All the talk about the quarterback prospects in the 2021 NFL draft begs a question: How good is this class really going to be in the long term?
There are those who have utilized analytics to project how much success a quarterback prospect will have, and among them is Football Outsiders' QBASE.
Football Outsiders first rolled out the model in 2015 and it, in part, explained why analytics saw things differently from NFL scouts when it compared the top-2 QB prospects at that time — Jameis Winston and Marcus Mariota.
This year, Football Outsiders rolled out a new version of QBASE which takes into account functionally mobile QBs, such as Patrick Mahomes, Deshaun Watson, Josh Allen, Lamar Jackson, and Kyler Murray. QBASE is projecting the likelihood of whether a quarterback will be elite (Hall of Fame player), upper-tier (franchise QB), adequate starter (can get by with him but you want to upgrade), or bust (backup at best).
I wanted to look at both models for some perspective, but keep in mind that the 2021 QB draft class is evaluated with QBASE 2.0, not the previous version.
How Both Models Evaluated Past QBs
Let's consider the Denver Broncos quarterbacks that have been drafted in the past few years in the top-100 picks: Brock Osweiler, Paxton Lynch, and Drew Lock.
The first version of QBASE projected a negative DYAR for Osweiler in years 3-5, a tally that came to pass. In other words, the model predicted Osweiler as more likely to bust than become at least an adequate starter.
If you go back to the 2016 projections, it gave Lynch a 67% chance of busting and a 33% chance of being at least an adequate starter. Looking at the 2019 projections, it gave Lock a 59% chance of busting and a 41% chance of being at least an adequate starter.
As for QBASE version 2.0, it remained negative on Osweiler, with Lynch and Lock each barely making it as an adequate starter. Of course, time has told the story about Osweiler and Lynch, while the jury is still out on Lock.
Still, if you use QBASE as your end-all, be-all analysis, you're likely prepared to move on from Lock. But let's compare Lock to the projections for other QBs.
Under the old model, Lock's 59% chance of busting was lower than Josh Allen, and under the new model, Allen was still projected to bust. This helps explain why people keep citing Allen as the Lock comparison.
Meanwhile, the new model puts Lock about the same as Lamar Jackson in terms of projections, but the older model gave Jackson a better chance of being at least an adequate starter (54%) than a bust (46%).
However, in my previous discussion about QBs and what likely comes into play to ensure they are successful, I argued that both Allen and Jackson were drafted by regimes who saw them as their guy, with the Buffalo Bills able to build around their guy and the Baltimore Ravens building like they had their guy before Jackson became available.
I would argue that, if you want to ensure Lock's success, you need to build as if he were your guy, and it's only more recently that the Broncos started building the team like they had their QB. Remember that Denver's 2015-17 draft classes netted just two long-term starters, while 2018 and 2019 each suggest at least two, and 2020 has potential for the same.
How Accurate has QBASE Been?
Before we make any conclusions about any prospect, it's fair to ask how accurate QBASE is in determining projections. Keep in mind that no model is perfect, but the best models should be good overall in projecting outcomes, and if they missed on a projection, the brainchildren of the model need to dig deeper into the details to determine why.
Both QBASE models loved Mariota, who most everyone would call a bust in retrospect. However, there are a couple of factors that didn't help him in his first few years.
Among them was the firing of Ruston Webster after 2015, the general manager who drafted him, the firing of head coach Ken Whisenhunt midway through the season, while his successors — head coach Mike Mularkey and offensive coordinator Terry Robiskie — failed to develop Mariota. Both were fired after 2017. Injuries didn't help Mariota's cause either.
Both QBASE models also love Baker Mayfield, and I'm not sure you can say the consensus is that he has solidified his status. Of course, Mayfield has gone through three different head coaches and two GMs in three years, so that could be an issue. We'll find out in 2021 if he improves upon a modestly successful 2020 campaign.
As for notable top QBs, both models liked Mahomes but didn't give him high odds of being an elite starter, while the old model was lukewarm to Watson and the new model was slightly more favorable. Regarding the 2020 draft class, both models liked Justin Herbert, Joe Burrow, and Tua Tagovailoa, while the new model likes the first two but isn't as high on Tagovailoa.
The first version of QBASE liked Josh Rosen's chances of being at least an adequate starter, and we know what came to pass there. However, the newer model doesn't like Rosen, which likely takes into account his lack of mobility.
One might argue the model can project whether or not a QB will at least be adequate, but isn't necessarily the best model for projecting who will be elite.
On the other hand, both models have been good about projecting busts. The old model didn't like Winston and the new model wasn't a big fan either. The old model gave Sam Darnold a 52% chance of busting and the new model isn't big on him. Neither model gave Dwayne Haskins a good chance of being at least an adequate starter.
And when it comes to those QBs that are projected as backups by most scouts, QBASE has generally seen them the same way. For example, both models saw Christian Hackenburg as a clear bust, and we all know the story there.
I won't go into every QB and how QBASE projected his success level, though. Instead, let's look at the prospects that everyone is talking about.
2021 Draft Class
To sum up, QBASE V2.0 doesn't see Clemson's Trevor Lawrence as a guaranteed success but likes his chances to be at least an adequate starter with a low chance of busting.
Regarding BYU's Zach Wilson, QBASE likes his chances of making it and sees a low bust rate, but not as low as Lawrence. If those two go No. 1 and 2 overall as expected, and QBASE proves correct on both, the Jacksonville Jaguars and New York Jets should be quite happy.
QBASE gives Ohio State's Justin Fields a 40.8% chance of busting, but that still means his chances of making it are better than not. However, I do think it indicates has to see Fields as their guy and do a good job of building around him and developing him.
The same applies to North Dakota State's Trey Lance, only more so, because QBASE gives him a 43.4% chance of busting. But, among the top four, QBASE thinks they have a better chance of making it than not, though it doesn't think any are guaranteed to be elite.
QBASE is not a big fan of Alabama's Mac Jones, though, giving him a 54.8% chance of busting, mostly because of his lack of mobility and the fact he had quality teammates. I would definitely say Jones is not worth a top-10 pick, but I suspect he goes somewhere in the first round.
Again, you can check Football Outsiders for more details, but I think the same principles I've discussed remain: If you love a QB, you should get him, but only if you love him — and if you do, be sure to build well around him and properly develop him.
The one thing I can take from QBASE, though, is this: If it thinks the QB has more than a 50% chance of busting, you should proceed with caution with the QB, and only take him if you really love him.
Follow Bob on Twitter @BobMorrisSports.
Subscribe to Mile High Huddle on YouTube for daily Broncos live-stream podcasts!