clock menu more-arrow no yes

Filed under:

Statistics, Bias and the Draft, Part 4

New, comments

Drafting Starters

NCAA Football: West Virginia at Baylor Ray Carlin-USA TODAY Sports

I was originally inspired to write this series of articles as a reaction to what I perceived to be some common misconceptions about the draft that are frequently repeated in comments threads on Hogs Haven. Along the way, the focus has switched to this year’s hot topic: should the Redskins pick the best player in the draft at #2 or trade back to add more picks?

The second article in this series examined the probability of drafting an elite playmaker at the #2 overall pick, as well as a selection of draft picks in the first three rounds that might be involved in trade down scenarios with Miami or the Chargers. I used the well known, if somewhat outdated, Jimmy Johnson/Mike McCoy Trade Value Chart (TVC) to determine fair value in the trade scenarios. In the process, I discovered that the relative values assigned to the selected draft picks by the TVC appeared to correlate with the probabilities of selecting elite players at those picks.

In the third article, I validated that finding by demonstrating a strong correlation between TVC values and the hit probabilities for elite players across the first 96 picks. Because of that correlation, trades valued by the TVC give each trade partner roughly the same probability of selecting an elite player. The one benefit of trading down I demonstrated is that spreading that hit probability across multiple draft picks reduces the risk of losing the drafted players to injury. This may have left some readers wondering if there really is that much of a benefit to trading down, particularly if it means missing out on a “once in a generation” talent like Chase Young.

In this final article in the series, I will show that elite players are just part of the story. To really appreciate the full picture of team building through the draft, though, we need to indulge in a bit of football philosophy. And who better to enlighten us, than the philosopher king of NFL team building, Scott McCloughan.


NFL-Minnesota Vikings at Washington Redskins Photo by Jonathan Newton/The Washington Post via Getty Images

The Wisdom of Scott McCloughan

“He’s not the most flashy guy, he’s not the prettiest guy, but he’s a football player and that’s what I’m talking about with a red… Blues are like perennial Pro Bowl players. Reds are like really good football players…So, if you’re lucky, you have four or five blues and hopefully one of those is your quarterback. If you have another 30 who are reds, starters, solid backups, core special teams guys, then you have a chance. That’s how you build your roster. That’s how you build a team.” – Scott McCloughan to Bleacher Report

It takes 53 players to man an NFL roster. In the salary cap era, most teams can afford to keep four or five star players, maybe a few more if their QB is on a rookie contract. What Scotty M is telling us, if we are willing to listen, is that the core of a competitive football team is the 30 or so quality starters, backups and special teamers who make it possible for the stars to shine brightly on game days.

How does the Redskins roster compare to McCloughan’s ideal heading into Ron Rivera’s first draft as head coach and with plans to switch to a 4-3 defense? Starting with the blues, we have Landon Collins and Brandon Scherff, if he bounces back to pre-2019 form, and perhaps Tress Way. Technically, Trent Williams is still on the roster, but he has vowed never to play for the team again and does not appear to fit Riverboat Ron’s vision for the future.

In the red department, by starting position, we have:

RB – Adrian Petersen

C – Chase Roullier

WR1 – Terry McLaurin*

DT – Daron Payne, Jon Allen, Matt Ioannidis, Tim Settle

DE – Montez Sweat*, Ryan Kerrigan, (Ryan Anderson?)

MLB – SDH, John Bostic

SLB – Cole Holcomb

CB – Kendall Fuller, (Ronald Darby?)

FS – Sean Davis

K – Dustin Hopkins

LS – Nick Sundberg

*Indicates red rookies flashing blue potential

By my count, that’s three blues – one playing a franchise tag and one punter - and sixteen reds. Some might even call my characterization of reds overly generous, but I’m sticking with McCloughan’s definition which includes solid backups and core special teamers. At least five offensive and defensive starting positions are not manned by reds or blues: LT, TE, WLB, LG, WR2. And it would be fair to say that big questions remain about CB1 and non-starting positions slot WR and 3rd down back.

Behind the established players, there are a few younger players who might develop into blues and reds and fill some of the vacancies this year, including Dwayne Haskins, one of the injured RBs, Steven Sims, Kelvin Harmon, Wes Martin, Fabian Moreau, Jimmy Moreland, and maybe Hale Hengtes. But even if half of them make good, the team is still well short of fielding a competitive starting roster, and I haven’t even really talked much about depth.

