How does SR affect DR for ranking

[fastone371]

Your door number is not your expected finishing position, ever, no such thing exists in the game.

You gain and lose points by beating or losing to other drivers, beat S drivers as a lower rank and you'll gain more, lose to enough lower rated drivers as a S rated driver and you'll lose a lot of DR points but that is a simplistic way to look at it, the other threads explain it much more.

In any case it's not really something anyone (other than people who are testing how it works) should concern themselves with.

OK, thank you, I guess

 

[Voodoovaj]

Well that's not "affecting DR ranking", that's just matchmaking. @Famine already explained how SR can affect that pretty clearly.

Matchmaking is a black box that we can't experiment with, so we'll likely never know exactly how it works. It does seem to do different styles for different races even when my SR remains at 99. One factor in which style it chooses is certainly the number of players available, but I've also had it flip back and forth between well-matched and not, at times when I wouldn't expect it to be tallying with the number of players available.

But in all cases it always seems to match primarily on SR. There's no indication that it uses the results of some imagined magic_function(sr,dr) for its bucketing.

My bad on terminology, I meant matchmaking.

To me, it seems it does have a "magic function". It seems too consistent. It seems a little too predictable that I drop a few SR points and there happens to match me with the lower DR's. I get it that there is likely more people in the 80-90 SR range in the C and D ranks, but that doesn't explain how dropping SR rank to A or B will still match you up with DR C's and DR D's with SR S ratings.

If the matchmaking were just trying to keep people of the same SR together first, then an SR B player should never see an SR S player of any rank.

It would be easier to execute matchmaking using a single value rather than two values.

 

[Outspacer]

My bad on terminology, I meant matchmaking.

To me, it seems it does have a "magic function". It seems too consistent. It seems a little too predictable that I drop a few SR points and there happens to match me with the lower DR's. I get it that there is likely more people in the 80-90 SR range in the C and D ranks, but that doesn't explain how dropping SR rank to A or B will still match you up with DR C's and DR D's with SR S ratings.

If the matchmaking were just trying to keep people of the same SR together first, then an SR B player should never see an SR S player of any rank.

It would be easier to execute matchmaking using a single value rather than two values.

I wouldn't know, I haven't ever dropped an SR rank But seriously, I have watched quite a few races on youtube when I was looking into DR scoring, and wide variations in SR are pretty uncommon at any level. Just looked for races now, and found a race with a mix of SR A and B, with DRs of B and C. That is certainly what's typical, and what you'd expect if it was going by SR primarily.
edit:

Stream ended, but basically here's a player who's SR is falling, and getting matched appropriately - starts with him at SR A with mix of SR A and S, then a race with all SR A, then the race described before (mix of SR A and B) when he's become SR B.

With DR scoring we worked it out because we could see the input and output, and figure out what went in the middle. With matchmaking, we can't. At any time you only see one race of the many it's matched, and we have no idea of the input - how many players there are and what data from each that it uses. It could well be using more than simply SR/DR to match on. And it's pretty clear that it isn't a simple sort and then fill rooms starting from the top of the list. For example, when I'm the rabbit in one race there must be other rabbits in other races - I've had times like that where I've been with the other rabbits in the races before and after, and that change in style hasn't been because my ratings have changed.

So if you have a theory that's great, but don't forget to look for the counter-examples

 

[John Crellin]

OK, thank you, I guess

DR was explained in this thread:

https://www.gtplanet.net/forum/threads/how-exactly-does-driver-rating-work.363037/#post-12203023

 

[Voodoovaj]

I wouldn't know, I haven't ever dropped an SR rank But seriously, I have watched quite a few races on youtube when I was looking into DR scoring, and wide variations in SR are pretty uncommon at any level. Just looked for races now, and found a race with a mix of SR A and B, with DRs of B and C. That is certainly what's typical, and what you'd expect if it was going by SR primarily.
edit:

...For example, when I'm the rabbit in one race there must be other rabbits in other races - I've had times like that where I've been with the other rabbits in the races before and after, and that change in style hasn't been because my ratings have changed.

So if you have a theory that's great, but don't forget to look for the counter-examples

You haven't noticed that this "style change" coincides with certain DR and SR values? For instance, you will always be the rabbit if you are less than 95 SR.

Everyone takes an unintentional penalty. See if the style of matchmaking changes for you after that happens.

