Offense or Defense: What Do You Draft?

Quarterbacks always have the inside track when it comes to being selected at the top of the draft board. But that aside, do NFL teams favor one side of the ball or the other when drafting? Do teams prefer to load up at a position early in the draft and then ignore that position later in the draft, or vice versa? This article addresses those issues through the examination of the 2005 through 2014 drafts.
Offense vs. Defense
While there might be short-term fluctuations due to supply or demand at a position, over time it is reasonable to expect that number of offense players and defensive players drafted should be about even. When looking at the 2005-2014 drafts that does turn out to be the case. Of the 2501 non-kickers drafted, our count is that 1245 were offensive players and 1256 were defensive players. That’s about as even as you can get.
There are more significant variances on a team-by-team basis. Certain teams, even over a 10-year period, do show at least some indication of bias on one side of the ball or the other. A short-term bias is certainly understandable as a team looks to plug holes wherever they exist. Over a 10-year period, though, all that should even out and it may come to a team’s drafting strategy or pure chance.
The following table shows the percentage of offensive draftees for each NFL team for 2005-2014.
The ends of the spectrum are the Jets taking offensive players with 58% of its selections and, on the other end, the Falcons taking defensive players with 60% of its choices. Is the variance from the average a matter of a team’s strategy or is it just random based on a team’s draft board and the players available? It is impossible to say without being in the draft room or being part of a team’s management.
The first three rounds of the draft produce most of the starters in a draft class. Over those three rounds, the numbers historically lean slightly toward defense, but it is still a relatively even split with about 49% of the draftees being offensive players versus 51% on defense. The team distribution changes, though, indicating that some teams favor one side of the ball in the first three rounds and then the other side in the final four rounds. Here is the chart as previously shown, but for the first three rounds only and with the scale slightly changed to accommodate the 49%/51% split.
The extremes for the first three rounds are the 49ers with 59% of its selections from the offensive side and the Saints with defensive players making up 62% of its selections. The same question as above is still applicable regarding whether this is a planned strategy or just a matter of chance.
The differences by team can be seen more clearly in the next table. This table shows the percentage of offensive draftees by playing position and team for round 1-3, rounds 4-7 and overall. Defensive draftees are, of course, 100% minus the percentage of offensive draftees.

Bias by Playing Position
 Within the offensive and defensive splits presented above, there are also biases by playing position. Before looking at the information by team here is a breakdown by playing position for the first three rounds, the last four rounds and all rounds. The percentages represent the portion of all drafted players in each grouping from 2005-2014.

This table shows that quarterbacks, wide receivers, defensive linemen and corners account for 50% of draftees in the first three rounds but only 44% of players drafted in rounds 4-7. This indicates a bias towards drafting those positions in the top three rounds.
To get a better feel for the teams that are most and least likely to draft players at those four positions, the following tables show the distribution of players drafted by playing position and NFL team.
The first table shows the quarterbacks drafted in the first three rounds. The Browns, still searching for a quarterback, had the most with five. Six teams did not draft a quarterback in the first three rounds. The Texans are one of those six, though they are not settled at the position.

The next table shows the distribution by team for wide receivers drafted in the first three rounds. The Giants and Titans are the leaders with eight and seven, respectively. The Titans are far from settled at the position. Five teams drafted two or fewer receivers in the first three rounds. With the exception of the Cowboys, none of the teams are well set at the position.

The next table shows the distribution for defensive linemen. The Eagles had the most with 10 and the Redskins the least with one.

And finally, the distribution for corner backs is shown in the next table. The Rams led with eight corners drafted in the first three rounds while the Eagles had only one.

This addressed only one aspect of the positional bias issue. A “shortcut” way of looking at the possible existence of a bias is to find situations where the number of players drafted at a position is somewhat greater (or less) than 1.5 times the number of draftees for the first three rounds. The 1.5 factor is based on averages by position as discovered in this study.
The logic would be applied as follows:

  • The 49ers selected eight offensive linemen in the first three rounds and six offensive linemen in rounds 4-7.
  • This indicates a bias for selecting offensive linemen earlier rather than later as using the 1.5 factor they would be expected to have drafted 12 lineman in rounds 4-7, or double the number drafted.
  • On the other hand they selected two corners in rounds 1-3 and nine corners in rounds 4-7.
  • This would indicate that they believe they can find corners later in the draft and do not need to draft them early.
  • The expectation would be that three corners would have been selected in the later rounds, and not the nine actually selected.

The following table shows selected instances of positional biases for each NFL team and the number of players drafted in rounds 1-3 and then in round 4-7.. The number of instances was capped at three. The column labeled “bias” indicates whether the bias was in favor of drafting a position early (like the 49ers offensive linemen) or in favor of drafting a position later. 1-3 indicates that the bias is toward drafting early. 4-7 means that the bias is toward drafting later.

