Last offseason, as the usual discussion about Washington’s quarterback situation was running its course, I tried a little experiment.
The idea arose out of an article in which I compared the strategies the most successful teams over the past decade had used to find their franchise quarterbacks to the variety of approaches that Washington has attempted. I found that, with the sole exception of the New Orleans Saints, the perennial contenders from 2011 to 2021 found success with quarterbacks that they drafted, without making franchise-crippling trades to acquire them. The Redskins/Football Team/Commanders have tried just about everything else.
I concluded that Washington has been making the search for Mark Rypien’s successor much harder than it needs to be. To test that idea, I partnered with the team at BrisVegas Systems, a shadowy Brisbane-based technology firm, to develop the DraftBot.
DraftBot 1.0 used a conventional microprocessor which was programmed to execute a simple rule set for selecting quarterbacks in the NFL draft as follows:
IF the team is not set at quarterback;
AND it is holding a pick in the first half of the first round;
AND there is a quarterback with a first-round grade available;
The demo run, used nothing more than consensus rankings from well-known media draft analysts to rate QBs. The results demonstrated that, even without the aid of an NFL scouting department, the DrafBot’s algorithm would have selected Drew Brees, Ben Roethlisberger, Aaron Rodgers, Lamar Jackson and Tua Tagovailoa in their respective drafts rather than the players that Washington selected.
Leveraging that successful proof-of-concept demonstration, BrisVegas Systems was able to raise substantial funding from sources that I am not allowed to talk about for an ambitious expansion of the DraftBot’s AI decision-making platform for applications in another field that I’m also not allowed to talk about.
The good news for Commanders fans is that the successful demo established the value of the NFL draft as a test bed for the technology, and my colleagues at BrisVegas Systems have decided to extend the partnership. The DraftBot is back for the 2023 offseason with expanded capabilities to take on the full seven rounds of the NFL draft.
The main upgrade to the DraftBot was replacing the standard CPU with a state-of-the art neural network processor with highly sophisticated machine learning capabilities. This was loaded with data from all previous NFL drafts and player seasons to allow the DraftBot to identify trends and patterns that would have been undetectable using conventional statistical methods.
Due to certain national security considerations, I am unable to reveal the details of how the DraftBot works. What I can share is the main findings that arose from its training on the NFL draft dataset, which will inform its decision making in the draft.
The primary outcome of the DraftBot’s data mining exercise was the realization that, since the introduction of the salary cap, NFL teams have a continual need to get younger, better and cheaper at every position on their rosters. Rather than focusing on immediate needs, teams that have achieved the greatest benefit from the draft have used it as the primary platform to add premium talent and continually renew their player pipelines to prepare them to address future roster needs as they arise.
That doesn’t mean that the DraftBot ignores team needs in the draft. Rather, it has developed a different concept of “need” than many fans are used to. When the DraftBot is deciding between two similarly rated players on its board, the key question it asks is which player has the greatest potential to improve the team.
In the early rounds, this might be a player with potential to fill a vacancy or replace a weak starter. However, the DraftBot expects the pro personnel department to have patched any glaring roster holes ahead of the draft, and is just as likely to select a potential impact player at a position held by an average starter, or even a good starter on an expensive contract. Later in the draft, its focus shifts to building the development pipeline to prepare the team to address future needs. These can’t be predicted a year or two in advance when the players it selects are likely to be ready to contribute, so immediate needs become irrelevant.
Analysis of its decision making in beta testing has revealed two of its main operating principles:
QB takes precedent: The original principle that inspired the DraftBot’s creation has become hard wired in its network architecture. An above average starting quarterback has greater impact than an elite player at any other position. If the team needs a QB and one with a first-round grade is available that’s who it will pick.
Never let an average player stop you from drafting a potentially elite player: The DraftBot has discovered that teams benefit the most by adding impact players in the draft, rather than focusing draft capital on the weakest areas of their rosters. Having a passable starter in place won’t stop it from drafting a potential star.
Too Early Mock Draft
As the title suggests, it is still too early in the offseason to make even an educated guess about what the Commanders might do in the draft fourteen weeks from now. The final draft order hasn’t even been set, the final evaluation of draft prospects has just begun, comp picks have yet to be awarded, and major decisions still need to be made about which players will be re-signed, let go, or acquired in free agency.
In order to run a mock draft now, a few assumptions need to be made. This mock draft is based on the premise that the Commanders will release Carson Wentz, and will not be able to reach a deal to extend Daron Payne. The cap savings from those moves, and potentially others (Logan Thomas, Chase Roullier, etc.), will be used to sign free agents to patch the biggest roster holes at OL, LB and CB.
