From Bell Curves to Le’Veon Bell: A reason to have faith in the Steelers’ Draft

USA TODAY Sports

When the Steelers drafted a running back in the second round of the 2013 draft, it caught many in Steeler Nation by surprise that they selected Le’Veon Bell. Given how anxious Steeler Nation was for the Steelers to have a successful and productive draft, many questioned the level of competition he faced while at Michigan State; others how much he had left given his NCAA leading number of rushes he achieved his senior year.

And yet after the draft, Steelers GM Kevin Colbert was quoted as describing Bell as an "NFL back", and offensive coordinator Todd Haley called Bell a "three down back".

While the more cynical might attribute Colbert's and Haley's comments as at best post-draft hyperbole, and at worst self-serving rationalization for their choice, we'll only really know after Bell has played a few years.

In recent years Steeler Nation has had reason to have doubts about the drafting efficiency of Colbert and Tomlin, and 2008 certainly validates those doubts (as I demonstrated in this article). It was the research for that linked article that got me wondering whether the commonly lathered praise for the Steelers drafting abilities held up to closer scrutiny. It's one thing while pointing out a Troy Polamalu or Ben Roethlisberger to cavalierly brush off a Limas Sweed or Ricardo Colclough as being aberrations; it's quite another if those players who make up the middle ground between the stars and the busts are in fact mundane and well, average. So I set out to see just where the Steelers' drafting history puts them in relation to the other 31 teams in the NFL.

The metrics used herein are (as in the previous articles) the CarAV and DrAV stats as compiled by Prof-Football-Reference (hereinafter referred to as "PFR"). I'll leave it to you to follow the link and refresh your memory on CarAV and DrAV.

I've added the Yrs (average career length) metric, which is simply the average career length of each draft choice as assigned to the team who initially drafted him, regardless of how long he actually played for that team. This metric is important because there is an obvious correlation between the "quality" of a player as measured by the CarAV and the length of said player's career; a high quality player is far more likely to have a long career (barring injury or other non-football related matter) than a player of poor quality. Conversely, a poor quality player rarely if ever has a long career; it is in defining the difference between "poor" and higher levels of quality (that is beyond the scope of this article) that blurs the line of causation between the quality of player and length of career

The first step is to determine whether averaging the scores for each team's draft choices in terms of CarAV, DrAV and Yrs for each year from 1990 through 2008 would result in a normal distribution curve (see link for basic explanation of Normal Distribution).

For the sake of brevity (hah!) of this article, I will only detail my analysis of the CarAV metric, although the process I've undertaken is identical for DrAV and Yrs.

As you can see from the graph below, yes it does.

Pa1_medium

What this graph shows is that taken collectively, each teams draft class CarAV over 19 years falls within a range of points within which a linear trend line can be derived. Each of the points on this line contain the CarAV scores of each of the 32 teams, so in reality the numbers would be scattered above, below and along the trend line shown.

From here, if you calculate the average of all 608 scores (32 teams' annual draft classes x 19 years of data), you find the "Mean" or single average CarAV score for that period for the entire NFL. Then, you apply the calculations associated with Normal Distribution to create the bell curve shown below. Finally you then derive the standard deviations of all the scores from the Mean; that is the shaded portion below the curve:

Nfl_bell_curve_medium

What this curve illustrates is the fact that collectively, the average CarAV score of each of the 32 teams' draft classes for each of the 19 years analyzed (1990 through 2008) all fall within a determinable range of numbers, and within this range there is a consistent separation of the scores into three distinct segments called Deviations.

The color coding of the standard deviations for the graph above represents the three "deviations" from the overall average CarAV score for the NFL (the "Mean"); red represents the Third deviation ( -3 and +3) which are furthest away from the Mean; yellow represents the Second deviation (-2, +2) and the green represents the First deviation (-1, +1), as shown below.

Std_dev_color_medium

The data being analyzed should adhere to a general rule of Normal Distribution: the "68-95-99.7 Rule"; meaning 68 percent of the data will fall within one standard deviation of the overall Mean (or average); 95 percent within two standard deviations, and 99.7 percent within three deviations.

The "68-95-99.7" breakdown is also known as "Confidence Intervals" meaning that within a data set that complies with Normal Distribution, you can have 68 percent confidence the data will fall within the First standard deviation range, 95 percent confidence that a particular datum point will fall within the First or Second standard deviation ranges, and you can be quite confident (99.7 percent) all the data will fall somewhere in the three standard deviation ranges. Thus in the analysis we see that 22 teams, or 68.8 percent of the NFL had their CarAV score fall within the first deviation (green). No team's score fell within either the upper or lower Third deviation (red) for CarAV, while three teams fell within the lower 95 percent confidence level (negative Second deviation - yellow on left) and seven teams fell within the upper 95 p ercent confidence level (yellow on right). This means the average CarAV score of these seven teams' draft classes are higher than the average of all the teams' Mean by an amount sufficient to push them into the upper echelon of quality drafting teams.

