Week 12 GFT NASCAR AI Driver Rankings: Chase Elliott Moves to P1 after win at Dover
5 May 2022
By Bob Francis, Managing Editor
Go Full Throttle Racing News
It was a great week for the GFT AI models as our “SkyNet” NASCAR models correctly predicted Chase Elliott as the winner at the Monster Mile in Dover, DE. Elliott was sitting P2 in the GFT AI Driver Rankings after Week 10 at Talladega, but the algorithms detected something in the data and delivered the winning prediction.
From our April 28, 2022 article: Week 10 GFT NASCAR AI Driver Rankings: Byron Still P1 after Talladega
Predictions for Dover
With no data on the Gen 7 car at Dover and little comparable data, the predictive model confidence level is in the medium zone. The data says the 4 Hendrick cars, who were all strong here last year, have a higher probability, as do Blaney and Kyle Busch. Specifically, the GFT Prediction Model says Elliott is the favorite to win.
Big Movers after Dover
Moving up 4 spots to P11 after his 4th place finish at Dover is Christopher Bell. With a 3rd place finish, Ross Chastain moves up 2 spots to P3 in this weeks rankings. Losing 2 spots was Austin Cindric after a disappointing 36th place finish (last place) and the first “victim” of Miles the Monster.
Cole “The General” Custer is in
With a 15th place finish at Dover, Cole “The General” Custer breaks into the GFT NASCAR AI Driver Rankings. Both Byron and Blaney had difficult days finishing 22nd and 26th respectively, but each only fell 1 position in the rankings as a result of their overall AI score.
Predictions for Darlington
As with Dover, there is no data on the Gen 7 car at Darlington and little comparable data, therefore the predictive model confidence level is in the medium zone. However, if we push the models to pick a winner, “The machines” like Kyle Larson as the favorite for Throwback Sunday.
How do the Go Full Throttle AI models work? (Repeated for context as we have new subscribers joining each week)
The Go Full Throttle AI Driver Rankings is a cloud based predictive analytics system that uses our proprietary algorithms utilizing artificial intelligence and machine learning technology to dynamically tune and improve accuracy over time. Data feeds into our system include our Pre-Season rankings and the data available from NASCAR and other trusted NASCAR Cup Series sources that include (but not limited to) inputs such as Driver points, Stage points, Stage wins, top speed, lap times, and laps led from all practice sessions, qualifying and the races.
How do predictive models work?
Can artificial intelligence and machine learning algorithms really make accurate predictions? YES! In fact, our proprietary models accurately predicted 8 out of 10 F1 race wins in 2021, including Max Verstappen’s Driver Championship over Lewis Hamilton. Our NASCAR model was equally as accurate correctly predicting 7 of the drivers making the Round of 8 in the 2021 Cup Playoffs as well as predicting Kyle Larson winning the Cup Series Championship. These models will only get “smarter” in 2022 as more data and “learning” will improve accuracy.
Context is key!
“Artificial intelligence refers to the general ability of computers to emulate human thought and perform tasks in real-world environments, while machine learning refers to the technologies and algorithms that enable systems to identify patterns, make decisions, and improve themselves through experience and data” according to Columbia University’s School of Engineering. But Microsoft’s definition breaks this complexity down even more: “AI and machine learning enable companies to discover valuable insights in a wider range of structured and unstructured data sources. For better, faster decision-making, data scientists use machine learning to improve data integrity and use AI to reduce human error — a combination that leads to better decisions based on better data.” For us motorsports fans, we might use the analogy that artificial intelligence (AI) is the car and machine learning (ML) is the engine — you need both to win.
Dynamic tuning to improve accuracy
When we say “dynamic” we are referring to “near real time” data continually feeding the models. Besides the obvious final race results, as an example, our system will “watch” a number of data feeds looking for trusted information, such as the F1 live leaderboard with lap times during practice or qualifying results, laps led on race day, and final results.