Follow Tony on Twitter @draftmetrics

Rankings of NFL Teams by Draft Class

For all the talk about the draft and teams that do well, there is one point that must be stressed. No team does well in the draft every year. Taking that statement to the next logical step, a team does not have to do well in the draft every year to be successful.
To study this premise and associated issues, we reviewed drafts from 2005 through 2012. The drafts from 2013 and 2014 were omitted because it is too soon to make even a preliminary judgment on those. Before discussing the results we should discuss the key elements of the study.
First, contrary to most of the other studies we have done, only a player’s career with the team that drafted him is considered. For example, the Jets do not receive credit for the years Darrelle Revis spent with the Bucs and the Patriots.
Second, despite considerable thought on the matter, we could not come up with a formula or measure that would yield a meaningful comparative rating for the teams, so the ratings are subjective based on several factors. These factors include:

  • The number of two-year starters in a draft class
    • The cut-off of two years was used so that we could include as many current drafts as possible
  • The number of games started by a player over the course of their career with the team that drafted him
  • The number of Pro Bowl selections
  • The number of All Pro selections

It is acknowledged that there is an inherent advantage in being drafted by a “bad” team as the path to a starting job is likely to be an easier one, but this was not considered in the rankings. Draft position and number of choices were also not considered, so the result is an absolute and not relative grade.
Based on these factors the drafts of each team for each draft class were analyzed with the teams ranked one through 32. The full rankings are shown at the end of this article. It is recognized that the rankings for at least some of the years might change over time as careers ebb and flow. One would expect, for example, that more post-season honors are likely to be won in the future by some of the more recent draftees.
Perhaps the most striking results from the study is that no team was ranked #1 more than once. Similarly, no team was ranked last more than once either. While other factors may be at play, this supports the variability of draft results and the presence of the “luck factor” in the draft process. The same management team following the same process can garner different results in different years.
Seven teams were ranked in the top 10 in at least half of the eight drafts. Those teams, with its ranking based on its won-lost record for 2005-2014 in parentheses, are:

Several things stand out in the analysis:

  • The Packers were actually ranked in the top 10 in five of the first six years of the study before earning lower rankings in 2011 and 2012
    • The Packers have been in either the top or bottom 10 in each of the eight drafts
  • The Seahawks have finished in the top three in the 2010-2012 drafts
    • They are the only team with three top three grades
  • The 49ers were ranked in the top 10 in each draft from 2005-2007 but only once since then
  • Neither the Chargers nor the Raiders had even one top 10 grade
    • The Raiders had only one bottom ten grade as well
  • The Patriots, with the top won-lost record for 2005-2014 by a comfortable margin, were in the middle of the pack as far as draft grades with three top 10 rankings and three rankings in the bottom 10

Seven teams finished in the bottom ten at least half the time. These teams were:

A few notes about the teams finishing at or near the bottom:

  • The Saints have finished in the bottom 10 for each of the last four drafts
  • The Jaguars and Giants have finished in the bottom 10 for each of the last three drafts
  • In addition to its four bottom 10 grades, the Bears have ranked 11th from the bottom on two other occasions
  • Every team but the Titans have finished in the bottom ten at least once

Here are the complete annual rankings for the 2005-2012 drafts:

Follow Tony on Twitter @draftmetrics

Which NFL Teams Draft the Best?

Evaluating the draft performance of an NFL team is difficult at best. The ultimate test of a team occurs on the field, of course, and is a mixture of many elements, with draft performance being just one. There are a number of issues when trying to isolate draft performance. Teams do not have the same number of selections, drafted players are moving into different playing situations, and so forth.
In this article, the number of starts from 2005 through 2014 was used as the most reasonable metric for measuring draft performance. This is not perfect but allows a window into draft performance. Starts for a player are credited to the team that drafted him and includes games started for either the drafting team or a subsequent team. Drafts for the past ten years (2005-2014) were studied to see what conclusions could be drawn.
Each NFL team starts with 70 selections for the 10-year period (10 years and one selection by round). Adjustments are then made for selections lost due to penalty or used in the supplemental draft, compensatory selections awarded and the net number of choices gained or lost through trades (mostly trade-up or trade-down transactions). Here is a summary by team, sorted by total number of selections between 2005 and 2014:

There is quite a range in the number of selections with the Saints having the fewest with 60 and the Packers the most with 96. All draft selections are not created equal, however, as a first round choice is worth considerably more than a seventh round selection. It is fair to say, in my opinion, that no one views the absolute number of starts as being fully indicative of a team’s expectations.
A measure that does a better job of weighting the draft choices is the number of starts a team could historically expect, given their number of selections and the location of their selections. This sets the expectation but does not provide any information about actual results. The following ranks the NFL teams by the number of expected starts for players drafted from 2005 through 2014.

The revised rankings push the Rams into the position of being the team that should have expected the most help from the draft over the past 10 years, with the Packers dropping from first to second. There is little change at both the top and the bottom of the expectations between the first and second tables.
We next ranked the teams by the actual number of starts by players drafted by each team in the 10-year study period. This gives great weight to the total number of selections by a team, particularly early round selections. The following table shows those rankings:

As can be seen from reviewing the previous two tables, the difference between actual and expected performance can be rather significant. The Ravens, for example, are ranked fifth in actual starts compared to 18th in expected starts by players selected in the study period.
The above table still places, in my opinion, too much weight on the actual number of draft choices. What I believe to be preferable is to rank the teams by number of actual starts as a percentage of expected starts. The following table shows the percentage by which a team’s actual performance exceeds or lags expected performance. Players drafted by the Seahawks, for example, had almost 19% more games started (1785 divided by 1505) than a historical analysis would suggest. A number in parentheses indicates that a team did not achieve expectations.