Given the state of the Redskins’ roster, is the best move with the second overall draft pick to chase a single, high-impact blue chip player, such as Chase Young? Or would the team be better off trading down to make it possible to address multiple positions on the roster with day one and two draft picks? In the previous article I estimated the probability of drafting elite players. To address the second half of the equation, I would now like to take a look at the probability of drafting starting quality players at different draft positions and the impact of trading down on the team’s ability to fill out its starting roster.


Probability of Drafting Starters

As in the previous two articles, to estimate the likelihood of drafting a starter at different positions in the draft I compiled data on the number of starts at each pick number over the 20-year period from 1998 to 2017, using data from Pro Football Reference. Since the Redskins have immediate needs at multiple starting positions, I looked at year-1 starts (1998 to 2017). And since it often takes players a few years to earn a starting spot, I also looked at eventual starters by measuring games started in the third year post-draft (2000 to 2019).

The first graph shows the average number of starts by players drafted at each pick number in their rookie season. As we’ve seen with all the draft stats thus far, there is a fair amount of variation from pick to pick, but there is a clear downward trend, as illustrated by the curve fit (dotted line, polynomial curve). Average number of starts maxes out just shy of 12 at the top of the first round and then declines fairly steeply to about three by the end of the third round. A team needing to fill multiple starting spots this year would do best to focus on the first round and first half of the second round. Unless they have a lot of first-round picks, like Miami this year, free agency is probably a better place to find players to start this year.

The next figure shows the average number of starts in players’ third years post-draft. By the third year in the league, the average number of starts has increased at all the draft positions. You may have noticed that the curve is not as steep as in the first figure. That is because the biggest increases in numbers of starts are in the late first to middle rounds of the draft. That should not come as a surprise. On average, players drafted early on day one tend to start early. It often takes a few years for players drafted after that to earn a starting position, and this effect gets bigger as the draft progresses into day three. Players drafted later on day three have a low chance of earning a starting spot to begin with and that doesn’t improve a great deal by year three.

In order to directly compare different trade scenarios in terms of potential to add starters to the roster, I calculated the proportion of starters drafted at each pick number over the past 20 years, which provides an estimate of the future probability of drafting starters at those positions.

The trickiest part of this analysis was deciding on how to define a starter. I wanted to exclude players who just played a few games in a season. On the other hand, I wanted to include players who took a few games to earn a starting spot in their rookie years, and not exclude genuine starters who missed a few games in a season due to injury. So I settled on eight or more starts. I could have gone with 10 or even 12, but I thought the higher thresholds would be unfair to first-year players. And remember, this analysis is aimed at estimating the chance of players who are good enough to earn a starting spot, not necessarily superstars.

The next figure shows the proportions of first-year starters at each draft pick number over the same 20-year period. The red line, hugging the bottom of the scatter of points, shows the Jimmy Johnson/Mike McCoy Trade Value Chart (TVC), scaled to the maximum of the curve fit to the starting data. The thing that really pops out here is that the TVC, which does a really good job of tracking the probability of drafting elite players, grossly undervalues all draft picks after #1 overall in terms of potential to add starting quality players.

That has a clear implication for trade value. A team looking to add quality starters, rather than superstars, can achieve surplus value by trading down in any pick swap that has equal value according to the TVC. The maximum surplus value is achieved where the two curves are farthest apart, from the middle to the end of the first round. But there is a large surplus value favoring the team trading down from the first overall pick right through the third round. As a result, the team trading down may receive a package of picks that give the same cumulative probability of drafting an elite player, thanks to the equalizing function of the TVC, but which are heavily discounted relative to their ability to yield starting quality players. The net effect is to increase the overall probability of drafting starters.

To illustrate this effect, consider a trade of Washington’s #2 pick for Miami’s #5 and #18 picks. According to the TVC, #5 is worth 1700/2600, or 65% of the value of Washington’s pick, while #18 is worth 900/2600, or 35%. However, the relative probability of drafting a first-year starter at each of the later picks, as read off the dotted blue curve, does not drop nearly as much as the TVC values. The probability of drafting a first-year starter at #2 is 0.76 versus 0.72 at #5 (95% vs. 65% relative TVC value) and 0.57 at #18 (75% vs 35% relative TVC value). The cumulative probability of drafting a single first year starter in this scenario increases from 0.76 at #2 to 0.88 (1 – (0.28*0.43)) with picks #5 and #18.

