General AI Performance Discrepancy Reports (No Discussion)

AI 113 diff at Montreal modern layout. Dry conditions, tested it with the P cup 4.0, I assume it's the same with other cars.
Turns 8 and 9, the chicane just before the hairpin, the AI seems to run too close to the wall on exit, by clipping the banana curb you can gain easily 3 or 4 tenths on them every lap.
They could also use the first banana kerb on the last chicane to gain some speed. AI is easily half a second slower per lap than what they should be if those lines where corrected.

 

As I reported elsewhere Group C AI is significantly faster in the wet. The opposite happens with GT1 cars, though not that dramatic
Settings: Laguna Seca, 120 AI, Sunny, Heavy Rubber preset -- Rain, Wet preset / Storm, Wet to soaked

Differences in Player vs AI:
Group C Dry: +0,337
Group C Wet: +2.977

GT1 Dry: +1.163
GT1 Wet: -0.600

 

I've found a similar issue with the Formula Retro Gen 1 - but similarly to the Group Cs, maybe they are a bit faster, but the biggest culprit is the tire wear is really intense for the player, and it doesn't affect the AI as much. Maybe I just need to get good, and I know wets are supposed to be softer than slicks, but I don't think I'm sliding or locking up all that much, but the tire wear is just so intense.

For Formula Retro, It was Monza and Österreichring on a soaked track, and for Group C its Daytona. The first lap I'm able to keep up alongside those I qualified with, but like a current day Haas, I drop like a stone to the back of the field by race's end. It could also be my lack of speed but I've noticed I would get slower and slower each lap, but the AI do not. And really worn wet tire is really scary to drive on, too.

I've noticed they can break much later, and get on the power faster when they have the wet tires on. But it might be just the tire wear for the AI or the player needs to be rebalanced.

 

GT3 gen2 at Daytona.

1. The AI seems to be way off top speed wise compared to the player, easily between 8 to 10 kph lower at the end of the straight, almost makes me think the AI is running the normal configuration and not the Low downforce variant of the GT3s.

2. AI constantly makes mistakes at the bus stop, almost every lap they seem to crash each other off making racing them imposible.

Tested this with 100L fuel, current date January 2024. Track default progressing from a green practise session.

 

LMDH behaviour on the pits.

Been doing a couple of 50 min races with the LMDH, tried Le Mans aswell as Spa and in both cases I've seen the same thing, the LMDH cars take way to long in the pits, 50 min is just the sweet spot to splash and dash at the end, 110L fuel tank + a quick pitsop to add maybe 10L more to finish the race, my pitstop with the porsche lasted arround 4 seconds + entry and exit to the pits, but for whatever reason the AI seem to lose an added 5 to 10 second extra on the pits. Don't really know if they're putting more fuel than necesary or what but I've seen this only in this car. Done multiple F1 races with multiple F classics and FV10's and I haven't seen this at all.

 

Multiclass at Lemans 24. 112% mid aggression. Field: gt3 gen 2 , lmdh, gte and p1 gen 2. Dry race, heavy rubber.
I was in the Sigma p1 g5. My fastest lap in this 10 lap race was 8 secs faster than the fastest ai in the same class.

 

AI. Daytona. The ai is broken at bus stop. They will slide and crash into each other. Seen with gt3s and lmdh.

AI slipstream vs player. The ai will never take the player slipstream if they are 0.1 from the player’s car. 0.2 and above is when they will tuck behind you. The problem is that they lose momentum and will miss an overtake opportunity. Ithink this is what causes the ai not toovertake including passing other ai cars.

 

AI Discrepancy Report: F-Ultimate Gen2 @ Imola (modern), 107% strength, 25% Aggression, Medium Rubber; Single Race, 50% Distance (32 laps). 2x Tyre Wear

The AI qualified 1.6s ahead of me. I tried this several times and the outcome was the same. I also didn't get any better .





In the race the AI was also much faster than me but pitted three times, around lap 8, 16, 24. I pitted once. The lap times of the top AI driver started with 1:18.8 and then got slower to around 1:19.6.

 

Version: 1.5.5.1.
Car: BMW M4 GT3 G2 (same GT3 G2 class race).
Track: Kansai GP (default progressing rubber, 15min P, 15min Q, end of qualy)
AI: 115 strength, medium aggression.
Setup discrepancie: over baseline setup (front/rear):

• Brake bias: 59/41%
• Brake ducts: 50/30%
• Rear wing: 8
• Tire Pressures: 1.60/1.55 bar
• Springs: 8/7
• Slow bump: 11/10
• Slow rebound: 12/11
• ARB: 13/13
• Electronics: ABS 4, TC 4

Report: AI seem to be way faster in sector 1, 1.1 s, where it's only 0.1 s faster in sectors 2 and 3 (with respect to P1 where I stayed P19, second screen capture). They are still waaay faster exiting turns, so the player hardly can reach them. Race line seem to be mostly fine, but they're still eagger to jump over curbs and grass to overtake.




