MULTIPLAYER Update 0.6 - MP & Ratings Rebooted

Very hard to tell, to be honest. I know what the math does, and I know how the rating reacts to my driving (and my desperate efforts to drive differently, which is ridiculously difficult).
I probably will never say "it's perfect", but have some facts I collected about the CC in total

1) working with debug output (so overly informative) for over a year now made me a much better driver. Actually I'm shocked about a) my pace, despite that I do only testing and not really driving/training - and b) my consistency, which comes from the fact that the CC rating heavily punishes overdriving. So in total, I think the approach and the math makes a lot sense

2) looking at data, an incredibly low amout of drivers does stay on or below the limit. Most people overdrive or overturn, trying too hard

3) very good drivers and aliens (according to laptimes and hotstint results) are rewarded with close-to-perfect CC ratings, except they go for Top10 in the hotlap events. It appears that the fastest possible 1-in-a-million-lap is driven above the limit CC sets, and still delivers the best laptimes - but that's probably only true for a few dozends of aliens out there.

4) Back in our Spa competition in 2018, I've seen two 100 CC laps filed during the event, and I had the luck to observe one live. It was a top3 laptime in the end, Kamil would even have beaten Enzo's time with his lap if there wasn't two minor line mistakes (which CC does not look at, only how well you use your tyre's capabilities). This is a bit weird and contradicting the 3) observation, but the guys possibly didn't drive 1-in-a-million-laps in an event where you have to wait 1h for a 10 6 minute timeslot - and sitting at home with one month to go changes this
Edit: Link for some impressions about the Spa event;
And Enzo's winning lap had a 97 CC rating

5) We also witnessed very good drivers having bad CC rating, which made them understand that they can become better when taking it a bit easier. Reference story is this, absolutely stunning where a factual alien with lots of experience changed his style a bit and ended up winning several seconds of laptime, arriving in the Top10 of that leaderboard. You won't find many simracers who are brave enough to admit this in public, but I have at least 30 private messages with similar reports.

6) Rain: For light/medium rain situations, for 90+% of the drivers the CC ratings actually go up. The reason is that that most drivers find theirselves (slightly) above the limit, and twichy rain conditions force them to drive more carefully and not overdoing it. While a slick on dry easily forgives 30% of overdriving (which induces other problems, and is not really faster), this won't work in the wet. So you could say wet forces you to drive better, and the CC rating reflects that.
However I read many reports that rainy races do drop the CC rating; atm. I'd like to investigate this. Realistic outcomes:
- there is a bug
- it works, and if the CC drops during rain it means you have to calm down in wet conditions (which is good news as you have an easy way to become faster)
- the race makes you lift often, and not use the tyres potential which reduces CC. That will become better with some adjustment, but atm I simply do not have the data
- there are side effects, like having too high TC/ABS settings derived from the wet setup. CC will not like TC/ABS settings that are *way* off

7) You probably can already see this based on my answers; the adjustment process sadly is very slow. The rating works very well for my own performance level (which is around 93 CC, roughly equals a bit more than 1s off the top times), but I need observations, reports and data both for the high-level aliens *and* the many variations of less quick and experienced drivers. Interestingly the 2nd category is much harder, but way more important (as we have hundred thousands of drivers like that). We have quick drivers, real drivers, real aliens in our testing realms - but beginner levels are quite very tough to investigate.

 

I always had this intuition, and on several occasions the reality baffled me.
Back when the prototype for the CC rating was invented - that was way before I was on the list to the Kunos team and realized in an AC1 app - I had very interesting results for my own driving, and also massive improvements in my driving. But then I started to ask a two aliens, which I knew one was known for being super smooth, and the other one was super quick but also had a very aggressive style. Turns out - nope, those words describe something else. Yes you can watch both replays and tell who is who just by watching how they drive, how their steering is and how they unsettle the car.

But already in this very rough algorithm, this prototype did look at the correct things, and in the result data you could not distinguish both drivers anymore. Both had a close to perfect 99.x % "rating" (although this isn't comparable to what we today know as CC). I think I spent hours to check the data and watch the replays over and over again, because I couldn't believe it. I though my driving was really good, I think I had the data for a 2:15.9 lap (R8 LMS at Spa after trying for 2000 laps). But their ability to go that fast without overdoing it was incredible - while I immediately understood how much time I actually lose on my "perfect" corner entries and transitions between braking and turning.

In a nutshell: No, although my intuition says the same, my experience tells me Alonso would score very high CC values, just like any other real driver (*all* of them were very good at CC once they chose to driver, opposed to play a game).

 

k then I know what to check for - this shouldn't be possible. While a good lap will let your CC jump immediately, the decrease when you keep doing worse laps should be quite slow.

This is how it works

 

No, for exactly that reason

I see a pattern here
Ya, this should be completely impossible.

 

That's what I wanted to express with statement 5) - you most probably are in the situation that your CC is not in line with the rest of your driving performance. It's just that nobody admits that a stupid rating system gave them a hint how to improve what years of simracing did not reveal. Not saying the rating is 100% sharp here, but there's a significant chance.

Well I didn't tell you about the linearity of this rating.. The distance between 96 and 99 is already kilometers Out of the head roughly as much as 70-80.

Very good, then you know how to do it, and there are no systematic flaws. You just need to focus on delivering more and more 90'ish laps, and less of the others
I guess 90 is a very reasonable target for anyone