It is interesting that this year’s Super Bowl participants (Seahawks and Patriots) are at the opposite end of the ratings. This reinforces the notion that there is more than one way to achieve NFL success.
The Seahawks achieved most of its success in rounds two through four, while being close to average in other rounds. The Patriots did better than expected in round one but were ahead of only the Lions in second round performance. The first round is the only round where the Patriots exceeded expectations by more than a nominal amount. While the Patriots did draft some very fine players in round two (like Rob Gronkowski, Sebastian Vollmer and Jamie Collins), they also had their share of flops and players approaching flop status in Ras-I Dowling, Darius Butler, Terrence Wheatley, Ron Brace, Tavon Wilson and Chad Jackson.
The best and worst performances by draft round are shown in the following table. All situations where actual was better or worse than expected by at least 100 starts are included. The table indicates that the Colts’ sixth-round selections, in the aggregate for the ten-year study period, exceeded expectations by 264 starts.
Follow Tony on Twitter @draftmetrics

Comparing Draft Years

Every year experts offer up their opinions on the upcoming draft, both on the overall depth of the class and the quality and depth by playing position. In this article we will explore how real these differences are and the degree of difference among draft years.
Year-to-Year Comparison
What is the best way to measure the differences? There is no perfect metric, but the most practical performance metric is the number of games started. Number of starts has the principal advantage of letting us use recent years in the analysis whereas other measures (such as number of five-year starters) required us to cut off the analysis well before the most recent drafts.
There is an inherent conflict between using current information versus complete information. It takes 15 years or so to “complete the book” on a draft class. This would mean that the most recent year that could be used in this analysis is 1999. That serves as a sound basis for the analysis but many readers would regard the information as stale and, therefore, not relevant.
It was decided to compare draft classes after their first, third, fifth and seventh years. The draft years used were the most recent 10 drafts for which the number of starts data was available as follows:

The following table summarizes the data by draft class for each of the four time periods. The “ratio” row represents the ratio of the number of starts for best draft compared to the worst draft. A high number shows a lot of variation and a low number shows little variability.

This table shows that the degree of difference, as expressed by the ratio, is greatest for the first year and levels off after that. The 2006 draft year has held up best over time. The draft class in each of the last four years has exceeded the number of first-year starts from 2006 but both 2011 and 2012 have fallen behind 2006 by the end of the third year and it is too soon to tell for 2013 and 2014. It remains to be seen if the advantage of the 2006 draft class will hold up.
The 2008 draft class is also an interesting case. That class is ranked last in first year starts but moves up to the middle of the pack for the subsequent measurement periods. This reinforces the notion that the first year does not tell anywhere near the whole story when it comes to evaluating draft classes.
This analysis shows that there is a reasonably large difference among draft years. This makes characterizing draft classes as good and bad a logical exercise.
As an aside, data from the draft classes that are virtually complete as far as number of starters (1995-1999) is instructive regarding the evolution of the impact of a draft class. The cumulative average percentage of total starts by year for the five draft classes is shown in the following table. The table tells us that 91% of the starts from a draft class occur in the first 10 years.
Playing Position Comparison
It should be no surprise that the level of variability increases by playing position. There are fewer data points and this leads to greater variances in results. The analysis by playing position was confined to the most recent ten draft classes and excluded fullbacks and special teams players. Again, the number of starts was used as the performance metric.
The following table shows the best and worst classes for each playing position after one season by draft class.

This table shows that there significant differences among playing positions with the offensive line and defensive backs showing the smallest, though still significant, differences.
The next table shows the same information after three years for each playing position by draft class (which excludes the 2013 and 2014 draft classes).

This table shows that the ratios have tightened significantly by the third year but that they still exist. Offensive linemen show the smallest difference between the best and worst year with running backs having the largest difference.
It is interesting to note that while 2006 is a highly rated draft class, it does not appear as the best year for any playing position.
Follow Tony on Twitter @draftmetrics

The Rest of the Equation

The article “Draft Probabilities by Playing Position” laid out the probabilities of achieving various milestones for players selected in the NFL draft. What was missing from that discussion, though, is the time element. That is, how long does it take a drafted player to become a starter or achieve another relevant milestone?
When a team signs a free agent there is a general expectation, unless a special team player is signed, that he will move right in and be helpful. For drafted players the expectations are generally lower in terms of providing immediate help. But what exactly should those expectations be?
Three milestones were used to measure the time element for the purposes of this article. The milestones were the length of time used to become a one-year, three-year and five-year starter. Our definition of a starter is a player who starts at least eight games in a season. The averages include only the players who achieved those milestones. For example, take the case of a player who started for four years. He would be included in the one-year and three-year calculations but would be excluded from the five-year calculation.
The best “score” a player can earn is 1.0, indicating that he became a starter in his rookie season. He would receive a 2.0 if the becomes a starter in his second season. There are no partial seasons awarded, but the averages do show partial seasons (kind of like the Census Bureau saying there are 2.58 people per household).
The logical expectation is that players drafted earlier will start faster (and more often as indicted in the “Draft Probabilities by Playing Position” article). Whether this is because the players are better or because they get more chances due to being highly paid is a matter of debate. The truth is probably that it is a combination of the two.
The following table shows averages by playing position and Draft Range for the one-year milestone and indicates that about 42 players per season start as rookies. It is probably not surprising that running backs take the longest to start for any playing position among early round draftees. For the group a whole (as indicated by the “All” column), though, it is defensive linemen that take the longest to start their first season. The average for the rest of the positions are pretty close.
The table also confirms that the earlier a player is drafted, the sooner he will start. Players selected with the first four picks start much sooner than a player taken later in the draft.

