12-03-2020, 10:46 AM
(This post was last modified: 12-03-2020, 10:47 AM by Simone Saviolo.)
I think the gist of this has already been suggested somewhere, but I couldn't find it. So here is my suggestion.
Rating and Incident Average are currently a running, historic tally. I think that more weight should be given to recent results, to better represent what a driver is like now (or rather, what h'e been like lately).
Historic results mean that people who ran a lot and achieved a lot in the past, but are not currently active anymore, rule the standings and will never get down from there. Hypothetically, a driver could be somewhere around #100 in the overall standings, yet he could be the best guy around these days.
Or, someone who got a bad start with lots of incidents and now is a much more careful driver could have a high IA, even if his last races have been cleaner than that. The reverse is also true: if someone was very careful and he raced a lot in the past, he can easily look good while having an aggressive behaviour these days.
So, what I suggest is to give a lower weight to results older than, say, a month, and maybe even lower weights for other thresholds. For example, 0.7 if older than a month, 0.4 if older than a year, 0.1 if older than two years. Of course this would need to be adapted to the actual Rating algorithm, which is rightfully kept secret.
I realize that this may not be possible due to how Ratings and Incident Averages are computed and stored. But if it was, I think it would help people understand who they're racing against.
Rating and Incident Average are currently a running, historic tally. I think that more weight should be given to recent results, to better represent what a driver is like now (or rather, what h'e been like lately).
Historic results mean that people who ran a lot and achieved a lot in the past, but are not currently active anymore, rule the standings and will never get down from there. Hypothetically, a driver could be somewhere around #100 in the overall standings, yet he could be the best guy around these days.
Or, someone who got a bad start with lots of incidents and now is a much more careful driver could have a high IA, even if his last races have been cleaner than that. The reverse is also true: if someone was very careful and he raced a lot in the past, he can easily look good while having an aggressive behaviour these days.
So, what I suggest is to give a lower weight to results older than, say, a month, and maybe even lower weights for other thresholds. For example, 0.7 if older than a month, 0.4 if older than a year, 0.1 if older than two years. Of course this would need to be adapted to the actual Rating algorithm, which is rightfully kept secret.
I realize that this may not be possible due to how Ratings and Incident Averages are computed and stored. But if it was, I think it would help people understand who they're racing against.
