General AI Performance Discrepancy Reports (No Discussion)

GT3 too slow in the rain in Austria.

I qualified P4 on 113 strength. In the rain however (also in other classes than GT3, so maybe this is even something for the AI line redo thread, but I didn't notice this so much in the dry) the AI lose a lot (!) of time through the two long off-camber left handers, especially on the exit of the second one.

They also brake too early for the downhill braking zone before T5.
This is also something that happens in other classes and also happens in the dry.

 

Not sure if it’s been mentioned already here, but the ai line severely overlaps the pit exit line on the run up to T1 at Azure Circuit. This becomes a problem for the player when exiting the pits, as the ai cars will either ram you from behind at full speed, or phase through you completely, which is particularly jarring in VR!

it happened to me twice during practice/qualifying.

 

A thing I've noticed with AI when using DRS. The AI only seem to use DRS once, if for some reason they come off the throttle when the DRS is active, they don't seem to activate it again. Easy to see an example here:

 

Tried to do a race at Jacarepagua, 61 laps, car fully fueled for the entire race, but frankly lost motivation because:

F-ClassicG4 120 AI in qualifying: 1,21,0 vs me 1,20,1

F-ClassicG4 120 AI in the race on lap 2-3: 1,24,5 vs me 1,26,5

One of these needs an adjustment of three seconds. I don't care which, but AI still loses by far not enough time with high fuel load (or doesn't win enough with no fuel). I also wanted to test if they got slower towards the end of the race, but I crashed before. I guess they kinda have to be slower at the end - after all with no fuel in quali they also are a second slower than me.

 

Its not their pace, its the fuel weight issue, there are some classes in need of fine tuning.
RetroG3
HitechG2 (Haven't tested the G1)
ClassicG1&G4 (in the G1 its mainly the Lotuses that where OP before, haven't tested them in this latest release)

I had no issues however with the following:
RetroG1&G2
ClassicG2
F-V12
F-V10G1
F-Reiza
F-UltimateG2

Haven't tested the others properly, but will do when I get the chance. I think it might actually even be track dependent, but cant confirm, only that I had no issues with the HitechG2 at Imola, but it was pretty bad in Barcelona.

 

I posted this in the Bug Reports thread but it's perhaps better off in here, won't repeat so heres the link to a pretty major AI issue at Donington;
Bug Reports - Automobilista 2 (Read the Opening Post)

 

Baseline Strength: 108/Medium Aggression
Car Used: Group C/Sauber C9
Track: Barcelona 91
Session: Qualifying/Race

The AI is pretty slow here, specially on both fast chicanes on sector 2.

 

Great AI update for 1.5.3, the overall behaviour of the AI is really fantastic and most of the immersion breaking stuff like insane weaving, driving on grass etc are dialled right back. A lot of tracks have vast improvements - it's great to see the AI no longer eat bollards like smarties at Oulton for instance Still a few discrepancies (dry weather, quick races):

Bathurst/tin tops
Tin top AI is pretty wild at both Bathursts IMHO. I've tried a number of classes but the GTEs are the most egregious. At anything above 90% skill the AI are faster than the player on mountain straight, up the mountain, over the mountain, down the mountain, and 20+kmh faster down Conrod. The only places they can be caught are Griffins and the Cutting. They tailgate and attempt insane passes down the mountain just for good measure. This is a track I know like the back of my hand from years of sim racing and yet I end up drenched in sweat (for the wrong reasons) battling to survive against opponents in classes and skill levels that I can just cruise to a podium with on most other tracks.

Road America/F Inter
Even at low skill (<95%) the AI magically gain 1sec on the player up the main straight, and kettle straight. If you are behind them at this point, they pull away like drafting doesn't exist

F Junior
And while on the subject of discrepancies, I still think that F Jr AI are overpowered relative to the player. The have better brakes, and more grip and acceleration out of corners, and higher top speed on many tracks. I usually need to turn AI skill down by 5-10 to compete.

I know this is not the place to ask it, but too bad: will we ever see the return of the per class and per track difficulty slider?

 

Yeah this is not a tin top, but an universal problem. I reported some of this in the other thread, too, but so far Bathurst hasn't made it on the "to be redone" list.

For the formula classes the AI definitely has too good traction out of the Cutting, but are rather slow in the Esses.

 

Perhaps the Group C AI is not yet calibrated but I'll post my yesterday experience nonetheless.

Baseline Strength: 103/Medium Aggression
Car Used: Group C (driving the Porsche)
Track: Silverstone 1991
Session: Qualifying + Race
Conditions: cloudy

Copse: AI could be a bit faster here but basically nothing to worry about
Becketts and Chapel: exit Becketts the AI is a bit slow
Club: AI is very slow from mid to exit, they could be much faster
Priory: AI imo could brake harder/later for this corner
Brooklands & Luffield: also the AI could be faster throughout this tight section

 

Baseline Strength: 105/Max Aggression
Car Used: Sprint Race
Track: All
Session: All

I've been doing a custom Sprint Race championship and I've noticed the AI is 5-6 seconds slower than me in wet conditions. As the track begins to dry the difference will diminish until it's completely dry and they are at the same pace as me.

For now I'm not using wet weather, but I hope it gets fixed soon.

 

Copa Classic, Londrina long version, at 120% difficulty, in qualification and race i'm 2.5 seconds faster than the second. And i do suck, definetely not a pro, not even close to esports drivers.

 

GT4, AI 95%, any track
In full wet weather the AI brake too early and are very slow into corners, but then gain impossible grip and acceleration out of the corner (absolutely cannot be kept up with). They are 10+kms faster than the player on straights. The slow entry and fast exit/straights don’t cancel each other out, so the AI just end up several seconds faster than the player

 

Baseline Strength: 105/Max Aggression
Car Used: Sprint Race
Track: Londrina short
Session: All

It feels like they are way too slow on both chicanes.

Also my country flag is wrong, it should be Brazil.

 

McLaren MP4/6 F1 Classic Gen 4 , IA Starts Very Slow

 

Autoclub Road Course - the AI will NEVER attempt a passing move when approaching this corner no matter how much space is given, instead decide to go single file.

 

I did some wet AI testing in the F-Vintage G1&2 and the Retros G1&2. Tested in Kyalami HIstoric, 108%, medium aggression.

Vintages: These need a serious review, I was 10 seconds faster on the wet than on the dry in the G1 and 7 seconds on the G2. Tested in Silverstone to see if it was a track problem and nope, 8 seconds faster on the G1. Now the issue I think is not entirely the AI being slow, I felt I didn't lose nearly the amount of grip you should lose for the conditions, only lost 4.3 seconds from my dry time, the AI lost a more realistic 14 seconds. So I think the tires have too much grip for the player and hence the difference.

Retros: These were much more straightforward, the AI lost 2 seconds to me from dry to wet on both classes, the grip level in the G1 felt fine although I would expect a bigger delta than 9 seconds. In the G2 it was a little too easy to drive and the delta was too low to me as well, only losing5.5 seconds from the dry time.



Again 108% difficulty is my baseline as I have to push to the max to finish in the top 5 in the F-UltimateG2 at Barcelona, so it should be my proper difficulty for all classes per the AI discrepancy report thread.

 

So ummm maybe I am just that good?

Seriously though, I am on 1.5.3.5.2458 which I believe is the latest main release. Did a file integrity check and it didn't change the results, they gained 2 seconds but that's not enough.

Can anyone do the same test to see if the results are the same?
AI 108%
Medium Aggresion
Kyalami Historic
Rain with the wet track state at the beginning