The Draft Research Project: Draft Classes

The Overview

While there are many truisms to the draft, one that analysts, fans, and coaches all seem to agree on is that not all classes have an equal number of talented prospects. However, it is unclear how much variation actually exists. After all, at some point each team needs to find new players to fill vacancies left by free agency, retirements, injuries, and outright cuts. The draft is the primary source of new players. To some extent, then, it seems like saying a drafted player actually appears in a game just means that the team had an opening, not that he was any good.

Put simply—saying that a player started for a team does not mean that he was any good, only that the team he played thought he was better than other (possibly worse) options. The team needs someone to be a starter. If this line of reasoning is valid, then draft classes should be relatively consistent regarding how many spots on game rosters each draft class fills and regarding how many players from each draft find starting roles. Somebody’s got to wear the laundry.

As before, this study therefore looks at the seven draft classes (2011-2017) of the modern rookie salary era across just the first five years of their careers. Very superficially, a review of classes provides some evidence for the idea that classes have a certain amount of consistency. However, even this global view suggests shows important levels of variation:

The last two years under consideration (2016 and 2017) produced appreciably more games played than other years, and that’s even though they were “short” classes of only 253 players instead of the average of 254 (mean and median). The average games played per player also stands out for those two. Meanwhile, 2014 and 2015 are notably low despite being in “long” classes of 256. It is awkward to use proper statistical terms with sample sizes this small, but 2014 is more than a full standard deviation below the mean, and 2016 is likewise more than a full standard deviation above the mean.

Meanwhile, 2017 was amazing at producing Pro Bowlers and All-Pros while coming in second in terms of games played and average starts. By contrast, the same two underperforming classes (2014 and 2015) just did not seem to produce 1^st-Team All-Pros, and 2015 was short on Pro Bowlers generated.

Finally, it’s worth looking at the rate at which players earned a fifth year. In 2011 there was a much lower rate of players making it to a 5^th year, but 2016 saw a dramatic increase before settling back down at a number still 10% higher than 2011 and 7% higher than the next highest year.

Year-By-Year Outcomes

With a global view established, it is time to consider what this means in terms of individual players. Pulling terms across from the round-by-round piece, we have the following groups:

• Starters: players who started in at least 40 games
• Impact Players: players who appeared at least 40 games and earned at least one Pro-Bowl
• Stars: players who earned at least two Pro Bowl honors or at least two 1st-Team All-Pros
• Failures: players with careers lasting under 4 years

This chart is information-dense, but a few data points stand out. The apparent weakness of 2014 as presented in the global view disappears. It actually produced as many starters as any other class. It also has the most star and the most total Pro Bowls, despite having a very mediocre number of impact players, All-Pros, and an unexceptional (if high) number of failures. This class seems both rich in talent and lacking in low-end depth.

By contrast, the 2017 class produced only an average number of starters but more impact players than any other year. It also had an extraordinarily low number of failures and an extraordinarily high number of All-Pros earned.

To simplify things, here is the same chart with the numbers removed, but with any outcome that is at least one standard deviation below (-) or one standard deviation above (+) the mean highlighted. Looked at this way, 2013 and 2015 are apparently poor classes, while 2014, 2016, and 2017 are seemingly strong classes.

To make sure relative outliers are not driving the results, the “middle” of each draft is worth considering. Removing the top 64 players from each draft in terms of starts and Pro Bowls and then the bottom of each draft in terms of games and starts, the 130 players who represent the “core” of each draft produces outcomes that are consistent with earlier results. The core of each draft class provides a mean of ~6000 (+/-300) games played, a mean of 80 players (+/-10) who earned 5^th-year contracts. Likewise, they produced 27 players (+/-6) who failed to see the fourth year of a contract. Here are specific results with outcomes more than one standard deviation outside of the mean highlighted:

Note how the general direction of games contributed, starts, and successes is still upward (with a corresponding decrease in failures). When 2018 is available for study, it will be worthwhile to note whether 2016 was a peak for draft outcomes of if there has been a general trend toward increased efficiency in drafting. Even without that data, however, these seven classes can still be analyzed in their own right. 2016 was obviously far deeper than the other classes, as even with the best two rounds’ worth of players removed, it stands out in every category. Meanwhile, 2011 remains problematic in most categories. 2012 and 2017 stand out a little, but only by the same standards as before. 2014’s problems do not extend into its core. Instead, it seems as if this class had truly poor showings from the lowest performers, dragging down earlier averages.

Here is a summary of the results so far, simplified to note findings that are at least a single standard deviation outside of the norm:

• 2011 was below average in games per player and core strength with a high failure rate, but it was above average in total All-Pros earned. Essentially, this was a poor class bolstered by a very few select players (J.J. Watt, Richard Sherman, and Patrick Peterson stand out).
• 2012 was below average in failure rate but otherwise very average. Given that it shares this very high failure rate with 2011, one explanation could be teams not yet maximizing the efficiencies of the rookie salary scale.
• 2013 was below average in producing starters, stars, and total Pro Bowls. It was thoroughly average in producing rank-and-file players, however.
• 2014 was below average in total games/games per player and All-Pro rate, but it was above average in terms of starters, stars, total Pro Bowls. While there was a very strong “top tier” of talent available, there was an equally poor “bottom tier” leading to notable failure rates and a corresponding poor average performance in some categories. This class was a class full of haves and have nots.
• 2015 was the worst draft class studied. It was below average in games per player, Pro Bowler rate, impact players, stars, total Pro Bowls, and total All-Pros.
• 2016 was an amazing class that was above average in total games/games per player, starts per player, 5^th-year success, starters, and impact players while maintaining a very low failure rate.
• 2017 competes with 2016 for the title of the best class studied. It yielded above average games/games per player, Pro Bowler rate, All-Pro rate, 5^th-year success, impact players, total Pro Bowls, and total All-Pros (with a very low failure rate).

Conclusions

In practical terms, the swing between a strong and a weak draft class could be as many as 13 starters and six impact players. Some draft classes produce twice as many stars as others, but even a strong draft class will not produce one impact player per team, and even if stars and starters were evenly distributed, it would take two strong years for every team to draft a star and four starters.

Previous analyses, including my own, have shown different results here in that they have counted drafted players as starters on the basis of their eventual status, not how they performed under their first contracts. In simple terms, a number of drafted players grow into starters only after they have had a chance to become free agents.

This overview potentially has implications for draft strategies at this point. If the goal of the draft is to acquire starters, then most years produce at least 60-70 chances to find starter-level players. Trading down and accumulating additional chances increases the chances of acquiring more players to suit up in games as role-players and more starters. However, if the goal of the draft is to acquire impact players or stars, then the strategy of trading down possibly leaves a very narrow margin. In a weak year such as 2013 or 2015, there might not be enough high-potential players to make this approach work.

However, before drawing such conclusions the next pieces need to consider the positions drafted and a results-driven breakdown of draft order.