Windy City Gridiron
We developed a sophisticated AI model to help with the Chicago Bears coaching hire
For the fifth time in the George McCaskey era that started in 2011, the Chicago Bears will be looking for a head coach. After failing miserably on the first four hires, one would assume the Chairman would be looking for any and all information to help him with his search.
In aid of this search, this most crucial of hires for the Bears, I built the HALAS 2025 model. HALAS, so named after the founding father, George “Papa Bear” Halas, stands for the Head coach Approximation of Legacy Analytic Standard. The inputs used to develop this rich, deep, almost life-like model include 105 years of NFL historical data, weighing more recent years higher than decades in the past. However, because HALAS cares deeply about legacy, extra weighting is used in cases of dynasties like the 1940s Bears, Lombardi’s Packers, 1970’s Steelers, Walsh’s 49ers, the 1990s Cowboys, and of course, Belichick’s Patriots, and Reid’s Chiefs.
Do you want the Bears to be merely competitive, or do you want to make a hire for immortality?
We then feed that data into a model that uses advanced AI tools and techniques that can simulate a multiverse of scenarios with economic levers like salary cap increases and the impact of NIL money on the future readiness of college athletes, down to even societal levers and advances in technologies that can impact outcomes. This model is so advanced that the United States military uses a version of it to predict global conflict risks.
Some might suggest that using a model of this sophistication may have been a waste of time and resources, but this is the Chicago Bears we’re talking about. They need all the help they can get, and this is the future.
After calibrating the model, a tedious process that adjusts the values of the inputs to better mirror real-world observed data in the outputs to a 0.97 or a 97% accuracy rate, I called it “good enough.” Pushing for even an extra percentage point would have taken additional months of work, and my deadline was today.
It’s important to note that this is not 100% accurate (no model is), but we solve for that by running a multitude of scenarios with this well-calibrated model, which will provide us with the most likely successful coaching hire. We’ll then add up the number of scenarios each coach “won” and determine who the best hire is for the Bears.
Once the model calibration was complete, I ran the model one million times. While some might say that one million scenarios are too many, I needed to be sure. This is the Chicago Bears we’re talking about. No stone was left unturned, and while the multi-verse is infinite, one million was a conservative estimate to use based on the prevailing literature. Again, we’re just doing our best to simulate reality, not attempting to identify all infinite possibilities.
After plugging in all of...
