Official General AI Comments & Discussion Topic

I think I read in the update notes they are still calibrating the tire wear and AI pace in the cars with the new physics, should be fixed by the official 1.5 release! Hopefully!

 

1.4.8 RC, i have the ai at 105 difficulty and medium aggression. With the f-usa 23 at spa 2022, the ai are a bit quick on the kemmel straight, but are really slow around the twisty middle of the track. They also are very slow at stavelot, they should take more outside kerb and carry speed but instead they park it on the apex.

 

Formula Junior
Any Track

I know they're in RC but the AI for this class are quite overpowered compared to the player, even with the skill down at 92-95. The Formula Inter at the same difficulty is easily kept up with.

 

Game version: V1.4.8.0.2299AVX (releasecandidate branch)
Car: F3
Track: Kyalami (modern)
AI settings: AI skill 105, aggression max
Report: Got brake-checked by an AI car on the Kyalami grid straight.

 

not sure but I think it could be unforced mistake by AI.
So nothing wrong with that I suppose?

 

It's a straight. The AI is not changing its line. It should be holding down the gas pedal for another 400 meters or so, and at this point on the straight it should have no problems with grip.

If this was 200 meters back at the exit of the last corner when they're just going on the throttle, I might agree. It's not though. If it's an unforced mistake, then there's something that needs fixing in the logic for unforced mistakes, unless it's also simulating things like a player getting attacked by a cat in their rig mid-race.

The AI also did not run out of fuel, and there was no evidence of a mechanical failure that I noticed. I could theoretically believe this to be a move to let a faster car pass (which I wouldn't mind as a feature, especially when it comes to quali behaviour on the AI's outlap or blue-flagged AI cars in race sessions, which this car wasn't), but if that is the case then the logic for facilitating that pass needs to be made safer.

This just really did not feel right to me.

 

Seems to me it slowed for the cars exiting the pits.

 

That crossed my mind, yeah. Still doesn't make sense for the AI to do though, there's no reason for it at all.

 

Is there a limit from 19 AI drivers in Championchip Mode although more liveries are available?

 

The maximum opponents are set by the tracks, so if you have a 5 race championship and 1 of the tracks has a maximum of 19 slots, that will be the maximum for all the other races.

It would be cool if that could be changed, so we could better recreate vintage seasons where the amount of cars would change a lot, race by race.

 

Thank you. @Mazdaspeed
How can I see the max. slots of the tracks?

 

You have to check it track by track in single race mode, dunno if there is a faster way in game.

 

Just ran a timed 1 hour Endurance race with the P1-P4 class, thunderstorm to dry weather, on 1.4.8.1. AI was on 113. Half the P1 class (and several others, I suspect but can't know for sure) refused to pit for dry tires, losing 6-7 seconds a lap, allowing me (who had crashed in the rain) to catch up almost a minute and a half against AI that were in a 1 second window compared to me in terms of laptime.

Not sure if this is supposed to simulate strategical errors?

 

The AI seem to have issues dealing with wet to dry races, I had a race in the RetroG2 where they kept pitting each lap on a drying track but never changed to slicks, was at Spielberg historic.

 

Today I have testet every single track of my custom championship in single Race Mode. Here I can select more than 20 opponents on every track.
Than I createt a new championship with just one track (Silverstone). And also there the game again allows only 20 cars.

 

Couldn't find an specific thread to ask this, but does the AI loose performance due car damage?

8:47

The Porsche loose it's rear wing, but keep as fast as if not. Before it, two cars had their front bodies badly damage and still keep a strong pace.

 

Car / Track: Formula Junior / Tarumã
AI Strength: 107
AI Aggression: Medium
Damage: High
Session: Weekend (P: 30min / Q: 10min / R: 20min)
Other settings: Dry. Tyre wear and fuel use (Standard)
Report: Abnormal difference in lap times between P/Q and R sessions. As an example average time of 5/6 leaders:
Practice and Qualy = 1:16.0 ~1:16.5
Race = 1:14.0 ~1:14.5
They are on average 2s faster in the race. I understand that for a category like this with few adjustments and fuel consumption, the times between sessions could even be close, but I find a discrepancy to be 2s faster in race.

To reproduce, follow the settings mentioned.

 

Car : Formula Vintage Gen 1
Track: Brandhatch
AI Strength: 92
AI Aggression: high

Ai speed is very slow on the start/finish straight. I am on race pace with them in the corners but at the straight they are way to slow. Any higher difficulty and I cant keep up with them anymore at the rest of the track so its a bit weird to race.

 

Anyway in race AI should be at least a little slower than qualy. Even without taking into account the car's setup, in qualify driver risks more and pushes more aggressively than in the race.

 

Adding:
New test with same specs, except for ai ability now at 100% - The difference is even greater, on average 3.5s to 4s.
P/Q = 1:18.5 ~ 1:18.9
Race = 1:15.2 ~1:15.6