There's A Lot of Consensus Boards Out There. Why Ours?

There's a lot of consensus big boards out there. Let me sell you on ours

There are a lot of consensus big boards out there purporting to give us information on the NFL draft. Which one should you keep track of?

Wide Left will be publishing its Consensus Big Board soon, aggregating data from over 100 analysts. That piece, like this one, will be free to all readers.

Back in 2015, when I started the Consensus Big Board project, there weren’t many resources that gathered together the accumulated wisdom from draft watchers. Over ten years later, that’s changed.

There are about a dozen places to find a consensus board, including multiple former employers of mine, as well as more open fora like Reddit or social media platforms. On top of that, we have access to other methods for aggregating opinions on player prospects in the draft, including the aggregated mock positions from GrindingTheMocks and the NFL Mock Draft Database.

Is there any serious difference between them?

To some extent, we explored this two years ago when discussing the different consensus boards.

Why come to Wide Left when all of these other options are available?

I wanted to answer that question with some rigor, so I contacted Benjamin Robinson, who runs Grinding the Mocks, and we worked together to test the boards. What I found might not be surprising. But first, a primer on how these boards are constructed.

Why Are These Rankings Different?

There are essentially four categories of player rankings I’ll be looking at: the limited-rank-average model employed by most consensus boards across the web, the mock draft aggregation produced by Robinson, the expanded-value-average model for Wide Left’s board, and the blended board/mock model employed by the NFL Mock Draft Database.

Most consensus boards take rankings from 10 to 16 or so analysts and average the ranks. So, Fernando Mendoza might be ranked {1, 1, 1, 1, 1, 2, 2, 3, 3, 4, 4, 5, 6, 7 and 8} by 15 boards, his average rank would be 3.27. Arvell Reese might be ranked {1, 1, 1, 1, 1, 2, 2, 2, 3, 3, 4, 4, 4, 4 and 5} for an average rank of 2.53, so he would be ahead of Mendoza. In this exercise, Jeremiyah Love ranks {1, 1, 1, 2, 2, 2, 2, 2, 3, 3, 4, 4, 4, 5 and 6} with an average of 2.80.

Oooh, preview!

So, an averaging approach would have those players rank Reese first, Love second and Mendoza third. Easy to understand, intuitive, and gets the broad picture across easily.

That might tell us who experts think are the best players in the draft, but it doesn’t tell us very much about where those players might go.

Grinding the Mocks focuses primarily on mock draft data (with a small amount of big board data) to construct a regression model that is tuned to better reflect expected player position. As the model gets more robust, it can be used to construct probability ranges for players.

This year, Robinson’s data projects Mendoza to go first (an “expected draft position” of 1.3), Reese to go (generally) second (EDP of 3.0) and Love to go fourth (EDP of 4.7).

Robinson’s model weighs more recent and more accurate mock drafts heavily, a method meant to take advantage of the most up-to-date information we’ve learned about teams or prospects, which may better tell us what might happen on draft day.

NFL Mock Draft Database’s “Consensus Board” also focuses on grabbing data from mock drafts as well as big boards, using over 1,000 mocks and nearly 150 big boards gathered throughout the year to produce a composite ranking that blends the two approaches.

There is also a decay function in the NFL MDDB, with no boards over four months old included in the calculation at any time and with recency weighting, as with Robinson’s approach.

Wide Left’s approach ignores mock drafts and assigns values to each player based on their rank from each analyst. Some consensus boards (like the Athletic’s) do the same, assigning 100 points for the top-ranked player on a board, 99 points for the second-ranked player and so on.

The value calculation Wide Left uses is modeled after various trade charts (such as the Jimmy Johnson, Rich Hill, Fitzgerald-Spielberger, and Chase Stuart charts), assigning significant value to the first overall rank and much less to the 100th-ranked player.

This way, the differences between players are better captured; assigning a player a top-ten rank suggests that the player is much more talented than the 32nd-ranked player, but assigning a player a rank of 101st is not that different from designating them as the 132nd-ranked player.

A logarithmic decline in value can capture changes in rank differences in a way that the average rank value cannot. In order to reduce the impact of outliers, Wide Left also uses the “interquartile mean” of all the values assigned by each analyst.

That means tossing out the values in the top and bottom 25 percent of the range and averaging the middle 50 percent. Then, to break ties, the outliers are reintroduced with a much lower weight.

In the above example, with Reese, Love and Mendoza — where Reese ranks first, Love second and Mendoza third by a simple average — Mendoza finds himself second, not third, because his outlier ranks of 7 and 8 are thrown out.

Photo by Michael Hickey/Getty Images

Not only that, the value of having that many first-overall ranks exceeds the value of the number of second-overall ranks Love receives.

