Hi! I’m the person who did the analysis, thank you for sharing it here! It was a very pleasant surprise to see something mine in the forum in which I spent the last years lurkingclownfish wrote: ↑27 Apr 2022, 15:45Your thoughts on this? (There are further details on the approach here: https://www.buymeacoffee.com/F1DataAnal ... s-mercedes )
The analysis is my first one of this kind, and is still relatively rough, but still I tried to take all the reasonable steps to improve its accuracy (taking the local track slope into account, considering only straight sections in order not to have additional drag due to cornering forces, appropriately filtering the data etc…)
The points considered are relative to all the DRS open sections, with >=99% throttle, brake=0, rpm>10000 and ax>-0.5m/s^2 (to avoid clipping, but to include both signs of the noise to avoid a bias). The lap considered is the best one in qualy for that driver.
Main limitations are: power is considered constant (while it changes with the rpm, although slightly due to the fuel flow being capped above 10500rpm, and with speed due to the energy deployment strategy), and the drag coefficient too (and it too changes with speed, due to squatting etc). At least considering only the DRS=open sections reduces its variability, and with DRS open the drag reduction due to squatting is lower (less downforce to push the car down, and less drag to be removed). What I get is, we could say, the average P/m and D/m values for each car through the straights: in reality they vary with speed as I said, but after all they are the single value that best describes the straight-line performance, which is what matters! The values provided (2.0% and 1.6%) could be misleading: these are the values, rounded to the first decimal number, that I got, but this surely does not mean that the accuracy is in the order of 0.1%! I believe the accuracy to be in the low single-digits, I will try to estimate a confidence interval for the next analyses.
Thank you again!!