Post by Purple Pain on May 18, 2019 11:55:49 GMT -6
[OC] Attempting to Create an Objective "Runner Rating" by Citronaut1
Rest at link:
old.reddit.com/r/nfl/comments/boxqwe/oc_attempting_to_create_an_objective_runner_rating/
Also:
Since the 1973 season, the NFL has recognized passer rating as an official statistic in order to evaluate and rank quarterbacks. The passer rating formula has stood the test of time and, while complex, has served as a fairly accurate system to objectively view NFL quarterbacking.
Today, we will be making an attempt at creating another objective grading system, this time for the running back position. Using the passer rating formula as our template and a hint of statistical analysis, we should be able to generate a similar equation that can objectively rate runner performance.
Why Are We Doing This?
If the term "runner rating" sounds familiar, that's because it is; there have already been several attempts at creating such a formula, with various levels of success. Football Outsiders uses a system called DYAR, a complex observational strategy that utilizes success rate to evaluate players.
Another attempt was made by Daniel Heard, a former graduate student at Duke University. His methodology (similar to the one I will be using) was to adjust the passer rating formula to apply to the running back position. Heard's paper can be found here and I highly recommend looking at it prior to reading the rest of this piece.
Although both of these methods are good starts, there are a handful of issues that I have with each. Starting with Football Outsiders' DYAR ratings, their methodology is somewhat subjective and attempts to categorize players as "above" or "below" average. The goal of this exercise, however, is not to find an "average" line for running backs to be measured upon; rather an objective rating system that doesn't depend on outside influences.
Moving onto Mr. Heard's study, there is one key factor that I can simply not get past: the receiving stats. In Heard's introduction he states: "...I weighted rushing 87% and receiving 13% (based on rushing making up approximately 87% of the total touches for players in my data set.) " By basing a part of his formula off of a time-specific data set, this could cause some problems, especially for players who don't catch out of the backfield much. It could also undervalue players such as Alvin Kamara and Christian McCaffrey, who receive far more than 13% of their touches through the air. For that reason, we will be avoiding receiving stats altogether and focusing only on running performance.
Breaking Down Passer Rating
As stated previously, we will be using the passer rating formula as our template for this exercise. The formula is as follows:
Essentially, there are four separate equations evaluating three key categories: (1) efficiency, (2) scoring, and (3) turnovers. The four equations mentioned are:
The NFL's passer rating formula seems very "messy" at first, but is much simpler when broken down like this. To make this process as simple as possible, nearly all modifications to the formula will be done in these four equations (you'll see what I mean by that later on).
For any of those who are math savvy, you'll notice that each respective equation can result in some crazy totals. For that reason, the formula automatically places a value cap of 2.375 and value floor of 0. For example, if a player throws a single pass for a touchdown, equation (c) will round that respective value down to 2.375, rather than 20. We will be applying the same boundaries here, for simplicity's sake.
Creating the Formula
For starters, we need to adjust the amount of equations that we will be dealing with in our formula. When passing, there are four major stats that we use: completions, yards, touchdowns, and interceptions. Unfortunately, we only have three major running stats: yards, touchdowns, and fumbles. In essence, there is no statistic that can accurately supplement completion percentage, so we will be excluding equation (a) altogether.
To account for this change, we will simply change the denominator "6" in Figure 1 to "4.5". By doing this, we can still obtain a maximum "runner rating" of 158.3, just like passer rating.
Finding the Averages
Here is where things tend to get tricky. To accurately modify the equations, we will need to establish what an "average" running performance is. When the passer rating formula was created in 1971, statisticians based the formula on passing stats from 1960-1970, giving an "average" value of 1.000 for each respective equation (more on that here).
Through their research, they determined that a perfectly average passing game is 7 yards/attempt, one touchdown per 20 attempts (or 5% touchdown rate), and one interception per 18 attempts (or 5.5% turnover rate). We will be following the same method here, compiling all rushing stats from 1960-1970. Note: NFL and AFL stats have been combined.