Later in the draft, we find quite a few players ranked by some boards and not others. How consensus approaches deal with absent players in the rankings matters.

Different top 100s will have a different spread of players, typically encompassing a universe of 120-140 players.

A good way to understand the impact of failing to account for player absences in boards is a simple example. Let’s assume we have 100 analysts ranking players 1 to 100. We know that not every analyst will rank the same 100 players; most leave players on the cutting room floor who others have included in their top 100.

In this example, Player A is ranked by 30 boards as the 75th-best player in the draft and is unranked on the other 70 boards. Player B is ranked by all 100 boards as the 76th-best player. Most approaches to consensus boards will rank Player A as the better player, even though the average opinion on that player is that he is worse than Player B, not better.

Wide Left’s consensus board, with its points-based averaging approach and varying penalties for absences on different-sized boards, avoids that problem (at least partially). There is a bigger penalty to a player left off of a 500-player board than off of a 100-player board, and the value calculations reflect that clearly expressed preference from the analysts.

There is also a sample advantage when using 80-plus boards over 12-16 boards and it seems to play out when we look at outcomes.

OK, So Which Board is Better?

Trying to determine which board is better depends on what you’re trying to accomplish. In general, we rank players for two reasons: 1) to figure out where they’ll go in the draft or when they will no longer be available and 2) to figure out which players in the draft are most likely to succeed.

Predicting the Draft

For that first exercise, we can determine the average pick distance each board has from the NFL’s actual selections. For the averaging consensus board, I used the rankings I’ve gathered from legacy media and well-read draft analysts over the years, picking between 12 and 16 boards, depending on which ones were available. This meant grabbing the rankings from CBS Sports, The Athletic, various NFL.com boards, ESPN and so on.

Grinding the Mocks does the best job of giving us insight into where a player will fall in the draft. Using that as a guide to determine whether a team was right to trade up or to take a shot and reach instead of trading down seems like an appropriate use case.

The NFL Mock Draft Database, despite using a very similar dataset, is an entire pick or more off from the Grinding the Mock’s approach throughout the top 100, which is a bit of a surprise. Nevertheless, it significantly outperforms Wide Left’s consensus board and the typical average-rank consensus board we see on the web.

The fact that the Wide Left consensus board outperforms the typical consensus board in the first round by a good margin is an interesting insight, though that advantage disappears heading into the second round or the top 100.

Predicting Player Outcomes

But, which should we use if we want to figure out which players are best. Given how good the NFL is at rank-ordering players through the draft selection process, wouldn’t the board best at predicting how the draft will fall be the board we want to use for that exercise?

Not necessarily.

The NFL draft market isn’t very liquid; teams don’t organize optimally to make sure that player value, need, and resources all line up. A team that needs a quarterback is going to reach, and trades are pretty rare. Teams, who are risk averse, would rather stay put and make a selection than move up or down the draft until there’s significant reason to move.

Using a combination of player grades, postseason honors, production and contract value, I’ve put together a crude ranking system to order the players from each draft class. It works alright; Dak Prescott ranks first from the 2016 draft class, followed by Jared Goff, Carson Wentz and then Jalen Ramsey – in part because quarterbacks are worth so much more than other positions.

Following that are Joe Thuney, Tyreek Hill and Hunter Henry. In 2017, the highest-ranked players are Patrick Mahomes, Deshaun Watson, George Kittle and Marlon Humphrey. In 2018, it’s Josh Allen, Lamar Jackson, Baker Mayfield, Mark Andrews, Quenton Nelson, Dallas Goedert and Jaire Alexander.

These rankings are by no means perfect, but they’re close enough to perform an analysis on how rank-ordering worked for boards and the NFL. Because we don’t have as much information on the performance of players from the 2024 and 2025 draft classes, they’ve been excluded.

Draft values and contract values both tend to organize themselves with exponentially increasing or decreasing values, so after broadly rank-ordering the players from the drafts, we can assign draft value to them.

That way, we can compare the implied value from the big boards against the produced value of the players who were selected and treat them like the logarithmic trade values used in the Wide Left consensus board rankings.

From there, we can see how close each board approach was to determining a player’s long-term outcome value.

In terms of projecting player outcomes, it doesn’t look particularly close; the Wide Left Consensus Big Board vastly outperforms the mock draft database and the average ranking approach in the first round, second round and beyond. The Grinding the Mocks database, perhaps by virtue of getting closer to NFL draft selections, comes in second in each of those value buckets.

The expansive nature of the Wide Left Consensus Board really shines on Day 3, where having data on over 800 players organized by perceived talent allows for more granular grading on players who end up going on Day 3.

Which aggregate board should you choose? It depends on what you’re looking for.