Bert Bell used to own the Philadelphia Eagles back in the franchise’s infancy and he is credited with “inventing” the annual Draft as a way to increase competitive balance – and thereby entertainment value – in the new pro football league. That concept has been adopted by most US professional leagues, and I have argued here before that the idea is theoretically good but practically of marginal value.
One glaring problem is the propensity of teams that are bad enough to “earn” a top draft pick finding a way to draft a guy who simply cannot play at the pro level. Let me refresh your memory about some humongous NFL Draft busts from the recent past:
- Ryan Leaf – – Chargers – – 1998
- Matt Leinart – – Cards – – 2006
- Trent Richardson – – Browns – – 2012
- JaMarcus Russell – – Raiders – – 2007
- Akili Smith – – Bengals – – 1999
By the way, I have no interest here in picking on the folks who draw up draft boards for NFL teams; this happens elsewhere too:
- Anthony Bennett – – Cavaliers – – 2013
- Markelle Fultz – – Sixers – – 2017
- Kwame Brown – – Wizards – – 2001
- Darko Milicic – – Pistons – – 2003
- Adam Morrison – – Bobcats – – 2006
- Michael Olowakandi – – Clippers – – 1998
The explanation usually given for such a sour turn of events is that scouting and projecting player performance is not a science; it is more akin to an artform. And maybe that is the case and maybe that will be the case forever. Except …
- Maybe avoiding huge “Draft Busts” is beyond the normal skill set of the people who wind up making Draft decisions – – but what about Artificial Intelligence?
Current AI models learn by reading and agglomerating “knowledge” from a broad set of sources. In doing that AI models like ChatGPT or Gemini or CoPilot can easily compete and defeat humans in trivia contests. But suppose a team hired a cadre of elite AI coders – – a Coding Cadré of “CC” – – that developed a player model based not on comments by observers but on measurables demonstrated by actual successful pro athletes. Granted, I have no idea how the members of “The CC” would measure competitiveness levels or emotional stability for potential draftees, but they might be able to use micro-measurements of successful players’ on-field action(s) to build models that may compare apples to apples.
NFL Players go through a standard set of drills at the NFL Combine. Someone somewhere decided that those drills were the “correct” means to measure the basic skills needed to play effectively in the NFL. I submit, respectfully, that the truncated list above demonstrates that not to be the case. Remember, all those players running a 40-yard dash and doing a standing broad jump are performing without any opposition. Such is never going to be the case on the field in an NFL contest. Perhaps what is needed is far more precise measurement and correlation analysis of data that never shows up in the game box scores. For example, in MLB, people now measure the spin rate of each pitch; no one did that in 1980; are there analogs for football and basketball that remain untapped in 2025?
For example, look at highly successful NFL QBs and ask how close do defenders have to get to those QBs before they begin to abandon the pocket and “scramble”. Does that measurement correlate with anything else that might be measured such as the time each QB needs to move his arm through a throwing motion? And are either of those measurements or a combination of those measurements indicative of completed passes? To be clear, I am NOT suggesting those measurements are of any value in building an AI Draft Model, but I do wonder if there are a series of measurements that might make the “Draft Process” a lot more “science-like” and less of a “crapshoot”.
Here is a measurement I would use as a starting point for offensive linemen, defensive linemen and edge rushers:
- How fast do those players react to the movement of the ball by the center to initiate a play?
I am talking about measurements in the hundredths of a second here. Who gets “off the ball” most quickly because in the extreme, that is a success strategy for those position players. So, I might begin by asking “The CC” to find correlation among that measurement and other measurements in highly successful linemen that could then be applied to potential draftees using game films where there are opponents out there trying to thwart the potential draftees’ intentions.
Maybe all this is fanciful; maybe this is beyond AI capability. Nevertheless, it might be a more interesting use of AI as compared to creating fake videos of a political candidate having carnal relations with a goat.
Finally, lest one gets carried away with my rosy projection here for AI involvement, take into account this from neuroscientist, Vivienne Ming:
“AI might be a powerful technology, but things won’t get better simply by adding AI.”
But don’t get me wrong, I love sports………