 

[Outspacer]

You haven't noticed that this "style change" coincides with certain DR and SR values? For instance, you will always be the rabbit if you are less than 95 SR.

Everyone takes an unintentional penalty. See if the style of matchmaking changes for you after that happens.

You honestly think I wouldn't think of that?! Like I said, that usually isn't because my ratings have changed... my SR stays at 99 nearly always (112 of the last 129 races, 30 of the last 31), and my DR doesn't change by much these days.

 

[Outspacer]

Hey @Voodoovaj - pure coincidence, just caught you racing with DooDoo who was streaming

So here's some data for you in case you're interested...

Seems to me that matching is essentially on SR for this race.

Spreadsheet with all the DR calcs...

 

[Voodoovaj]

Hey @Voodoovaj - pure coincidence, just caught you racing with DooDoo who was streaming

So here's some data for you in case you're interested...

View attachment 727112

Seems to me that matching is essentially on SR for this race.

Spreadsheet with all the DR calcs...

Ya, that was after my disastrous Monza race.

That "matchmaking" looks awful. How do you match a person with a DR of 654 with people who have 30k-50k DR? That makes no sense.

 

[Outspacer]

Ya, that was after my disastrous Monza race.

That "matchmaking" looks awful. How do you match a person with a DR of 654 with people who have 30k-50k DR? That makes no sense.

It is crazy. You didn't have any complete newbies in that race, but easily could have since SR starts at around that level. I was actually looking out for DR resets due to SR falling, but no chance there - most were +SR at the end despite the rough nature of the race, some even made it straight up to 99 SR. It isn't even closely matched on SR, but maybe that's just a sign of how random it's become.

So I don't think a fn(sr,dr) is in play. It's not a bad idea though, since it could be tuned to avoid matching like that race. I still think it should weigh mostly on SR, but perhaps slacken off if the range of DR exceeds some amount. But it's all moot while SR is varying so wildly

 

[Voodoovaj]

It is crazy. You didn't have any complete newbies in that race, but easily could have since SR starts at around that level. I was actually looking out for DR resets due to SR falling, but no chance there - most were +SR at the end despite the rough nature of the race, some even made it straight up to 99 SR. It isn't even closely matched on SR, but maybe that's just a sign of how random it's become.

So I don't think a fn(sr,dr) is in play. It's not a bad idea though, since it could be tuned to avoid matching like that race. I still think it should weigh mostly on SR, but perhaps slacken off if the range of DR exceeds some amount. But it's all moot while SR is varying so wildly

I recall some of the names in that race you posted. Looking at those DR scores (1, 150, 469) it makes sense to me why it was a circus. There SHOULD be function in place to arrive at an overall score for matchmaking and I, foolishly I suppose, expected them to implement a system like that to keep some of these matchups from occurring. That race highlighted the STARK difference in how penalties work. Those lower DR drivers were playing as if there were no penalties at all.

 

[Outspacer]

I recall some of the names in that race you posted. Looking at those DR scores (1, 150, 469) it makes sense to me why it was a circus. There SHOULD be function in place to arrive at an overall score for matchmaking and I, foolishly I suppose, expected them to implement a system like that to keep some of these matchups from occurring. That race highlighted the STARK difference in how penalties work. Those lower DR drivers were playing as if there were no penalties at all.

Probably because they don't get anything like the same penalties!! Such a bad decision from PD to have a hard cut where it changes behaviour.

I'm sure the matchmaking would be great if there were millions playing each night... but it's a few thousand at best (50k globally in 24 hrs, split across 3 regions, multiple time zones, varying play hours, and the 3 daily races plus fia). So many more players here in Europe, but still it comes up with very mixed fields outside of prime-time.

Anyway, here's the race from DooDoo's POV:

 

[Voodoovaj]

Probably because they don't get anything like the same penalties!! Such a bad decision from PD to have a hard cut where it changes behaviour.

I'm sure the matchmaking would be great if there were millions playing each night... but it's a few thousand at best (50k globally in 24 hrs, split across 3 regions, multiple time zones, varying play hours, and the 3 daily races plus fia). So many more players here in Europe, but still it comes up with very mixed fields outside of prime-time.

Anyway, here's the race from DooDoo's POV:

Well, you can see how my race went right at the start