The next table shows the averages for achieving the three-year milestone. Running backs again take the longest to start among the early round selections. This holds true through the 4th Draft Range (which is through the 46th pick). Overall, defensive linemen tend to take longer than average to achieve the three-year milestone. Offensive tackles and linebackers are at the other end of the spectrum and reach the milestone sooner than average. The amount of time to start for a third season ranges from 4.0 to 4.8 seasons.

The final milestone reviewed was the amount of time it takes a player to become a five-year starter. It is interesting that the range tends to narrow at this point because only the better players are likely to achieve this milestone and be included in the calculation of the average. The range is from 6.0-6.7 seasons, slightly narrower than the range by position for achieving the three-year milestone. Among the earliest draft selections (first 14 picks) Running Backs and Quarterbacks tend to take the longest to reach the milestone. Offensive Tackles drafted in that same range are the fastest to reach the five-year milestone. Overall, Guards and Offensive Tackles are the fastest to reach the milestone, while Defensive Ends and Quarterbacks are the slowest.

Click here for the complete breakdown of “Draft Ranges”
Follow Tony on Twitter @draftmetrics

Time to Get Rid of NFL Compensatory Selections

Accompanied by the customary media hype, the NFL recently announced the awarding of 32 compensatory draft selections. The determination of the selections to be awarded has evolved into a science. A handful of website and bloggers have “cracked the code” and do an excellent good job of projecting the compensatory selections.
I am not a fan of the current compensatory selection process. It is my view the selections provide minimal recompense for the losses a team suffers during free agency. Either there should be no compensation at all or the compensation should be fair. Instead, the NFL has taken a “something is better than nothing” approach.
Supporters of the current system can cite a list of compensatory selections that have gone on to be successful. In my opinion, this is strictly anecdotal information and there is no reason a compensatory pick should be more or less successful than the rest of the selections that surround them. As will be supported later in article by hard data there is no reason to expect that the compensatory selections will yield more than two or three starters each year.
Is there a better way to do it? I think that maybe there is. This article is intended to encourage discussion of the process without regard to the feasibility of actually pushing through a change. I acknowledge that the NFL and NFLPA have negotiated the matter of compensatory selections so it would not be easy to change. Feasibility will be discussed in the course of the article.
Before getting into that, let us review the process in more detail so that the proposed change is considered in the appropriate context. Compliments of the NFL, here are the players lost and signed by the 14 NFL teams that have been awarded compensatory picks.

The rules governing compensatory picks are impossible to find in writing but the aforementioned cap “scientists” have discerned the most important elements based on historical NFL awards. We do know that the number of picks is limited to 32, with no more than four awarded to any single team. What are the 32 compensatory selections likely to yield? Based on 20 years of history and considering their location in the draft, this year’s compensatory selections can be expected to yield the following:

  • 26 players who will play in the NFL for at least one season
  • 11 players who will last at least five seasons in the NFL
  • Six players who will start for at least two NFL seasons
  • Between two and three players who will start for at least five NFL seasons
  • One player who will be selected to at least one Pro Bowl

The value of these selections is further diminished by the fact that the compensatory selections cannot be traded. This means they cannot be used as a throw-in that might facilitate a trade. I am not sure what the trade prohibition is intended to accomplish.
The Chiefs, with four selections between rounds three and six, received the most value in 2015 of any of the NFL teams. History tells us there is less than a 50% chance that the Chiefs’ compensatory picks will yield even one five-year starter. So there should be no illusions that the teams are receiving even moderate value in exchange for their free agent losses.
I would argue that a superior option, and one that does a better job of contributing to league parity, is to adjust a team’s cap through a “luxury tax” based on free agent signings. Let us look at the Chiefs to see how it would work. Here is a summary of free agents lost and signed by the Chiefs, with dollar values representing the 2014 cap for each player as published by

For each player signed or lost, the contract amount would be multiplied by a percentage (say 17.5%, which is the Major League Baseball luxury tax rate) and the resulting product would be transferred from one team’s next annual cap number to another team. The Chiefs would receive a annual net cap increase of almost $2.4 million. The tax rate could be higher or lower depending on how much parity the league wants to encourage and based on negotiations.
I think there are several reasons why my proposal is better than the current system:

  • Perhaps most importantly, the proposed process provides greater flexibility in that the additional cap space could be used to either sign one or more free agents that could provide immediate help or allow a team to retain a free agent it may otherwise lose
  • The proposed process established a direct relationship between the value of a lost player and the compensation received
  • All teams affected by free agent gains or losses would be included
    • This is not true of the current system where the 32-choice limit leaves some teams with a net loss of free agents and no compensation