In the table below, I've averaged all 32 teams' individual CarAV Mean for the years being studied to derive an overall collective average NFL CarAV score, or Mean, which is 19.4339, with the three standard deviation scores color coded accordingly.

Std_dev_table_with_teams_medium

Within each Second deviation I've listed the teams that qualified; the teams listed at the top of each column have Means that are closest to second deviation from the overall Mean. The teams at the bottom of each column are furthest away from the Mean. Thus, the Cleveland Browns (CLE) with the lowest overall CarAV for years 1990-2008 are at the bottom of the yellow column on the left; the Indianapolis Colts (IND) averaged the highest cumulative CarAV of all teams' draft picks, as shown at the bottom of the yellow column on the right.

This analysis has been performed for all three metrics: CarAV, DrAV and Yrs, with the following results:

3_metrics_table_medium

From the 3 Metric Table above, we can see that there are eight different teams (Indianapolis Colts, Baltimore Ravens, Green Bay Packers, Steelers, New York Jets, Seattle Seahawks, New York Giants and the New England Patriots) whose ability to draft talented players has been consistent enough over the 19 season period to generate a cumulative score substantially above the league average for each metric while eleven teams (Cleveland Browns, Detroit Lions, Houston Texans, St. Louis Rams, Chicago Bears, Oakland Raiders, Atlanta Falcons, Jacksonville Jaguars, Kansas City Chiefs, Washington Redskins and San Diego Chargers ) consistently drafted players who performed less than the league's average in one or more of the metrics.

In the GIF below, the collective NFL is represented by the solid blue bell curve with its Standard Deviation range represented in the green yellow red band used in the charts and graphs above. The first CarAV bell curves for the bottom three rated teams appear, then the top seven scoring teams; all are overlaid on top of the NFL curve in dashed lines, with the Steelers black vertical dash lines representing its standard deviation shown last.

Top_and_bottom_teams_gif_medium

So, what do these bell curves tell us? First, let's look at the NFL, PIT and IND overlaid. If you recall, the Colts had the best CarAV of all the teams for the period in question, but both the Colts and the Steelers fell within the NFL's second standard deviation (the yellow column in the CarAV measurement in the 3 Metrics Table).

What_the_bell_curve_tells_us_medium

The Colts' standard deviation is represented by the solid red block in the graph above; it is partially coverd by the NFL's green deviation with the Steelers' deviation represented by the vertical black dashed lines.

There is a substantial amount of solid red peeking out to the right of the NFL's green, and a sizable piece extending beyond the Steeler's black. This illustrates the difference between the NFL's Mean of 19.43391, the Steelers' 22.61239 and the Colts' 23.12726; these differences signify that the apex of the Colts' bell curve is more to the right (higher on the X axis) than either of the other two. You will note however that the Colts' curve is "squatter" than the Steelers' or the NFL's; it is lower but broader. The difference in the three curves is a result of the differences in the three standard deviation scores from the Means for each.

The Colt's bell curve looks as it does because of where each of its draft classes falls along the standard deviation line:

Ind_carav_std_dev_table_medium

Now, compare the above to the Steelers:

Pit_carav_std_dev_table_medium

The Colts have a standard deviation (StdDev) of 10.2985, while the Steelers is a smaller 6.2826; in other words, as the IND table illustrates above, the Colts have a "wider spread" of draft class CarAV scores than the Steelers, and thus for its data to keep its bell curve shape, the curve is wider than the Steelers' curve. Note the Third Deviation to the right (Red Right) for both teams; the Steelers and Colts both have one draft class that falls within that Deviation. However, the Steelers have two draft classes (2002, 2003) that exceeds the Third deviation, while the Colts have three draft classes whose CarAV scores exceeded that Third deviation (1998, 1999, 2001).

In 2002, the Steelers drafted Brett Keisel (CarAV 53), Larry Foote (CarAV 56) , Chris Hope (CarAV 44), Antwaan Randle El, and Kendall Simmons (CarAVs 42 and 31 respectively); only three of the eight players drafted failed to produce a double-digit CarAV score. In 2003, only two of the five players drafted produced a double-digit CarAV score, but those two players are Troy Polamalu (CarAV 88) and Ike Taylor (CarAV 49), while the players the Colts drafted in the three years to the right of its Third deviation include such superstars as Reggie Wayne (CarAV 112), Edgerrin James (CarAV 114), and Peyton Manning (CarAV 165).