Sorry for the setup display, I have several files in the folder, don't know which one was.

Edit: obviously there was a lot of margin to improve (I'm not fast enough in sector 1), but still, a lot of discrepancie between sectors with the AI .

 

Barcelona GP 1991:

AI is extremely slow in the middle sector between T9 (Campsa) and T11 (La Caixa). Especially in the fast Nissan chicane, they crawl through the curve.

 

Hockenheim 1988:
AI is way too aggressive in corners, especially in "Ostkurve" (second chicane), they would never let a faster car pass and rather provoke carnage and crash into other car or tyre wall.

 

Interlagos modern
Accelerated tire wear + Fuel
Formula Ultimate g2
30 laps
The AI's pace was very, very good, actually. Very good battles the entire race. However they had to pit 3 times during the race, where I could easily do 2 stops only - even by flatspotting my tires in the 1st stint. By the AI pitting a 3rd time, it gifted me the win.

I made sure to match the AI and only use softs.

 

I'll add to this. I did a 30-minute Monza + F-Ultimate Gen 2 race on RCO yesterday.
Time progression X2
Fuel usage X2
Tyre wear X2
Mandatory pit stop but no minimum tyre requirement.

I probably could've done the whole race on a single set of mediums, but I think all the AI pitted at least twice. (Sorry, didn't realize to take a screenshot of the results at the time to illustrate/prove.)

 

Baseline Strength: 108/Medium Aggression
Car Used: F-V10G2
Track: Montreal modern
Session: Qualifying

I went from 4 tenths down on the fastest time on the dry to 4 seconds faster on the wet, the AI lose a lot of time on the first chicane. Only took a screenshot of the wet session. This might be a track issue as the pace difference between dry and wet was spot on in Spielberg.

 

AI DISCREPANCY REPORT:
Ginetta G55 Supercup (same problem with GT4 class and probably many other car classes) @ Silverstone, 115% Strength, Medium Aggression, Medium Rubber.
AI becomes slower by 0.5-1 second over the duration of a stint, while the player gains around 1 second of pace due to lower fuel weight.

I missed the timing of the screenshot, so the sector times are missing, but the loss of time is balanced over the lap. Except that the AI is generally significantly slower in the last corner and the second. But this doesnt fall under your 1 second rule.

I logged some laptimes:
Lap Player AI
3 traffic 2:09.3
6 traffic 2:09.6
9 2:09.0 2:09.8
12 2:08.9 2:09.7
15 2:08.7 2:09.9
18 2:08.9 2:10.1
21 2:09.6 2:10.2
24 2:08.4 2:10.0
27 2:09.2 2:10.2
30 2:08.5 2:10.1
33 2:08.1 2:10.2
36 2:08.0 2:10.2

 

Baseline Strength: 108/Medium Aggression
Car Used: F-VintageG1
Track: Monza 71 and Kyalami 71
Session: Qualifying and race

It seems the player doesn't suffer any fuel weight penalty in the F-VintageG1 (might be the same in G2, haven't tested it yet). I can basically do the same time with 170l of fuel that I do with 30l. The AI meanwhile lose 3+ seconds.

 

Baseline Strength: 110/Medium Aggression
Car Used: P1 Gen 2, GT3 Gen 2, GT4
Track: Velo Citta
Session: Race

Was trying to do a 3 hour race with P1 Gen2, GT3 Gen 2, and GT4 at Velo Citta.

Problem I am running into in running these longer races is the tires seem to not be updating when the AI takes control. The UI at least isn't updating and the AI are pitting much earlier than they need to and I think it is because the tires are basically worn down to nothing 2/3rds through their stint.

That brings up my second issue. The AI according to the UI at least are still burning through tires at a much faster rate than the player does. You can see where the tires are at when the AI takes over. That was after 96L of fuel and about a 68 min sting.

The problem is the AI tires were showing at the level I have attached not even 60L and only 43 mins of racing. The AI pitted just a few laps later when like I said they had about 37.5L of fuel left. So while I could easily double stint these tires, it seems still impossible for the AI to come close to matching the player when it comes to tire wear.

I had these same issues in a LMDh, GT3 Gen2 race I did at Spa. I think all you have to do to replicate it is have tires get swapped at a pit stop and let the AI take control. The AI's pace seemed to also be negatively impacted by this as well, although I don't have exact numbers or data to report on that.

 

Bathurst
Formula Classic (All Gens), maybe other cars, too.

AI has some kind of magic grip and traction in the curvy sector at the end of sector one and the esses down the hill.
They can rush through the corners just as if they had some kind of TC and stability control enabled.
My last try was at 110% strength with Classic Gen 3. On other tracks, I can keep up with the AI, but in Bathurst, they drive circles around me because of this grip advantage down the esses.

AI is faster than the guys in TT on that track.