How would the owners, the union (“NFLPA”) and Roger Goodell react to such a proposed change? It is only fair to speculate that, unless it is a sold as “its good for the game” by Czar Roger, both the owners and the NFLPA are likely to be resistant to making a change. I have heard no displeasure voiced against the present system and without a “champion” any change is unlikely.
While there should be no change in overall spending (it is just a matter of the cap dollars changing pockets), the NFLPA may see it as having the potential to restrict player movement due to the luxury tax component. Just by the nature of the bargaining process, the NFLPA would want something in return for making even a neutral change.
It takes the vote of 24 owners to change an NFL rule. Whether that level of commitment is obtainable depends on the balance between teams that are buyers or sellers of free agents. The sellers would not like this change. The current system allows them to pursue free agents without having to surrender anything other than the compensation paid to the player they signed. The proposed process introduces an additional cost of signing free agents. For a team that essentially sits out free agency, I would think they would support this change as it provides value that could be greater and maybe more fair than the current system when free agents are lost. It is hard for me to imagine that 24 positive votes could be garnered.
The bottom line is that, my protestations to the contrary, we are stuck with the present system. There is certainly no discussion I have seen that indicates otherwise.
Follow Tony of Twitter @draftmetrics

NFL Draft: Is it really that simple?

In its recent Analytics Issue, ESPN The Magazine had the following to say about the NFL Draft:

“NFL Draft math is simple: Stockpile picks for more shots at a jackpot.”

But is the strategy really that simple? Of course a team would like to turn unwanted or unneeded players into draft choices and a team with a surplus of selections would like to turn them into earlier and more selections. But beyond such cases there is considerable debate, both about the strategy and ESPN’s meaning. For example, did they really mean to say stockpile high draft choices or all draft picks?
In short, I do not agree that formulating the right draft approach is all that simple. The key is to be opportunistic and to always make the higher value decision. Also, the probabilities play such a large role that it is also difficult to ignore the “luck factor” in a team’s success. Even give the luck issue, though, it is the ability to evaluate talent that is the differentiator among teams.
This article will explore various aspects of draft strategy. We will keep it simple and do the analysis by draft round rather than the Draft Ranges we prefer. Since it is also highly likely that the answers will differ depending on the metrics used, we will provide multiple answers as appropriate.
Kickers are excluded from the analysis for comparability reasons. The analysis was conducted for a 10-year period and includes players drafted from 2005 through 2014.
What Is a First Round Pick Really Worth?
First, let us take ESPN’s statement to the extreme. What if you had a first round selection and wanted to get the maximum possible number of picks in exchange. How many later picks could you possibly acquire? This article considers only historical trade information and individual trades that are straightforward (e.g., no trades where a first plus another pick is traded cases when a player is involved and trades involving selections for the subsequent year, etc.).

Here is what you might acquire in our theoretical exercise:

So a first round selection can, in theory, be turned into ten late round picks. (We are not commenting on the feasibility of actually being able to execute the required trades.). Now let us compare the value of a first round selection with the ten late round picks using various metrics. The next table shows the number of players that should achieve each metric. The analysis shows, for example, that there is an 87% chance that a first round draft choice will have a five-year or longer career, so a single first- round choice will yield 0.87 players who play five years or more.

This shows that the multiple selections yield more players who may contribute on a limited basis but less in terms of players who should make an impact (defined as a five-year starter or a player who earns post-season honors at least once). Our definition of Pro Bowl selection is stricter than most, and is limited to players who are original selections, not alternates or injury replacements.
We next looked to determine whether there is a “sweet spot” along the way to acquiring the maximum of selections that would yield the best result. The following table shows the comparative metrics after each of the theoretical exchanges.

This shows that there is little difference when comparing each step along the way to the ten selections and, therefore, there is no sweet spot.
Accumulate Early Picks vs. All Picks
We next looked at the crux of the ESPN statement – – more choices lead to more success. The first table shows the number of total selections by team and the aggregate number of wins by each quartile.
The first table shows the number of selections for all rounds by team:

The win totals and a review of the table show that there is no apparent correlation between the number of overall selections and on-field success.
The next table shows the number of selections for the first three rounds by team:

This table does show somewhat of a correlation between the number of selections and the number of wins. This supports our theory that loading up on early selections may be the best strategy.
The first three rounds of the draft provide most of the NFL talent:

  • 80% of All Pro and Pro Bowl selections are from the first three rounds
  • 50% of All Pro and Pro Bowl selections come from the first round
  • Setting the bar lower, nearly 70% of games started come from players drafted in the first three rounds

Despite the apparent correlation there are winners and losers at both end of the draft selections spectrum. A few highlights are as follows:

