophomore quarterback Kevaris Thomas. In 2018, Thomas posted 1 pass attempt and 6 rush attempts throughout three games. Given such a small sample size and a lack of significant data, neither a hypothesis nor a fair evaluation could be made.

The above is a statistical comparison chart of two WKU Football quarterbacks from ESPN.com. Based solely on this chart (excluding CMP, ATT, and YDS), QB 2 led the position group in seven of nine statistical columns.

WKU spent most of the 2018 season going back n' forth between two quarterbacks after Drew Eckels' injury. As the saying goes: "If you have two quarterbacks, you have none."

But there is a question buried in those statistics that begs to be asked. Unfortunately, it is incredibly difficult to answer and will not be answered until the spring or fall.

If QB 2 (Davis Shanley) had the better statistics (including pass yards/attempt (see Part 2)), why or how did WKU score (and win) with QB 1 (Steven Duncan)?

Reason #1: The sample size is too small and play calling per quarterback

Football's sample size (number of games) is extremely small. A normal, healthy, and seasoned QB 1 will start anywhere between 12-14 games per season (depending on post-regular season games). Even then, if the team surges ahead and maintains the lead, other quarterbacks get playing time aka "experience." In 2018, Shanley played in 8 games and Duncan played in 9. You can only forecast so much with such limited data. It is completely possible that there is no definitive answer for the proposed question. Neither quarterbacks' impact should be marginalized nor will their performance be a guarantee for 2019 (new coach, new system).

As mentioned before, there is no way for me to identify every single play-call. The only way to justly evaluate Shanley and Duncan would be to compare their performance on the same play-calls using a rating system. In other words, you cannot compare the success (or failure) of a quarterback sneak to the success (or failure) of hitting a slant route.

Reason #2: Defensive efficiency opened up the field for Duncan

So often we forget that a player does not independently operate. His stat sheet is a combination of his own ability and what players - teammates and opponents - are able to do (or not do). When we analyze a player's stats, it is not as simple as "He had a higher completion rate, more yards, and more touchdowns." If only.

WKU's defensive efficiencies were best during the UTEP (93.0%), Louisville (76.0%), and LA Tech (71.0%) games (Connelly, 2018). Two out of three of those games, WKU won.

Essentially, the defense's efficiency and havoc rate limited the opponent's offense: forcing the opponent's offense off of the field and providing WKU's offense with more possessions to make plays.

Reason #3: Ratings (PR and Total QBR) per game

What you need to know: Passer rating (PR) is a statistic used to evaluate a quarterback's passing performance. Total quarterback rating (QBR) is an ESPN proprietary statistic that measures the performance of quarterbacks based on their total contributions (pass, rush, turnovers, penalties, expected points added (EPA), win probability added (WPA), etc.).

Season totals - while important - summarize the season, not what happened per game. Stats can be skewed and wildly misleading if a quarterback only saw action on a few snaps.

As the chart indicates, Shanley had, both, the higher PR and Total QBR. Duncan saw higher Total QBRs in more games (Ball State, Charlotte, and UTEP).

Reason #4: Shanley's Marginal Efficiency and Duncan's Marginal Explosiveness

Bill Connelly's marginal efficiency and marginal explosiveness is defined as:

• marginal efficiency: a player's SR - the expected SR
• marginal explosiveness: a player's IsoPPP - the expected IsoPPP

Duncan had a marginal explosiveness of 0.1 while Shanley had a marginal explosiveness of -0.1.

When Duncan was on the field, he had more successful explosive run and pass plays. Explosive plays can increase a team's win probability (WP) exponentially. When Shanley was on the field, he had a better marginal efficiency (1.0%): meaning he was more consistently successful per play.

Reason #5: Intangibles

It is possible that Duncan possessed an intangible. Whenever Duncan went in – whether as relief or replacement – his debut series (as, specifically, relief or replacement) resulted in a TD (Ball State, Charlotte, FIU, MTSU). No data can explain "next man up," but it is an intangible that mattered a great deal.

7b. Receivers

A comparative chart of 2017 and 2018 receivers reveals some interesting numbers and trends.

1. Marginal Efficiency decreased from 2017 to 2018, while explosiveness improved.

The receivers were not picking up the efficiency (SR and expected SR), but they were executing more explosive plays (16+ yards).

