Driver Performance Index

[przemator]

DRIVER PERFORMANCE INDEX is a set of statistical figures calculated to determine the skill of drivers racing in NSGP Online Championship. From the beginning of season 7, it is updated on a separate website.

[przemator]

The Driver Performance Index uses an Elo-based formula to calculate driver strength.

• The default/average rating of a driver is 1000 points.
• The K-value for each race is 30 * sqrt(number_of_drivers - 1).
• A difference of 100 points between two drivers means that driver A will be better than driver B twice as often. So if driver A has a rating of 1200 and driver B has a rating of 1100, it is fair to say that driver A is two times better than driver B.
• The rating does not decrease due to absence, but inactive drivers are moved to a separate list.

[Alex Reeves]

I like it! One possible modifier - a differentiator for race importance based on average rating of drivers i.e. a race with an higher average rating of players should be worth more than one with lower. I suspect this would be a modifier on step 3.

This would add an elo-like relative performance rating to the race, but without the dramatic elo penalties for good/bad performance.

[przemator]

I like it! One possible modifier - a differentiator for race importance based on average rating of drivers i.e. a race with an higher average rating of players should be worth more than one with lower. I suspect this would be a modifier on step 3.

This would add an elo-like relative performance rating to the race, but without the dramatic elo penalties for good/bad performance.

Well, this leads to a cyclical reference, because we want to calculate an index using the average of indices. I should remind you, that after each step I saved the results to a different sheet, because they were lost after after the modifications. There are no real "steps" in the procedure, everything is done at once.

A possible solution to the cyclical reference problem would be to calculate the new index using old indices, which would have to be updated after each race. I will look into it.

[Brad]

Well, this leads to a cyclical reference, because we want to calculate an index using the average of indices. I should remind you, that after each step I saved the results to a different sheet, because they were lost after after the modifications. There are no real "steps" in the procedure, everything is done at once.

A possible solution to the cyclical reference problem would be to calculate the new index using old indices, which would have to be updated after each race. I will look into it.

I get a feeling your good at maths. Feel like doing my maths exam tomorrow?

[Cookie Monster]

Maybe you should stop mixing revisions and Spotify? @Brad

[przemator]

I made a lot of mistakes in my formula, I hope it's fixed now.

[Sida79]

A-w-e-s-o-m-e. May I have the excel file for this one as well?

[alexeilh]

Top ten oh yeah \o/

[Alex Reeves]

Index updates after race 1 plskthx?

[przemator]

OK, I updated the DPI with the results from S04R01. First of all, I have dropped my formula as I had some doubts about it. It's back to Elo now. I have adjusted the values: the default is now 1000, the K-value is 30 and if you increase your rating by 100 points then it means you doubled your skill. I also moved inactive players to a separate list.

Check the first post for current list.

If you want the excel file, it's here: DPI.XLSM

[Alex Reeves]

What was wrong with your index? Not that I dislike the ELO way of doing things, but I thought your new formula was quite clever in how it introduced aging and a slightly different way of creating a race score.

Although I guess pressure's on me now to live up to my no.2 ranking...

[Sida79]

Ah, same question as Alex.
Thanks for the file!

[przemator]

The formula was a bit faulty. The weight of the race was always a sum of ratings, no matter if you won or came last. So, coming last in a weak race would be less harmful than coming last in a strong race.

[przemator]

Time to refresh the list. I marked many players as inactive, namely: Blazej, Alex Lloyd and others, for the sake of clarity. Maybe it will be a little motivation to show up next time .