But the most important value gain for a team with a depleted roster is that there is now a 0.41 probability (0.72*0.57) of drafting two first-year starters, which is roughly comparable to the 0.44 chance of drafting one elite player (3 time Pro Bowler or better) at #2. And bear in mind, those two first-year starters are still first-round picks, with one elite prospect likely to be available at #5 (Jeff Okudah, Isaiah Simmons, Jerry Jeudy, Tua Tagovailova, a top-five left tackle, etc.).

The next figure shows the probabilities of players starting in their third year post-draft. The probability of players starting at each draft pick as increased across the board, and as a result, the gap between the starting players curve (dotted line) and the TVC has grown. This means that surplus value of trading down for drafting eventual starters is even greater than that for first year starters.

The probabilities of drafting first year and third-year starters at each of the draft picks that might be involved in some possible trade scenarios with Miami and the Chargers are shown in the table, below. Since the Chargers don’t have the same fire power as Miami in 2020, I’ll look at some trade scenarios involving 2021 picks. It is standard practice to value future picks one round lower in trades. And since it’s unknown what pick the trade partner will hold next year, they are estimated to be the middle pick of the round (#17) for purposes of both the trade valuation and the starting probability estimate.

Probabilities of Drafting Starters
Pick # Team TVC Value p (Year-1 Starter) p (Year-3 Starter)
2 WAS 2600 0.76 0.86
5 MIA 1700 0.72 0.83
6 LAC 1600 0.71 0.82
18 MIA 900 0.57 0.72
26 MIA 700 0.5 0.66
37 LAC 530 0.41 0.58
39 MIA 510 0.4 0.56
56 MIA 340 0.3 0.46
70 MIA 240 0.23 0.39
71 LAC 235 0.23 0.38
112 LAC 70 0.1 0.24
2021 1st rd. (#49 trade value, #17 probability estimate) LAC 420 (#49) 0.58 (#17) 0.73 (#17)
2021 2nd rd. (#81 trade value, #49 probability estimate) LAC 185 (#81) 0.34 (#49) 0.50 (#49)

To examine the impact of trading down from #2 on the team’s ability to add starting players, I used the probabilities listed in the previous table to calculate the cumulative probabilities of drafting starters in five trade-down scenarios, as shown in the following table. For this illustration, I used the probabilities of drafting third-year players, to put the focus on long term team building potential. I also included the RG3 deal to act as a kind of an upper limit on the value of trade offers Washington might reasonably expect to receive. By coincidence the 2012 first and second-round picks that Washington gave up are the same as picks held by LA and Miami in 2020, which is very convenient for comparison.

The probabilities shown are the chance of drafting one or more, two or more, three or more, etc. starters with the specified combinations of picks. As I should have explained in the previous article, the probability of two independent events both occurring is the product of the individual probabilities of each of the events on their own. So the probability of drafting two starters with picks 5 and 18 is 0.83*0.72 = 0.60.

The more complex calculations, such as the probability of drafting two, three, and four or more starters in a five-pick swap, involved calculating the joint probabilities of all the potential draft outcomes with the appropriate set of picks and summing the probabilities of all the outcomes with the specified number or more starters drafted. These became fairly large probability tables which would be pretty laborious to provide an example of here. If anyone has a question about these calculations, I’d be happy to discuss by private email or in the comments.

Probabilities of Drafting Numbers of Starters in Different Trade Scenarios
Picks TVC Diff. p(1+) p(2+) p(3+) p(4+) p(5)
2 - 0.86 - - - -
5, 18 0 0.95 0.6 - - -
5, 26, 70 40 0.96 0.72 0.21 - -
18, 26, 39, 56, 70 90 0.997 0.89 0.61 0.26 0.1
6, 37, 71, 112, 2021 2nd 35 0.98 0.84 0.51 0.17 0.02
6, 37, 2021 1st, 2021 2nd 135 0.99 0.89 0.57 0.17 -
RG3 deal: 6, 39, 2x future 1sts -350 0.99 0.93 0.67 0.24 -

The figures in this table illustrate the huge gain in effective draft capital achieved by trading down, if the goal is to fill starting positions on the roster, rather than adding elite talent. The Redskins enter the draft with a 0.86 probability of drafting an eventual starter with the #2 pick. In the previous article, I showed that they have a 0.44 chance of drafting a 3-time Pro Bowler or better at that pick. Trading down for the 5th and 18th picks, which is dead even by the TVC, increases the chance of drafting one starter to 0.95, and creates the new possibility of drafting two starters with a 0.60 probability. Thus, the chance of adding two starters in the trade down compares favorably to the chance of drafting an elite player if they stick at #2. And they will still be likely to be picking an elite prospect in any trade involving picks 5 or 6.