2. The quiet rise of Jacquez Sloan

Through the first four games of the 2018 season Sloan only had 8 receptions for 117 yards. Against Marshall, Sloan had 4 receptions for 87 yards and a long of 60 yards. Then, against Charlotte, Sloan had 6 receptions for 83 yards and a long of 42 yards. Although he did not score in 2018, it is obvious that Sloan is a speedy, elusive wideout who seems adept with the deep ball and creating separation.

Look no further than his marginal efficiency and marginal explosiveness. In 2018, Sloan had a marginal efficiency of 17.0% (best of WKU receivers) and the third best marginal explosiveness (0.15). Additionally, Sloan led in yards per catch and yards per target.

Simply put, what Sloan lacks in size (5'9"), he makes up for in successful and big plays.

3. Garland LaFrance

A name I would anticipate hearing a bit more in 2019 would be Garland LaFrance. In 2018, (in addition to his rushing stats) the 5'10", 175 lb [then] freshman running back went 20-for-20 for 148 yards and 1 touchdown. On the season, LaFrance had a [receiving] marginal explosiveness of 0.88: the highest on the team.

Given what we have seen of LaFrance despite such limited data, it is not so much about what he did do in 2018 but, rather, what he could do in 2019. Especially with a new coaching staff, it would not be surprising if LaFrance was utilized in the passing game as a slot receiver.

7c. Running Backs

WKU's running corps was unquestionably one of the team's most improved position groups of 2018.

2017 was disappointing for the Hilltoppers' ground attack. After rounding out the bottom of the FBS in total rushing yards, yards per carry, longest rush, and yards per game, 2018 saw a promising turnaround. In 2018, WKU improved to #108 in total rushing yards, #105 in yards per carry, #97 in longest rush, and #104 in yards per game.

Joshua Samuel had the highest marginal efficiency and explosiveness among WKU RBs

It is not uncommon to see a team's RB1 have the highest marginal explosiveness among RBs, but an average or below average marginal efficiency (i.e., FAU's Devin Singletary). This could be because the player has more carries which allows the defensive line more opportunities to stuff. In other words, there is a certain predictability to his usage; although he may be stuffed, he is still able to create explosive plays, elsewhere, in order to revert to the mean.

Even so, WKU's Joshua Samuel led the the running corps (excluding quarterbacks) in marginal efficiency (4.3%) and explosiveness (-0.01). D'Andre Ferby was behind Samuel with an efficiency of 0.7% and explosiveness of -0.03.

With the departure of Ferby, another RB will have to step up. WKU now has experienced options with Gino Appleberry Jr., Garland LaFrance, and Jakairi Moses (among others).

7d. Offensive Line

Like the running corps, 2017 proved to be less-than-stellar for the offensive line. Upon the season's conclusion, the Tops had allowed 48 sacks (3.7 per game) and the ground attack was dismal.

Offensive lines can be difficult to evaluate. Many reference, specifically, the o-line's allowed sacks and/or experience/starts as a measuring stick for evaluation purposes. While allowed sacks and experience matters a great deal, exclusively using the two for assessment is a disservice and marginalization of the o-line's impact on the passing game and ground attack.

Using data pulled from Football Outsiders, the following comparative table was made:

The o-line's setback in 2017 could be chalked up to youth and inexperience, especially after losing the likes of Forrest Lamp and Brandon Ray; however, 2018 saw even more youth and inexperience, but better stats than 2017.

Of particular note in regard to the 2018 season, WKU's o-line opportunity rate was better than that of 2016 or 2017. Essentially, the percentage of carries where 5 yards are available and the player is able to gain 5 yards, increased. Hence, the ground attack improvement from 2017 to 2018.

7e. Defensive Backs

A positive efficiency and explosiveness is associated with being "good" offensive statistic. For the defense, the lower the efficiency and explosiveness, the better. In other words, a negative (or low) efficiency or explosiveness means that the defense or player is doing their job and limiting the success or explosiveness of the offense.

However, defensive backs tend to have the highest (sometimes positive) efficiency and explosiveness. This is not necessarily "bad," but demonstrates how difficult it is to defend against the pass (think of Mike Leach's Air Raid offense at Texas Tech and Washington State). To put it into perspective, Greedy Williams (CB, LSU) - considered one of the best defensive backs in the nation - had a marginal efficiency of 24.8% and a marginal explosiveness of 0.10.

