Driver Rating was created by NASCAR in 2005 to provide fans with a metric to identify driver performance beyond qualifying and finishing positions. Below is Driver Rating methodology and a critique of its short-comings.
Driver Rating consists of 3 segments, Primary Statistics, Fixed Bonus and Variable Bonus.
Finish and Average Running Position While on Lead Lap points are calculated based on the 2004-2006 NASCAR points system (i.e. 1st=180, 2nd=170…43rd=34). The Average Running Position While on Lead Lap points are rounded to the nearest finish position points (i.e. 3.4=3rd=165 etc.)
Average Speed and Average of Fastest 3 Laps is based on MPH. Average Speed is calculated from laps on lead lap under green. Restrictor plate races calculate average speed from corner speed in place of the total lap speed average.
The Driver Rating allots points for various step categories based on finishing position.
The Variable Bonus is calculated by adding laps led and fastest laps on green flag laps. This number is divided by the driver’s total number of green flag laps and multiplied by 100.
Points from Primary Statistics, Fixed Bonus, and Variable Bonus are added together and divided by 9 to find the Driver Rating.
Critique of NASCAR’s Driver Rating
My critique of the Driver Rating formula that NASCAR uses is primarily rooted in elements of the Fixed Bonus section and the use of the 2004-2006 points system in the Primary Statistics section. These elements induce noise by not smoothing data. This reduces its value for predicting future outcomes. A random look at correlation between Driver Rating and Finishing positions shows an R2 between 0.7 and 0.85. Removing the noise NASCAR has induced in their Rating would likely put R2 consistently in the 0.9s.
Another critique of the Driver Rating formula is that it actually has a maximum Driver Rating of 166.667 based on the information that NASCAR has published, not 150. This is chiefly due to a miscalculation with the Variable bonus. The Variable Bonus counts green flag laps led and green flag fastest laps. This means that there is a possibility of having two times the amount of green flag laps, which makes the maximum points for Variable Bonus 200, not the 100 NASCAR states.
For my modeling purposes, I disregard NASCAR’s Driver Rating and use my own Race Index to assist with predictions. R2 for the correlation between Race Index and finish positions ranges from 0.9 to 0.97. This is a substantially better fit that Driver Rating. Below is my methodology.
Race Index = (FI+API+PCE+LLI+FLT+RCI)/10
n = number of cars in race
Qualifying Index (QI) = (n-Qualifying Position)/(n-1)
- QI is only used as a function to find the PCI
Finish Index (FI) = (n-Finishing Position)/(n-1)*2
- FI is weighted 200%
Avg. Position Index (API) = (n-Avg. Position)/(n-1)*2
- API is weighted 200%
Position Change Index (PCI) = ((1-Qualifying Index)-(2-Finish Index)
- PCI adds or subtracts from a driver’s index depending on how many positions they gain or lose during a race
Laps Led Index (LLI) = Laps Led/Driver Total Laps
Fastest Lap Index (FLI) = Fastest Laps/Driver Total Laps
Race Completed Index (RCI) = (Driver Total Laps/Race Total Laps)*3
- Race completion is weighted 300%