Just some random (and not very constructive) thoughts in absolutely no particular order.
1. I would guess that a time series analysis would be more appropriate here.
2. Predicting things that are directly influenced by humans is complicated. It might be best not to attempt to predict a very specific number, but rather give a confidence interval or a probability. That way, you are technically never wrong
3. Maybe the DR change per race would be better data than DR change per day (unless you drive the same amount of races every day). I'm fairly sure you don't have access to such data though.
4. As others have said, your DR will probably plateau at some point simply because the competition you face gets stronger. One way to quantify that "competition" factor, is to analyse other people's DR profiles and see if there is a common pattern. This data will most likely be polluted by those damn aliens, who simply win every race and whose DR practically increase exponentially.
5. Regarding the 75k points limit, looking at the DR rankings, some 75k drivers are ranked higher than other 75k drivers. So I wonder if PD actually does not cap DR for internal rating purposes and simply continues to add up DR points with no upper limit.
This is a great exercise though
Having all of that data (thanks to people like jasguer and Milouse) and doing nothing with it would be a waste.