Especially in thinking of the upcoming season, Ta'Corian Darden and Devon Key are the names to know and continue watching (if you have not, already). Key has been a workhorse the past two seasons: amassing 165 total tackles, 7.5 tackles for loss, 4 interceptions, and 8 pass breakups. Darden with equally impressive numbers, also, allowed the least amount of yards per play (7.4) and was one of the defensive back leaders in marginal efficiency and marginal explosiveness.

7f. Defensive Line & Linebackers

Ben Holt (LB) posts Joel Iyiegbuniwe numbers

Holt - the son of ex-WKU Football defensive coordinator Nick Holt - absolutely took off in 2018: tallying 82 more tackles, 9 more TFLs, 1 more sack, and 9.5 more run stuffs than he did in 2017. His 2018 stat sheet was almost identical to that of Joel Iyiegbuniwe (Chicago Bears).

On February 14, 2019, it was announced that Holt would be transferring and joining his father at Purdue.

Eli Brown (LB) makes an immediate impact

Bowling Green native and University of Kentucky transfer made noise when utilized. In addition to delivering some highlight worthy hits, Brown was one of the defense's leaders in yards per play allowed, marginal efficiency, and marginal explosiveness. With the transfer of Holt, Brown will have to step into a leadership role; he will be the most experienced linebacker on the team.

Jaylon George (DL) leads line

In 2018, George led the team and defensive line in yards per play allowed (0.5), marginal efficiency (-30.8%), and marginal explosiveness (-0.86).

7g. Special Teams

Kicking woes

Kicking woes headlined WKU's special teams unit in 2018. Combined, Ryan Nuss and Alex Rinella hit 10-of-17 field goals (58.8%).

By the end of the season, WKU had allowed the second most blocked kicks (6).

From a purely statistical standpoint, this may be why WKU opted in favor of some controversial and risky 4th down attempts. For example, making a 22 yard FG has a probability of, about, 87.0% (based on ten seasons of data); however, the 2017 and 2018 seasons saw that probability drop to 58.8%.

Part 8: Statistical Improvements and Declines

The above table is a rating and ranking system I created based solely on C-USA play.

Items of Note:

• WKU averages the highest rating and ranking through five seasons.
• UAB and MTSU are the most consistently highly rated/ranked teams. This is impressive given the fact that UAB's program was eliminated during the 2015 and 2016 seasons. They were able to jump back into conference play effortlessly.
• After the 2016 season, the home team advantage dropped to 2.4 points in 2017 and 0.2 points in 2018.

Ratings and rankings can, also, be applied to each teams' offenses and defenses.

As a five year average, WKU has the highest offensive rating (11.1) and the ninth best defensive rating (0.6).

The top offenses are as follows:

1. WKU (11.1)
2. MTSU (6.3)
3. LA Tech (5.2)
4. (Tie) Southern Miss. and UAB (2.8)

The top defenses are as follows:

1. Marshall (-8.8)
2. UAB (-5.0)
3. LA Tech (-4.0)
4. MTSU (-1.3)
5. UTSA (-1.0)

Part 9: Upcoming Talent & Returning Production

9a. Returning Production

Production is a better indicator of contributions returning, not who. For example, what percentage of passing yards are returning? What percentage of sacks are returning?

Returning production is neither "good" nor "bad." There is a lot that returning production cannot encapsulate (i.e., low returning production but with breakout performances or significant returning production with poor performances). Returning production does not equate wins and losses nor does it predict strengths and weaknesses. In other words, you cannot assume that WKU will defeat a team with a lower returning production and vice versa. In 2018, the top 10 teams in terms of returning production totaled a record of 68-59: none of whom made the Top 25 and only 1 team won a conference championship.

According to Bill Connelly, WKU is ranked #14 in returning production for 2019; the offense returns 81% of its production while the defense returns 72% (pre-Holt transfer). From there, I calculated returning production per position and statistical category.

Given the returning production, how does the 2019 class cohere to the program?

9b. Upcoming Talent

Analyzing the positional breakdown of incoming players can provide insight regarding the direction of the program.

Per WKU's "2019 Football Signing Class" the Tops brought in: 7 wide receivers, 1 running back, 5 offensive linemen, 2 kickers (1 K and 1 P), 1 defensive end, 2 defensive tackles, 1 linebacker, and 3 defensive backs.