What causes the Steelers bell curve to assume the narrower, taller shape it holds is the key to their success; it is the preponderance of draft classes whose CarAV performance that fall within the Second deviation; 12 of the Steelers 19 draft classes (63.2 percent) fall within the Second deviation, while only 10 (52.6 percent) of the Colts fall within that range. In other words, the Steelers are more consistent in picking players that produce at a certain level; the Colts have had more draft classes that are above and outside their "norm", but also more that are lower and outside that same "norm".

Consistency; a characteristic reflected not only in the quality of players the Steelers select, but is also reflected in the nature of the organization itself; it has had only three head coaches during the period of this analysis. The Colts have had eight; eight different coaches with eight individual drafting philosophies resulting in eight different drafting criteria, while the Steelers have maintained a consistent philosophy and thus tend to pick the same types of players every year. Look at the five draft class scores for the Colts that fall lower than, or within the lower Third deviation, and compare those scores to the four the Steelers had; there is only one Steelers draft class that scored lower than any of the Colts lowest scores (the infamous 2008 Steelers draft class).

It is this consistency that lends validity to my inclusion of Yrs (length of career) as a metric; because the Steelers consistently draft a preponderance of players whose scores within the CarAV metric fall within a higher and narrower band than the rest of the NFL, the characteristics that attract the Steelers to those players also result in more of those players playing longer; the DrAV score for the Steelers, ranking in the Third deviation as compared to the rest of the league in the table above indicates the Steelers select players over all who not just play longer than average, but stay on the team that drafted them longer. There have been notable exceptions, such as Mike Vrabel who couldn't crack the Steelers starting lineup after being drafted in 1997 and was released in 2000, only to go onto a stellar career with the New England Patriots. But those exceptions are few and far between.

But we're not done yet.

Since we're dealing with three separate Means, but trying to ascertain from all this analysis what team has been the most successful in terms of draft quality over the past 19 years, we need to standardize the data across the three differing Means and deviations; to do that, we'll look at the Z scores all of this data has produced.

Z scores determine how far a particular number is from the group Mean in terms of the group's standard deviation. In layman's terms for this analysis, each team's Z score for a given metric indicates how superior or inferior its score is for that metric than the average of the group of 32 teams as a whole, based in units of the group's standard deviation. The team with the highest Z-score ranks as the best team in terms of performance of the metric. As a test on the Z-score calculation, if you add up all of the individual Z-scores of a sample set, it should total zero.

Below are the tables showing the Z-scores for individual teams ranked highest (best) to lowest for each of the three metrics, color coded to show where on the bell curve they would fall:

Z_scores_table_medium

The Steelers rank fourth for CarAV and second for both DrAV and Yrs. The Steelers' score of 1.49 indicates that the Steelers' CarAV score was 49 percent higher than the Mean for all the teams combined; their DrAV score of 2.16 indicates the Steelers scored 216 percent higher than the league as a whole, and 212 percent higher than the league for average career length (Yrs.).

If you total up each team's Z-score for the three metrics, the overall standings look like this:

Final_combined_z_score_rankings_medium

Thus, while Steeler Nation has long prided itself with having what it considers the best drafting team in the league (frankly, the best overall team period), hopefully this article helps provide empirical evidence to support that belief.

While not No. 1 in any of the three metrics used in this analysis (CarAV, DrAV and Yrs), the Steelers consistently ranked near the top in all three; it is this consistency in finding high achieving players year after year, getting the most out of them while they were members of the Pittsburgh Steelers, and selecting players that had durability and a propensity for long careers, that makes the Steelers the league's premier discoverer, nurturer and developer of talent in the NFL. Of course, given the rise of Baltimore Ravens, especially since John Harbaugh took over as HC under General Manager Ozzie Newsome, it is also understandable why the Ravens come in a close second to the Steelers.

Le'Veon Bell was not viewed by many in Steeler Nation as the "right" pick at No. 2 in this year's draft, and the rookie running back faces a steep learning curve to reach a level of performance expected from a second round pick, but the Steelers' bell curves should illustrate to you that with the Bell pick, in the Steelers' front office we should trust.

X
Log In Sign Up

forgot?
Log In Sign Up

Forgot password?

We'll email you a reset link.

If you signed up using a 3rd party account like Facebook or Twitter, please login with it instead.

Forgot password?

Try another email?

Almost done,

Join Behind the Steel Curtain

You must be a member of Behind the Steel Curtain to participate.

We have our own Community Guidelines at Behind the Steel Curtain. You should read them.

Join Behind the Steel Curtain

You must be a member of Behind the Steel Curtain to participate.

We have our own Community Guidelines at Behind the Steel Curtain. You should read them.

Spinner

Authenticating

Great!

Choose an available username to complete sign up.

In order to provide our users with a better overall experience, we ask for more information from Facebook when using it to login so that we can learn more about our audience and provide you with the best possible experience. We do not store specific user data and the sharing of it is not required to login with Facebook.

tracking_pixel_9341_tracker