With those assumptions setting the stage, it is time to turn the controls over to the DraftBot 2.0 to run a mock draft using the Pro Football Network’s mock draft simulator.
Round 1, Pick #16
Bryan Bresee, DT Clemson
6’5”, 300 pounds
Damascus, Maryland native Bresee is one of the elite prospects in this draft class, and the DraftBot seemed genuinely surprised that he fell to pick #16. Like Jonathan Allen, whom the Redskins selected with the 17th pick of the 2017 draft, there are some lingering injury concerns, due to a season-ending ACL tear in 2021, which may have caused Bresee’s slide. If the medical evaluations give Bresee a clean bill of health, then the Commanders might have got a steal on Allen’s level, or even better.
How he improves the Commanders: Bresee is an athletic freak who deadlifts 585 lbs, has a 30-inch vertical jump, and runs the 40 in the 4.7s at 300 lbs. His greatest strength is as an interior pass rusher, where he combines the power to overwhelm linemen with elite instincts. In three years at Clemson, he recorded 60 QB pressures. He has played nose tackle for Clemson, but his athleticism gives him the versatility to line up all over the defensive front, including the edge. He has room to improve in run defense, but is not bad already and seldom misses tackles.
If he lives up to his potential, drafting Bresee could provide an upgrade to the interior DL, building on the Commanders’ key strength, rather than simply patching the hole left by Payne’s departure.
Washington received multiple trade offers, but all involved future picks. None of the trades offered sufficient net gain of draft capital or additional picks in this draft to make it worth passing on a player the DraftBot had rated 7th overall.
Why the DraftBot did not select a QB: Only two of the three criteria required to activate the DraftBot’s automatic QB selection protocol were fulfilled — the team is not set at quarterback, and it is holding the last pick in the first half of the first round. The most important criterion was not — there was no QB with a first round grade available at its pick. The three QBs with first round grades (Bryce Young, CJ Stroud and Will Levis) were all off the board by pick #9. The next QB on its board, Anthony Richardson, was picked sixth by the Detroit Lions in this simulation. The DraftBot has Richardson rated in the second round and would not have picked him even if he had been available.
Best Available Players on the DraftBot’s Board:
- Peter Skoronski OT, Northwestern
- Cam Smith CB, South Carolina
- Joey Porter, Jr. CB, Penn State
- Christian Gonzalez CB, Oregon
- Jaxon Smith-Njigba WR, Ohio State
- Broderick Jones OT, Georgia
- Bijan Robinson RB, Texas
Round 2, Pick #47
Darnell Washington TE, Georgia
6’7”, 269 lbs
With the second-round pick that was salvaged by Carson Wentz’s broken finger, the DraftBot selects the massive tight end from Georgia.
How he improves the team: The Commanders’ tight ends group experienced a significant decline in productivity down the stretch in 2022. There are question marks around Logan Thomas’ recovery from injury and none of the young tight ends stepped up to replace his production.
Washington’s unusual size and catch radius creates mismatches in the passing game like no other tight end in college football. In addition to his value as a receiver, Washington also excels as a blocker, and effectively adds a sixth lineman to the Commanders’ offensive front. He had the fifth-highest run-blocking grade among tight ends in 2022, according to PFF, while also catching 28 passes for 454 yards (16.2 yd average) and 2 TDs.
The Commanders only received offers of pick swaps, with no additional picks and essentially neutral value. Darnell was too appealing to pass up at #47.
Best Available Players:
- Devon Witherspoon CB, Illinois
- Drew Sanders LB, Arkansas
- John Michael Schmitz C, Minnesota
- JL Skinner S, Boise State
- Andrew Vorhees G, USC
- Cedric Tillman WR, Tennessee
- Tank Bigsby RB, Auburn
(Round 3 Comp Pick – yet to be confirmed)
Round 4, Pick #115
Andrew Vorhees G, USC
6’6”, 325 lbs
For the second time this draft, the DraftBot could not believe its luck. Vorhees was ranked not far behind Washington on its big board, so it was quite a surprise to see him still available after skipping the third round due to the Carson Wentz trade.
How he improves the Commanders: Due to roster neglect and poor FA choices in the 2022 offseason, combined with injuries at center, the Commanders have needs to bolster starter quality and build depth across their entire offensive line.
