DraftGeM: A New Tool to Compare College and Pro Football Players
College football is full of comparison debates.
Which Conference has the best players?
Tua or Trevor?
Fromm or Fields?
With only a 12 game schedule and 130 Football Bowl Series (FBS) teams, it is difficult to rank teams week-to-week with any certainty. Few inter-conference big games add fuel to these heated debates. The ability to compare teams is a real challenge. The polls, Elo, and FPI attempt to bring some sense of order to the chaos.
Comparing players across the skill-diverse college football landscape was even more daunting…until now.
Our rating system, DraftGeM, rates all FBS defensive players and skill position offensive players. Our player rating is updated each game based on player performance and the strength of opponent. We have game-by-game ratings of all players since 2005, far more data than any single person can retain. This enables intuitive visual analysis of a player’s college careers at a glance.
How long was his career?
Did he finish his career on a high or did he drop off his final season?
Was he consistent?
Did he excel or struggle against top competition?
DraftGeM brings player comparisons to life. We show how players stack up in multiple dimensions. We can compare players against their peers or historic players.
Did they have similar career performance (rating) curves?
Did they play in similar number or games?
Who is a player most similar to?
As an example, let’s look at the Tua or Trevor question. We can quickly compare the game-by-game ratings of the two star quarterbacks.
Tua Tagovailoa, Alabama’s quarterback, played in relief of Jalen Hurts in 2017 and rose from the rating baseline after his brilliant second half performance against Georgia in the CFP championship game. His rating rose steadily throughout the 2018 season and continues to rise into the 2019 campaign.
Trevor Lawrence, Clemson’s quarterback, backed up Kelly Bryant the first four games of 2018 before getting the starting job. His rating also rose steadily throughout the season, but at a slower pace than Tua’s. You can also see that Lawrence’s has had a shaky start to 2019, relative to Tua.
Above we show our composite or overall rating, but we can compare their ratings in any of the statistical categories shown above the graph (yards per attempt, yards per completion, yards per game, interceptions, completion percentage, touchdowns, and passer efficiency).
Other observations from the graph:
Both have relatively short careers, to date.
The length of the curve indicates the number of games played. The darker gray area represents the envelope (minimum and maximum values) of active college quarterbacks and the lighter gray all quarterbacks since 2005.
Both are at or above the NFL Average
The dotted NFL average line indicates the average final DraftGeM rating of all FBS quarterbacks (since 2005) who have performed well enough in the NFL to earn an approximate value (AV) of at least 1.
Interested in more detail? DraftGeM offers a detailed view that annotates each season’s performance curve with game-by-game opponents. We can see both quarterbacks had a bad week toward the end of the 2018 season. What happened there? DraftGeM allows you to easily analyze that, as shown below.
Tua and Trevor both struggled in their conference championship games. In Tua’s case, Jalen Hurts came in to save the day and led Alabama to the college football playoff. In Trevor’s case, Travis Etienne had a monster game to send Clemson to the playoff en route to their fourth straight postseason meeting with the Tide.
Want to see how Tua and Trevor stack up in each statistical rating in one glance? DraftGeM offers that as well. You can quickly see that Tua is ahead of Trevor in every statistical rating. The outer perimeter of the spider plot represents the all-time best and the dotted line indicates the average for all FBS quarterbacks since 2005. These superstars have convex shaped profiles, indicating well-rounded talent. This graph can highlight potential flaws when the shape is flattened or concave in a particular area. For example, an inaccurate quarterback will exhibit a dent where we evaluate completion percentage, and possibly their interception rating.
DraftGeM works for defense too. Take a look at our top five defensive linemen in the 2019 NFL draft. The SEC, ACC, and Big Ten are all represented, while the Big 12 and Pac-12 are not. Ed Oliver proves that NFL talent is not exclusive to the Power 5. The specific order is unimportant. Our ratings are intended to show NFL potential through college game performance. General managers, scouts, and coaches still need to have eyes on players to fully evaluate each player in their system. That said, our top five defensive linemen were drafted in the first thirteen picks of the draft.
This performance is not limited to last year’s NFL draft. Over the last ten draft years we have proven that DraftGeM successfully identifies NFL talent. We have identified undrafted talent, such as Pierre Thomas, LeGarrette Blount, BenJarvus Green-Ellis, Doug Baldwin, Danny Amendola, Cole Beasley, Willie Snead, and Case Keenum to name a few.
DraftGeM also warned away from players that have been overhyped, most notably Christian Hackenberg. His curve fell well below the baseline rating, nowhere close to the NFL average rating. His spider plot shape was concave for completion percentage. His size and athleticism may have been great, but our data showed it did not translate into on-field performance. DraftGeM would have pointed the Jets toward Dak Prescott for their second round pick.
Moving from past to present, we would have warned the Giants away from Daniel Jones, especially with a first round pick. DraftGeM rates Tampa Bay’s Blaine Gabbert as a similar quarterback at the finish of his college career. Their final college seasons track closely, as can be seen in their curves below. It will be interesting to see how Jones’ career pans out.
How does it work?
At its core, DraftGeM leverages the same methodology as Elo ratings. A player’s rating is dictated by his performance and the strength of the opponent in which that performance was achieved. We discovered a way to translate the Elo concept to individual player ratings based on their game statistics.
To accomplish this we must rate players and teams. Rather than players going head-to-head with another player, players compete against teams. The rating difference between a player and the opposing team translates to an expected performance for that player in that game. Think of it as the player’s own Vegas line or point spread.
The player is then evaluated based on that line. Exceed expectations and the rating rises. Fall short, and so does the rating. Ratings are adjusted after every game for players and teams alike. Over time, players tend to find their appropriate rating. With the complex nature of an individual’s performance within their team, there will always be some game-by-game volatility, but most player curves plateau. College players with short game participation careers may never reach their level state.
For those that make it to the NFL, we start over and rate their NFL performance using the same methodology. NFL curves tend to be longer, as many careers extend beyond the duration of college eligibility. NFL players’ curves tend to flatten and even drop toward the end of their professional careers. Just like the college data, the ratings and player curves make football sense.
Want to see how your favorite players rate? Interested in comparing with others, old and new?
Check out DraftGeM and tell us what you think draftgem@visionistinc.com. Note: DraftGeM is designed for desktop and landscape tablet use.
We update current college and NFL players each week so you can see how their ratings evolve throughout the season.
