Discussing ACC Safety rating

Ha! Now I know I'm not going senile...

In the past 4-5 days (maybe longer, it would seem that the driver ratings and stats data have been erased / lost / vanished. Wondering if this might be a sever issue, can't imagine what else it could be...
Things that make you go..hmmmm

 

Yes, stats are disappearing in the middle of the night...does it MEAN something?

 

Hello to all,

I hesitate to ask this question here but I need to understand how Safety Rating works on ACC.

To summarize my understanding, it depends on the distance covered with other drivers on the track without having an incident, all the while staying relatively close to the other cars, whether they are in front or behind. The exact distance is quite nebulous but let's say that as long as we are, in principle, less than 0.5s behind, the SA should go up.

So I did a session to get my SA back up (from 89 to 79 in one evening after 4 incidents, I find that ultra severe but not the point) either because of my own mistakes (a slight touch on braking behind an AI that had curiously braked way too early) or because of the rather strange behavior of some drivers or AIs (for example an AI that clearly divebombed when it clearly had no room to pass) For this session, the AI was set to 100 with an aggressiveness of about 50 (I don't remember exactly for aggressiveness). I followed a car for 18 laps, without ever touching any opponent or any off-road or anything abnormal, we can say that I was driving "quietly" behind an AI in short...yet, my SA didn't increase at all, staying at 79 after these 18 laps but even worse, on some laps my SA was falling to 78!!!

My question is therefore to know if it is normal that in such circumstances (and I assure you that everything happened exactly as I describe it, I am not inventing anything but of course I did not record the session so I cannot prove it) not only we do not progress at all at the level of the SA but that in addition we can lose some whereas we did not have strictly any incident, neither near nor far?

 

From my experience, but multiplayer only (I don’t race against the AI and in the very few occasions I did, I have ratings disabled): SA will be gained by racing other drivers, not simply following them, i.e., I need to actively race, trying to pass, challenge race positions, fight for turn in advantage, etc.
Simultaneously, racing must be done within the track limits, so no cutting nor going wide off track (even if just time deleted and not a full cut), no contacts, obviously, and without losing car control - locking, spinning, drifting, etc. (SA and CC are interlinked for SA gain purpose).

So, SA will be gained by truly racing other drivers, going right on the tail, moving to pass or try to pass, going side by side, etc. without contacts and with car in control.

Edit: I don’t remember surely, but, if I’m not mistaken, SA will increase slower offline than in multiplayer.

 

Ok, let's say it is better to actually "race" against others, I can understand this (but losing 1 or 2 points for a tiny touch while requiring A LOT of laps fighting with other to get it back is...well, let's say "not optimal") but how can I LOSE SA despite I :

1) didn't touch anything or anybody
2) have 100% of valid laps
3) did not spin, not drift, not missed apex, nothing...
4) made 18 laps...in a raw...

...all those laps were done behind an opponent, between 0.2 and 0.8s...

Isn't a bit strange ?

 

Like I said, I don’t have a reliable experience offline, but, basically SA will grow with mileage as long you’re in hot racing, without contacts - or few incidents for long mileage - with car control (CC) and without causing accidents, so achieving a good trust is paramount.
The other contribution for improving SA is racing accidents severity and consequences.

The colder the race is the less trust you earn (or none at all) and (I’m guessing), by increasing your mileage without adding any trust your SA rating can eventually start to drop as racing distance increases and no trust is gained, i.e, the ratio of trust gained by mileage will start to decrease and SA can drop slightly consequently.

Furthermore, if your trust is low or non existent for several miles and you are caught in an incident some points are still going to be deducted to your rating (more or less points will be deducted depending of the severity of the incident and how “guilty” you are).
By not having hot races where you can grow your trust relatively fast, you’ll need a fair amount of laps of just warm racing to regain the points lost in those incidents.

 

I read you but I clearly don't understand how driving for 18 laps right behind a car (100% AI so not a slow car by any mean, not the fastest but still not slow) without any incident of any kind could make you actually LOSE SA...I'm pretty sure it shouldn't be possible, at the extremely worst you shouldn't gain anything but losing SA ? C'm'on...

 

Considering (at least how I look at it): SA =(trust gained - racing incidents) / distance travelled; if you don’t add any trust, even if you don’t have any incidents, as the distance increases (adding number of laps), could lead to a slight decrease of SA.
Probably if your laps are only limited to following the car in front, you’re just adding mileage and SA can drop 1, but if you do the same laps actively racing, continuously challenging the other car(s) going for an overtake, your SA would increase by 2, 3 or even a bit more, as your trust skyrocketing.

 

As you can see on the SA rating graph, older races are phased out of the equation while new races get factored into your SA rating. So, what likely happened to you is that old races , where you gained a lot of SA, were no longer included in the calculation. So, you basically lost SA rating not because of your recent good races, but because of older races.

Also, it does not matter, if you actively race an opponent or just follow a car, trust is calculated by driving near other cars, if one is in front of, next to or behind other cars does not matter.