Ron Rivera has signaled to the league that he will not move from the #2 pick unless he is offered a huge haul in return. He has also indicated that he is unlikely to trade very far down from #2, which is how I interpret, “If you’re going to pass up on Player A and you go back and you take Player B, Player B has to equal Player A”. So I have included trade scenarios with a range of positive TVC point differentials, all of which still fall well short of the RG3 trade. And with one exception, I have focused on trades involving picks in the top six to give Ron the chance to pick a prospect who is in the same tier as Chase Young.

In three of these trade scenarios, the probability of adding two or more starters is comparable to the probability of drafting one starter with the #2 pick, and the probability of drafting three or more starters is greater than 0.5. The results illustrate that there is a clear trade-off to consider with the Redskins’ second overall pick. If a “fair value” trade offer is received from a team holding a high first-round pick, they have to decide whether the best move for future of the team is selecting prospect with a good chance of becoming an NFL superstar, versus having a chance to upgrade multiple starting positions, and at least some of those with players who have a decent chance of becoming very good to elite players.


Summary and Conclusions

In the previous article of this series, I demonstrated that the Jimmy Johnson/Mike McCoy Trade Value Chart, used by NFL teams for many years, closely tracks the probabilities of drafting elite players at different positions in the draft. The TVC places very high value on high first-round picks. Its pick valuations drop sharply from the first overall pick and reach an inflection around the middle of the first round, after which it becomes less steep throughout the later rounds. The availability of elite talents is greatly diminished by the end of the first round. Because of the huge variability of draft outcomes, it remains possible to find the occasional elite player right through the sixth round and in free agency after the draft, but those finds are extremely rare.

By contrast, in this article I showed that the availability of players good enough to make an NFL starting roster falls off much less steeply than elite players, and extends well into the middle rounds of the draft. This is particularly true for teams with the patience to wait a few years for their draft picks to develop.

While the TVC equalizes trade values for teams seeking elite players, the difference between the rates at which elite players and starting-quality players are exhausted as the draft progresses creates a significant opportunity for teams seeking starting players to build surplus draft capital by trading high first-round picks for multiple picks later in the draft. This effect is fairly large. As I have shown, trades with moderately positive TVC point differentials, probably less than what Ron Rivera is seeking, can result in probabilities greater than 0.5 of adding three or more early round starters to the roster. And the probability of that happening actually compares favorably to the probability of drafting a single elite player at #2 overall.

I started writing this series of articles with a fairly simplistic view of the value equation involved in trading down in the draft. That was based on an expectation that availability of elite and starting-quality players would decline at about the same rate as the draft progresses. The discoveries I have made along the way have changed that. In particular, the findings that 1) the most well-established, if now dated, trade valuation tool is tied to the probability of drafting elite players, and that 2) the availabilities of elite players and eventual starting-quality players decline at different rates as the draft progresses, add layers of complexity to draft strategy.

Based on these results, I would now say that teams seeking to add elite talents would have to base their draft trade strategies on a different curve to teams seeking to add the most possible starting players. There are times when either strategy might suit a team’s needs. Considering the state of the Redskins’ roster in the wake of the Bruce Allen era, the figures in the last table of this article make it crystal clear to me what the team should do if they receive the kind of trade offer that Ron Rivera has signaled he is seeking. But that is just my opinion, and now I would be very interested to hear what the Hogs Haven readers think.

Acknowledgement: I would like to give a big shout out to James Dorsett who, along with Bill in Bangkok, was one of the regular contributors whose articles originally inspired me to start writing for the Hogs Haven. James has provided valuable input, advice and editorial support to this series of articles and my first article last year. These articles would not have been possible without his help and, if I had managed to do it without him, they wouldn’t have been nearly as readable.

Poll

If you were Ron Rivera on the clock at #2, would you rather...

This poll is closed

  • 42%
    Take the highest rated player on your board
    (119 votes)
  • 3%
    Take a 60% chance of drafting two starters in the first round
    (9 votes)
  • 6%
    Take a 72% chance of drafting two starters and a 21% chance of drafting three
    (19 votes)
  • 47%
    Take an 89% chance of drafting two starters, a 61% chance of drafting three, and a 26% chance of drafting 4
    (133 votes)
280 votes total Vote Now