Vorhees has played at guard and tackle for USC, but projects to guard in the NFL. Echoing the DraftBot’s first selection, Voorhees is known for his freakish strength (#99), putting up 40 reps of 225 lbs in the bench press. He has the burst and power to overwhelm defenders in the run game and a strong anchor in pass protection. While the DraftBot normally expects to be picking depth players in the fourth round, Vorhees should be competitive with the players on Washington’s roster to earn a starting position in camp.
He would have been good value in the second round, and is potentially a major steal on Day 3. No trade offers even came close to matching the value of selecting Vorhees here.
Best Available Players:
- Tuli Tuipulotu EDGE, USC
- Tyler Steen OT, Alabama
- Zacch Pickens DT, South Carolina
- Sam LaPorta TE, Iowa
- Tanner McKee QB, Stanford
- Daiyan Henley LB, Washington State
Round 5, Pick #147
TRADE: Washington sends pick #147 to New Orleans for picks #156 and #204
Round 5, Pick #156
Tanner McKee QB, Stanford
6’6”, 230 lbs
The DraftBot has been looking for trades to add more picks in this draft and finally received an offer that met its criterion. Continuing its spectacular run of good luck, the DraftBot selects the player at pick #156 that it would have selected at #147. The DraftBot was expecting to find value in developmental QB prospects later in the draft, but McKee’s value here was too good to pass up. It had him rated in the third round.
How he improves the Commanders: McKee is a strong-armed, traditional pocket-passing QB who still needs to refine his mechanics and processing to make the most of his natural arm talent. He joins the Commanders’ QB room as the third man on the depth chart, taking over the developmental role that Sam Howell occupied in 2022. In the near term, he has potential to upgrade the backup position currently held by Taylor Heinicke. In time, he could eventually develop into a starter. The big question about this pick is whether a pocket passer with limited athleticism fits the Commanders’ offensive scheme. That will remain unclear until the Commanders hire an offensive coordinator.
Best Available Players
- Jaylon Jones CB, Texas A&M
- Nick Broeker OG, Ole Miss
- Jaxson Kirkland OT, Washington
- DeWayne McBride RB, UAB
- Noah Ruggles K, Ohio State
Round 6, Pick #179
Jaylon Jones CB, Texas A&M
6’2”, 205 lbs
Notice a pattern here? If the DraftBot hadn’t been swayed by the greater position value of a QB at #147, the pick would have been Jones. So just imagine its luck that he was still available in the 6th round.
How he improves the Commanders: Jones is a traits prospect at this stage in his development. He has a great combination of size, length, physicality and short area quickness that make him an intriguing developmental prospect to fill out the Commanders’ depth ranks in the secondary. While he is refining his coverage technique, he has a good chance to make the roster on special teams. He is the type of prospect that is likely to rise on draft boards after combine testing.
Around this point in the draft the DrafBot’s big board gets a little sparse and it becomes a challenge to say who really is the best available player.
Round 7, Pick #204
Dillon Doyle LB, Baylor
6’3”, 240 lbs
Doyle played ILB for Baylor. He is a good all-around athlete with very good short area quickness and field awareness, and is known for giving top effort on every play. He is a good overall linebacker, with value against the run, pass rushing and in coverage. His pass coverage improved every year in college, finishing with one pass breakup, 2 interceptions and an opposing QB rating of 77.9 in 2022.
How he improves the Commanders: The Commanders paid the price for failing to address linebacker in the 2022 offseason, experiencing a steep drop off in talent when backups Jon Bostic and David Mayo were pressed into service following Cole Holcomb’s injury. Mayo, in particular, had an abysmal 146.9 opposing passer rating in coverage. If allowed to compete, Doyle has a good chance to displace former Panther Mayo, and improve the Commanders’ depth at the position. He could provide additional value on special teams, where he has the best chance to get an early foothold on the roster.
Round 7, Pick #211
Zion Nelson OT, Indiana
6’5”, 309 lbs
In what has become a recurring theme in this draft, the DraftBot closes out by selecting a player who probably shouldn’t still be available in the seventh round.
How he improves the Commanders: Nelson is described as an excellent technician, who rarely gets beat in pass protection, but lacks the aggressiveness to excel as a run blocker. Perhaps some tough love from Matsko and Rivera can light the fire and turn him into another Day 3 steal. The Commanders are so thin on the OL, that you’d have to like his chances to displace someone on the depth chart.
Acknowledgement: Edited by James Dorsett
Rate the DraftBot’s mock draft
Too good to be true - PFN’s draft board needs work