Today, we will be making an attempt at creating another objective grading system, this time for the running back position. Using the passer rating formula as our template and a hint of statistical analysis, we should be able to generate a similar equation that can objectively rate runner performance.
Why Are We Doing This?
If the term "runner rating" sounds familiar, that's because it is; there have already been several attempts at creating such a formula, with various levels of success. Football Outsiders uses a system called DYAR, a complex observational strategy that utilizes success rate to evaluate players.
Another attempt was made by Daniel Heard, a former graduate student at Duke University. His methodology (similar to the one I will be using) was to adjust the passer rating formula to apply to the running back position. Heard's paper can be found here and I highly recommend looking at it prior to reading the rest of this piece.
Although both of these methods are good starts, there are a handful of issues that I have with each. Starting with Football Outsiders' DYAR ratings, their methodology is somewhat subjective and attempts to categorize players as "above" or "below" average. The goal of this exercise, however, is not to find an "average" line for running backs to be measured upon; rather an objective rating system that doesn't depend on outside influences.
Moving onto Mr. Heard's study, there is one key factor that I can simply not get past: the receiving stats. In Heard's introduction he states: "...I weighted rushing 87% and receiving 13% (based on rushing making up approximately 87% of the total touches for players in my data set.) " By basing a part of his formula off of a time-specific data set, this could cause some problems, especially for players who don't catch out of the backfield much. It could also undervalue players such as Alvin Kamara and Christian McCaffrey, who receive far more than 13% of their touches through the air. For that reason, we will be avoiding receiving stats altogether and focusing only on running performance.
Breaking Down Passer Rating
As stated previously, we will be using the passer rating formula as our template for this exercise. The formula is as follows:
Essentially, there are four separate equations evaluating three key categories: (1) efficiency, (2) scoring, and (3) turnovers. The four equations mentioned are:
The NFL's passer rating formula seems very "messy" at first, but is much simpler when broken down like this. To make this process as simple as possible, nearly all modifications to the formula will be done in these four equations (you'll see what I mean by that later on).
For any of those who are math savvy, you'll notice that each respective equation can result in some crazy totals. For that reason, the formula automatically places a value cap of 2.375 and value floor of 0. For example, if a player throws a single pass for a touchdown, equation (c) will round that respective value down to 2.375, rather than 20. We will be applying the same boundaries here, for simplicity's sake.
Creating the Formula
For starters, we need to adjust the amount of equations that we will be dealing with in our formula. When passing, there are four major stats that we use: completions, yards, touchdowns, and interceptions. Unfortunately, we only have three major running stats: yards, touchdowns, and fumbles. In essence, there is no statistic that can accurately supplement completion percentage, so we will be excluding equation (a) altogether.
To account for this change, we will simply change the denominator "6" in Figure 1 to "4.5". By doing this, we can still obtain a maximum "runner rating" of 158.3, just like passer rating.
Finding the Averages
Here is where things tend to get tricky. To accurately modify the equations, we will need to establish what an "average" running performance is. When the passer rating formula was created in 1971, statisticians based the formula on passing stats from 1960-1970, giving an "average" value of 1.000 for each respective equation (more on that here).
Through their research, they determined that a perfectly average passing game is 7 yards/attempt, one touchdown per 20 attempts (or 5% touchdown rate), and one interception per 18 attempts (or 5.5% turnover rate). We will be following the same method here, compiling all rushing stats from 1960-1970. Note: NFL and AFL stats have been combined.
Rest at link:
old.reddit.com/r/nfl/comments/boxqwe/oc_attempting_to_create_an_objective_runner_rating/
Also:
Stats required to obtain a perfect runner rating:
7.2 yards per carry
1 touchdown per 13 carries (7.69%)
0 fumbles
7.2 yards per carry
1 touchdown per 13 carries (7.69%)
0 fumbles
... and neither is this Chris guy.