  • 31 extra third-round selections were handed out as compensatory picks
    • Four picks were lost by penalty (Broncos, Patriots, Saints (2) )
    • Four selections (Redskins picking Jeremy Jarmon, Browns selecting Josh Gordon, 49ers picking Ahmad Brooks and Raiders taking Terrelle Pryor) were used in the supplemental draft
  • Despite losing a 1st round pick as a penalty for Spygate, the Patriots are tied with the Rams as having the most selections
  • Three of the Rams extra selections come from the RG3 trade
  • The Patriots extra selections came from trading players for selections (Deion Branch, Mike Vrabel, Matt Cassel) and trading down and accumulating extra picks
  • The Patriots record is not spotless, though
    • They traded up to take WR Chad Jackson while Greg Jennings was selected at the position they traded out of
    • Other players like Carl Nicks, Joe Staley, Clay Matthews and Darryl Washington slipped through their hands in trade-down transactions
  • The Seahawks lost six draft choices in the first three rounds through the acquisition of Deion Branch, Nate Burleson, John Carlson, Percy Harvin and Charlie Whitehurst
  • The Saints lost two second-round selections through Bountygate and did not receive any compensatory picks
    • They used high selections to move up and take Jamaal Brown and Jahri Evans
    • The Saints ranked last in number of picks in rounds 4 through 7
  • Besides the RG3 trade (which may or may not work out), the Redskins wasted high picks on Jason Taylor (waived after one season) and T J Duckett (38 carries for the Redskins)

Should Team Focus on Acquiring Fewer but Higher Draft Choice?
 Another strategy worth considering is trading late round picks to have fewer but earlier selections. The Ricky Williams trade in 1999 epitomizes this strategy as the Saints traded their 1999 1st, 3rd, 4th, 5th, 6th and 7th round picks and their 1st and 3rd round picks in 2000 to move up seven spots and take Williams in a trade with the Redskins
How did that work out? Williams lasted three seasons with the Saints and rushed for over 3000 yards and went on to play 12 seasons in his NFL career and rush for 10,000 yards. He was certainly a legitimately good NFL running back. The Redskins, though, landed Lavar Arrington and used a number of the draft selections acquired to facilitate trades for Champ Bailey and Jon Jansen.
Two tables have been compiled to show the minimum cost of moving up to the preceding round. This was done for both five-year starters and players earning at least one All Pro selection. The cost of moving into the first round using post-season honors is prohibitive but it is included anyway.
The following table shows the minimum cost to move up a round while using five-year starters as the metric. The rows in the table show the round a team is moving to. The columns show the picks that must be relinquished by round. An “X” indicates that a pick in that round is not being traded. The number in the column shows the number of selections surrendered in that round. For example, the table indicates that it would take a 2nd, 4th and 7th pick to move from the 2nd round to the 1st round.

The next table shows the cost of moving up a round based on achieving the metric of earning Pro Bowl honors at least once.

The one oddity in the above table is that it costs more to move from the 4th round to the 3rd round than it does to move from the 3rd round to the 2nd round. This is due to the 4th round having a higher percentage of one-time Pro Bowl players than the 3rd round. This is the only round where this occurs. It is hard to say that this is a better approach than adding draft choices. It all comes down to what the market will pay and correctly analyzing the respective opportunities.
Follow Tony on Twitter @draftmetrics

One week later: NFL Free Agency

The article “What Can You Expect from Free Agency” discussed, among other things, early free agent signings and how there would be plenty of them once 2015 free agency started. Well, the first week is now behind us and there was indeed plenty of activity. This article reviews the first week of free agency and focuses on the activity that took place through March 16.
By our count, 119 free agent signings took place as of the end of the day on March 16. This includes signing both free agents whose contracts have expired and players who have been cut for salary or performance reasons. Traded players and players claimed via waiver were not included. Resigning a team’s own players is also excluded for the purposes of this analysis. While this is important it strays from the purpose of the article.
Contract Length
This is just the beginning of free agency, of course, but the early trend seems to be that fewer one- year contracts are being signed. Some of the one-year deals are for players returning from injury (e.g., Adrian Clayborn, Henry Melton, Sean Weatherspoon, Tyvon Branch, etc.,).
The following table compares the 119 signings so far in 2015 with the 139 contracts signed during the first three weeks of the 2014 free agency process.
With possible exception of the Scott Chandler signing by the Patriots, it is our guess that all the contracts of unknown length are one-year contract. This would mean that so far in 2015 free agency slightly less than 25% of all signings were for one year versus 36% in 2014. As free agency progresses we are likely to see more veteran-minimum, one-year contracts so this gap should be closed somewhat.
The Largest Contracts
The Jaguars have been the preeminent team in signing 2015 free agents to longer contracts. They have signed seven players so far with four those receiving contracts of five years. This represents 25% of all free agents who have signed contracts of five years or longer. A fifth player was signed by Jacksonville to a four-year contract. Beside the Jaguars, only the Eagles have signed more than one player to a contract in excess of four years, and they signed two.
If all five contracts are combined, the Jaguars committed to contracts having a Guaranteed Value of $77 million. Only the Jets exceed that investment with the contracts of Darrelle Revis, James Carpenter, Marcus Gilchrist, Antonio Cromartie and Buster Skrine having a Guaranteed Value of nearly $90 million.
Most would agree that, with the ability of NFL teams to release players and void contracts, a contract’s Guaranteed Value is its most important element. Here are the eleven free agent signings with a Guaranteed Value of at least $15 million. Information is from a variety of published sources.