• The addition of Celestin Haba (OLB/DE) could alleviate the lack of linebacker production returning from 2018. Haba played at Scottsdale C.C. and had previously committed to MTSU before flipping to WKU. While at Scottsdale, Haba led all junior colleges in sacks (17.5). A playmaker at middle linebacker will still need to emerge.
• Five o-linemen may seem like a lot, but 1) as a position group, the o-line has the most men on the field and 2) without a good o-line, you likely do not have the time for big plays to develop, the running backs have no open lanes, and the quarterback gets hit (or rushed and makes bad decisions). It all starts up front.
• As indicated by Helton, one goal of 2019 (and onward) is to return to an explosive, exciting brand of football. To play an explosive game, you need explosive receivers (i.e., Taywan Taylor, Nicholas Norris, and Jared Dangerfield); hence, the addition of seven receivers (6 WR, 1 TE).
• The sole running back will be a walk-on. There is not an immediate need for this position group.
• John Haggerty (P) will bring back the rugby-style kick that had served WKU well years prior. In 2018, WKU's punt efficiency ranked #125 with a rating of 44.0%. Cory Munson (K) will join Alex Rinella on place-kicking and kickoff duties.
• With the departure of defensive backs DeAndre Farris (CB) and Drell Greene (S), the Tops add three DBs. I would expect some DBs to either move positions or WKU might use the bigger DBs in a nickel formation and/or as a more speedy linebacker.
• Finally, WKU added three defensive linemen (1 DE and 2 DT) to the roster. The 2 DT fill spots left from Evan Sayner and Julien Lewis.

Part 10: Looking Ahead & Conclusions

Based on the aforementioned data, I would anticipate the following:

Quarterback Competition

Given such a small sample size, there are no guarantees as to who should or will start for the Tops in 2019. At this point, anyone suggesting that one quarterback is superior to another is basing their suggestion on opinion or personal preference rather than empirical evidence (very little exists). Each has his strengths. Shanley is strategic and consistent while Duncan is explosive. There is little known [statistically] about Kevaris Thomas, and for that reason, you cannot count him out, either. Spring ball will be incredibly important for this position group and their preliminary evaluation under a new coaching staff.

More Explosive Plays and Players/Less Conservative Ball

Head coach Tyson Helton even said as much, so this should not come as a surprise.There is no denying the impact that explosiveness has on a game and season: whether it comes in the form of a 12 yard gain off of a sweep or a 75 yard gain off a flea flicker.

Congruency on Defense

From, both, a production and leadership standpoint, the defense will take a bit of a hit with the transfer of Ben Holt. However, the retention of the defensive staff in addition to the production that is returning helps maintain a level of congruency. The defense is still loaded with experience and players on the rise.

A System for the Players

...you have their back...you're going to allow them to be the playmakers that they are. Every player has a different skill tree and it's our job as coaches to put them in the position to showcase their talents. -Tyson Helton

It is not my place to predict the success of Helton's system. However, it seems that Helton and staff will be emphasizing and exploiting the skillset of each player: a system for the players, if you will.

References

Alamar, B. & Mehrota, V. (2011). Beyond ‘Moneyball’: Rapidly evolving world of sports analytics, Part I. Analytics: driving better business decisions. http://analytics-magazine.org/beyond-moneyball-the-rapidly-evolving-world-of-sports-analytics-part-i/.

Burke, B. (2014). Expected points and expected points added explained. Advanced Football Analytics. http://www.advancedfootballanalytics.com/index.php/home/stats/stats-explained/expected-points-and-epa-explained.

Connelly, B. (2017). 2017 college football statistical profiles. Football Study Hall.

Connelly, B. (2018). 2018 college football statistical profiles. Football Study Hall.

Connelly, B. (2013). Study Hall: college football, its stats and its stories. Createspace Independent Publishing Platform.

Golliver, B. (2015). TNT's Charles Barkley rants about basketball analytics, jabs Rockets GM. Sports Illustrated. https://www.si.com/nba/2015/02/11/charles-barkley-analytics-video-daryl-morey-houston-rockets-gm.

NCAA offensive line stats. (2016-2018). Football Outsiders.

Winston, W. (2009). Mathletics: How Gamblers, Managers, and Sports Enthusiasts Use Mathematics in Baseball, Basketball, and Football.