Maybe you should read this thread to gain more insight: https://www.assettocorsa.net/forum/...acc-rating-system-part-ii-long-version.51975/
Although some details regarding the severity of collisions have changed, this explanation by the guy, who wrote the SA rating code for the game should really answer all your questions.

 

I understand what you are stating but I'm still amazed that driving 18 laps (ok, Zolder so small track) right behind a car actually makes you lose SA, this is beyond my understanding, sorry. Of course if I was to drive without any car close to me I would understand (would still be stupid to LOSE safety rating while not even touching anything but why not ?) but how can I lose SA in this situation ? I'm not asking why calculus is making me lose SA in this situation, I'm asking WHY this calculus is that stupid and counter productive ?

SA is about safety, you shouldn't be able to lose any SA if you are driving correctly without any incident, yet not only I was not causing anything but I was clearly very close to another car, for a very long period and was still losing SA, how logical is this exactly ? I'm sure this is not what should be.

 

Sure, maths logic is clear but frankly, is it a clever system ? I am losing SA despite I'm very clean just because I drove several laps far from others and all of a sudden, I have some incidents (most of them weren't even my fault but like in the real world, if your car is involved in a crash, no matter who was responsible, your car is still crashed so I have no problem with this logic) so the "trust" is gone and now, even when driving cleanly behind another car for a long period, I am still losing SA...how productive is it ? How fair is it ? I mean is it even accurate as a global system after all ?

imho, something is terribly wrong.

 

I don't see what else I can say about it.
From my side, I've never had any issues with how the ratings work. In fact, I think they should be stricter than they are, SA in particular, because I find it way too easy to keep it high.
Below are my ratings and I race exclusively online, and they are around like that for a very long time. With very busy races and/or with changing weather conditions, both CN and CC can drop slightly, but SA only comes down if I ever make a very huge mistake and take someone out.
Looking at my SA you'll see that trust is often at a level that allows me to keep my SA without any issues.

 

Yzangard,

On Luis's graph the bars to the left are going to be removed from the SA calculation.
From what you are saying, I presume that if you show your SA graph, you had some tall green bars that exiting left and had smaller green bars on the right.
Each green bar shows the KM that you raced closely to the car behind (0.5 sec I believe).
I also believe that racing against the AI has diminishing returns at higher SA.
If you ran 18 laps close to the car infront, then you should see the KM.
Zolder around 4KM per lap.
18 laps = max of 72KM for a green bar.
If the bar is less, then this is due to racing against AI instead of online.

The above is what I believe and since I'm married, I understand that most of the time I am wrong is what I say, do, think, act, etc

 

Not Everyone is serious on the game, I would worry about it.

 

lol, feel that way most of the time as well

 

Again, I understand what you are saying but I don't think it is fair, I made indeed some pretty long sessions some time ago but since I prefer rFactor 2, I was focusing on it for several months...and decided to give ACC another try (and frankly, I still enjoy ACC, not as much as rFactor 2 but still a very decent, more than decent actually, simulator) and made some very short stints with bots...problem is : bots with high aggression are pretty much...well, aggressive, to say the least, so they are ramming you very often. I would understand that you can't get as much SA with AI drivers but when they ram you, you are losing A LOT of SA, so basically you are gaining less but losing as much, with AI that are just not respecting you...let's say that's ok but still unfair to lose as much but gain less. My Issue is not with this actually, my issue is about not gaining anything on my next "long" session, even LOSING despite I didn't had any incident. Clearly not what should be.

 

Yes, pretty much my case. Again, I get it, the logic is what it is...flawed imho but still, this is the logic. Now I still can't understand what math can make you LOSE SA when you are driving a pretty long stint right behind an AI without any incident, I'm pretty sure this is not intended.

Note that I don't really care since I don't want to play online but I don't understand how the system can work that way, super counter intuitive and clearly counter productive...it just prevents me from going further actually, the more I am clean, the less I have SA, not sure it is the goal of the system...

 

You write you understand how the system works, just to contradict yourself two posts later by writing that you don't understand the math behind it and think it is not working as intended. Make up your mind, please. Which one is it? Do you understand the system and don't like it or don't you understand it at all and feel treated unfairly?

Anyway, I think the system is pretty good and fair, as it encourages to drive safely every time. As the trust and OBWP average over one's last driven x kilometres is taken into account, the system does not provide as much instant gratification as most games try to hand out to the players in forms of achievements and similar recognitions of skill or time invested to ensure player engagement. Instead it ensures trolls and rammers can't boost their safety rating with one good race against AI, after having rammed everyone over the course of several races before. Those rammers would have to exhibit clean driving for longer periods of time, effectively preventing them from ramming while building SA rating.
So, I believe the system works and does what it is intended to do: To keep most of the unsafe drivers from reaching SA ratings of 90 or higher.

 

What I meant was that I understood what was being calculated but I don't understand why it's calculated that way.

It's not the calculation that I don't understand, it's the logic of having chosen such a calculation.

 

Bruh, I mean i wouldnt worry......