The Suh contract pretty much drawfs the other contracts signed and makes him the highest paid non- quarterback in the league. By way comparison, JJ Watt signed a six-year contract extension last fall and received a Guaranteed Value of $21 million (total contract of about $100 milllion), only one-third of Suh’s guarantee. Seven of the eleven contracts are for defensive players. It is a matter of conjecture whether this is a matter of chance or a commentary on the importance of defense.
Future Free Agent Signings
More players will be added to the free agent pool as the year progresses and teams make further cuts, most of which will be salary cap related. A remaining key date for free agency is June 1. Many teams have an incentive to make cuts after June 1, as that will permit them to spread a terminated player’s salary cap impact over 2015 and 2016. Otherwise the entire amount of a player’s “dead money” goes against a team’s 2015 salary cap, something most teams would like to avoid.
In addition, while many of the big-name players are already signed, there are still a lot of talented free agents available. By our count, over 250 free agents remain unsigned. Some will retire and others will attract no interest, but most will end up in someone’s training camp.
How Are Teams Doing So Far?
The emphasis should be on “so far” as there is a still a long way to go in the free agency process. In assessing a team’s performance it was decided to focus on participation losses. If player A is lost in free agency and had participated in 1000 plays from scrimmage in 2014, his team must find a replacement to play those 1000 snaps.
That replacement can come from the prior year’s roster, the draft or a free agent. In this analysis only free agency is considered so, if a player or players are signed in free agency to replace Player A, his team has suffered no quantitative loss in free agency.
A team that brings in more experience than they lost has a net gain through free agency. The following table summarizes the net gain or loss through free agency for each team. A number in parentheses represents a net loss. The “Added” column represents the number of 2014 scrimmage plays for signed free agents. The “Lost” column represents the number of 2014 scrimmage plays for players lost in free agency.

The Jets and Raiders have added the most while the Eagles and Packers have lost the most. The Packers do not typically chase free agents and others, like the Steelers, tend to sit out free agency until the prices come down.
As a matter of perspective, the Eagles lost the most 2014 scrimmage plays (4637), but those plays represented less than 20% of their total scrimmage plays. So the impact of free agency is relatively modest.
Follow Tony on Twitter @draftmetrics

Ranking the Draft Performance of Colleges

In the recent article “The Power Five Conferences and the NFL Draft” we promised to provide information, at a later date, for individual colleges. This article follows up on that promise and evaluates the draft performance of colleges for players selected between 2005 and 2014.
The methodology used to rank the colleges is the same as that used in “Rating General Managers in the NFL Draft”. This methodology compares actual results, expressed as the number of games started, with expected results based on historical outcomes.
As always, we must acknowledge that this is just one way (out of many) of viewing a college’s draft performance. We continue to search for the “perfect” methodology but suspect that it does not exist.
Repeating the example used in “Rating General Managers in the NFL Draft”, Frank Gore illustrates the concept. Gore was drafted by San Francisco as the #65 selection in the 2005 draft. This means that he could have started a maximum of 160 (10 years times 16 games/year) games. Historically, players drafted at that point of the draft started about 36% of the maximum or about 58 games for a player drafted in 2005. The actual number of games started by Gore is 134, so he exceeded expectations by 76 games. This is referred to in this article as his surplus. If a player has fewer games started than expected it is referred to as a deficit.
A college’s surplus or deficit is then determined by summarizing each player’s results. The number of draft selections for each college is also counted. The surplus or deficit is divided by the number of draft selections for each college to arrive at its surplus or deficit per draft selection.
As compared with the analysis for General Managers, the data points for each college are widely dispersed. General Managers generally have a full set of draft selections, or close to it, for each year they are employed. This does happen for many of the colleges. Thus the necessity to convert all data to a per draft choice basis as this facilitates analysis and provides comparability.
The results or our analysis are quite different than they would be if only the absolute number of draft choices and number of career starts are considered. Southern Cal, for example, would be the leader with 67 players drafted and 1994 games started by those draftees. Instead, our analysis is more of a judgment on the efficiency of each college.  A college that produces all seventh round choices could rank higher than a team with more or higher selections depending on the degree by which performance was better or worse than expected.
Only colleges with at least 10 players drafted in the study period were included in the analysis. There are 76 colleges that meet the criteria and they are listed in the following table.
The table includes the following information for each college:

  • The number of players drafted in the 10-year period
  • The actual number of starts made by those players
  • The surplus or deficit per player
    • A number in parentheses indicates a deficit

A zero value indicates performance is exactly average.  A surplus means that performance is above average, the higher the better. A deficit represents below average performance.
Some of the highlights associated with this list are as follows:

