2021 Dutch GP media ratings predictions

As stated in one of my previous posts, I will show the ratings that my model predicted for the 2021 Dutch GP. Unlike the previous 2021 Hungarian GP, I think this race was a bit more traditional, so hopefully we will see better results this time.

Just as last time, right now I’m just posting the predictions created by the model that I created a few races ago. Once the media outlets release their actual ratings, I will update this post to show the performance of our model.

As you may notice, I didn’t include Alfa Romeo in the current results. The model uses information about the driver’s previous race, and Kubica doesn’t have a previous race since he was a replacement driver. I didn’t want to retrain the model this time, but hopefully I will be able to include all the teams for the next race predictions.

Update

I was waiting for Sky Sports to upload their ratings but they never did. If they do, I will update the chart. In general, I think the model did a decent job. Sure, the predictions are not perfect, but they will never be. I am quite happy with it though, and I think that with more feature engineering I may be able to improve the predictions even more.

The predictions

Model performance

The model actually did quite a decent job for this race. The Race continues to be the most problematic outlet. Their ratings seem to be way too harsh or way too soft on drivers, and that makes it very hard to predict their ratings. Having said that, the model did a great job for the ratings of F1i. The average error was of only 0.58 , with almost 80% of the predictions being off by 1 points or less.

For Planet F1 the numbers are pretty good as well. 85% of the predictions had an error of 1 rating point or less, while for AMus, 80% of them had an error of 1 point or less. 

The worst predictions came from The Race and Crash. As I’ve said, The Race tends to be quite inconsistent, so predicting the ratings for them is quite hard. For Crash, 70% of the predictions had an error of 1 point or less, but unfortunately the model made some big errors, like predicting a 7 for Tsunoda when he actually got a 4.

Final remarks

First of all, thank you for taking the time to read this article. I will update this post once the ratings from the media outlets are released. No matter how or model did, good or bad, I want to show you that the world of analytics is not perfect, and that sometimes there are obstacles on the road.

If you want to support me, there are multiple ways you can donate some money to help me keep this project alive. You can find all the options on either the about tab or the my supporters tab in the main menu.

Latest posts