  • Central Florida, Purdue and Mississippi were the only three colleges with an average surplus per play of ten games or more
    • The combined draft selections of the three teams combined (51) was less than six colleges
  • Brandon Marshall and Josh Sitton, both fourth round selections, account for a significant portion of Central Florida’s positive rating
    • Bruce Miller, a 7th round selection, is also a major contributor
  • Purdue’s positive results were largely driven by Ryan Kerrigan (1st round pick), Bernard Pollard *2nd round pick) and Cliff Avril (3rd round pick)
  • Mississippi was led by 1st round picks Patrick Willis and Michael Oher and 3rd round pick Mike Wallace
  • Four colleges that had among the highest number of draftees (Florida State, Ohio State, Oklahoma and Southern Cal) were ranked in the bottom 20 of this table
    • Could player’s from the “big schools” be overvalued and possibly overdrafted?
  • Florida State’s position is largely due to the underperformance of its draftees in the second through fourth rounds
    • Chief underperformers included Everette Brown, Lorenzo Booker, Buster Davis, Willie Reed and Craphonso Thorpe
  • Ohio State had some underperformers in the first round (Bobby Carpenter and Vernon Gholston) and every one of their six third-round choices underperformed
    • The six third-round picks combined for a grand total of 19 NFL starts
  • Oklahoma had marginal success in the first two rounds but draftees in the remaining five rounds underperformed
    • Only four players of their 31 selections after round three had a surplus
  • While there are plenty of success stories, Southern Cal draftees as a group pretty much underperformed throughout the draft
    • Biggest deficits were from Matt Leinert (1st round), Dwayne Jarrett (2nd round) and LenDale White (3rd round)
  • While Rodger Saffold was a net positive performer, three wide receivers (James Hardy in round 2, Courtney Roby in round 3 and Isaac Sowell in round four) dragged Indiana into the basement

Follow Tony on Twitter @draftmetrics

Rating General Managers in the NFL Draft

An NFL General Manager has a full plate of responsibilities ranging from signing or retaining free agents to overseeing the draft process. All these duties must be carried out in the context of the Salary Cap, adding another layer of complexity. The ultimate judgment regarding how well a General Manager does his job is, of course, his team’s won-lost record.
It should be noted that not everyone in our analysis carries the General Manager title. Nick Caserio of the Patriots, for example, is the Director of Player Personnel but is the closest person to a General Manager in their front office.
This article focuses on only one aspect of the job – – managing the draft process. We have typically used some measure of being a five-year starter in evaluating performance. That would not be useful in this analysis, though, because it effectively excludes the last six years in a 10-year analysis.
In order to facilitate a more current look at results, our rating were based on a comparison of actual starts and projected starts for all players drafted between 2005 and 2014. Let us take Frank Gore to illustrate the concept. Gore was drafted by San Francisco as the #65 selection in the 2005 draft. This means that he could have started a maximum of 160 (10 years times 16 games/year) games. Historically, players drafted at that point of the draft started about 36% of the maximum or about 58 games for a player drafted in 2005. The actual number of games started by Gore is 134, so he exceeded expectations by 76 games. This 76 game “surplus” is credited to Scot McCloughan, the General Manager at the time, because he was good enough or lucky enough to select Gore.
The percentage used is calculated for each of the Draft Ranges, as defined in earlier articles. This means there is no inherent advantage in our analysis from having an early first round choice (i.e., the top pick in the draft) compared to a later choice (e.g., pick #32). That is taken into consideration in calculating the expected number of starts.
This calculation is repeated for each player drafted between 2005 and 2014. A summation of the relevant individual scores is then made for each General Manager. The resulting total surplus or deficit for each General Manager is divided by the number of years in his tenure between 2005 and 2014, resulting in an average annual rate. McCloughan was, for example, employed as a General Manager for five years and ended up with a total surplus of 295 games, resulting in an average annual surplus of 59. The conversion to an annual rate is done to provide comparability among General Managers with different employment tenures.
An average annual surplus of zero indicates a General Manager that performs at exactly the league average. A high surplus is good. A deficit is bad.
The table that follows ranks all current General Managers by their average annual surplus or deficit, with the largest surplus indicating that the General Manager was the most effective at his job on draft day. Please note that the drafting General Manager receives credit for a player regardless of whether the player remains with the team. If Gore had left the 49ers after three seasons and played for the Chargers, for example, McCloughan would still receive credit for all of Gore’s starts.
The table requires some explanation:

  • The first two columns are self-explanatory.
  • The third column cannot be greater than 10 (representing the 10 years studied) and represents the total number of seasons spent as a General Manager from 2005 through 2014.
  • The next column cannot be greater than the preceding one and represents the number of years spent as General Manager of the current team during the 10-year time period.
  • The “Average Annual Surplus (Deficit)” column was calculated as explained above and the table is sorted by those values
  • The final column represents the other teams for which the person served as General Manager during the 10-year period.


There is no perfect way to statistically analyze a General Manager’s draft performance. What is presented here is one way of doing it. A surplus can result from drafting more NFL starters, players with longer careers than typical or some combination of the two.
A few highlights from the above:

  • Six teams have had the same General Manager for at least the past 10 seasons
  • Almost half the teams (14) are within 10 games per season of the average
  • Scot McCloughan left the 49ers for “mutual reasons” and went on to serve as an advisor to John Schneider, the #2 rated General Manager, for four drafts
  • Mike Maccagnan, the Jets’ new General Manager, came from the Houston Texans where he was Director of College Scouting
It is also interesting to look at people who served as General Managers for at least three seasons over the past 10 years and who are no longer employed as a General Manager. It is no surprise to Lions fans that Matt Millen is at the bottom of the rankings but some of the names at the top might be surprising.

Follow Tony on Twitter @draftmetrics