Official General AI Comments & Discussion Topic

Velopark, stock cruze 2023, AI High aggro 115 diff.

AI seems to be quite slow in sector 2. The way they come out of the back straight makes them loose easily half a second. Everywhere else they're fine.

 

Track: Road America
Layout: Standard
Corner: Carousel; Specifically mid-to-late
Car: DPI Cadillac
Weather: Dry and sunny
Difficulty & Aggression: 113; High

Description: During a 15 min race, I noticed on numerous occasions that AI would drive straight into the gravel on the left side of the carousel. I was able to get this event to trigger by having an AI within ~1 second behind me as I went through the corner. It seems that they try to avoid driving directly behind me and inadvertently drive off the track in the process. This meant that I couldn’t be overtaken the entire race because any AI that got close would end up driving off track and crashing in the carousel.

 

A few posts back I mentioned an issue with a lot of F-Classic Gen 1 cars running out of fuel at Monza 1991.

Since then I've started a championship with these cars and found the same issue at Imola 1988 and Spa 1993. At Silverstone 1975 it was even worse - all the AI cars failed to finish!

So it seems something is off with the fuel loads and fuel calculations for these cars. I don't know if it is possible, but I think it would be good if the user could have the ability to set the fuel loads and strategies for the AI cars.

 

Car: Formula V10 Gen2
Track: Silverstone 2001
Settings: 105% Strength
Report: AI is extremely slow in the last two corners on the rain.

 

Car GT3's, all of em, and I don't think its just related to these cars, think its all A.I. cars
Track Laguna Seca, but again, think its any track
Settings: Happens with any skill level or track state
Report: Basically, A. I. (dry tires at least) are wearing immensely faster than players!

This is taken from a discussion about it in another thread, explains in a bit more detail:

"Im very glad you brought this whole thing up actually @Alexandre Sene, as it has reminded me to try something out which seems like a big problem, and after doing a little test, it indeed is a big problem!

See the driver swaps (I know this from recent streams of long races with A.I. driver swaps) are bugged, because when the A.I. takes over your car, the tires start to waer at an incredibly fast rate, and they keep pitting for tires! I wondered if this was related to just hwen the A.I. take over your car, or if there is something wrong with A.I. tire wear in general, and it turns out its in general!

So I tried running a a long race with the A.I. in the GT3 cars at Laguna Seca in hot conditions but with tire wear turned up to 5X (something I would rarely ever do apart from the purpose of a test), and sure enough, just like when they take over your car, they're tires seem to be waering at a much accelerated rate than yours, as by the time my tires were only about 1/8 worn (so barely worn), they were already pitting for new tires! Infact by the time my tires were just over halfway worn, they'd pitted 3 times!

I've just gone back and ran a shorted race to check this and again, they pitted when my tires were about 1/8th worn."

 

Car: F-Classic Gen1 Model 2, McLaren MP4/5B, McLaren MP4/6, McLaren MP4/7A.
Track: Any.
Settings: Any.
Report: AI is extremely slow on standing starts, always losing multiple positions before even getting to turn 1.

 

Car: any. Specifically GT1
Track: Suzuka Kansai GP and classic
Settings: 100% , 60 aggression
Report: AI is extremely slow on 130R Conner and very early braking before Hairpin corner

Car: champcar indy
Track: Daytona oval
Settings: 100% , mid aggression
Report: with some car settings, it's easy to get pile position by 0.3

Car: aus V8, F1 76
Track: brands hatch Indy
Settings: 100% , mid aggression, wet
Report: on wet brands hatch ai are slow everywhere

 

Car: Group C, GT1 (Potentially others as well)
Track: Any(?)
Report: Low downforce variant AI for both group C and GT1 does substantially faster and differently driven laps at a given AI rating than high downforce, resulting in poor raceability

i.e. for Group C, ~90-95% AI in low downforce config makes similar laptimes to ~105-110% AI in high downforce, with there being a slightly narrower gap for GT1 as far as I can tell. This results in the low downforce AI cars having extremely early braking points compared to the high downforce cars at a similar relative pace, but also having inhumanly good corner exits, which make them borderline impossible to properly race against

This difference existed on the Daytona Sports Car course, Monza 1991 version, and Hockenheim 1988 version, and probably other tracks as well. I used Spa, Buenos Aires No.15, and Road America as high downforce tracks to compare to and it seemed pretty consistent across all of them.

 

Car: GT4
Track: Spa 2022
Settings: 100% skill, high aggression
Report: AI is very slow off of Stavelot but disproportionately fast by the time they get to Blanchimont which doesn't make sense.

 

• Car / Track used: Formula V10 Gen 2 / Montreal GP
• AI Strength setting: 110
• Corner (if issue is corner specific): T8 and T13 (see map below)
• AI Aggression setting: High
• Session Distance: 20 Lap Race
• Weather when the issue was observed: Dry, Light Cloud, Default Date
• Player Tyre: Soft Tyre
• AI Tyre (if known to be different from player): Suppose it's Soft too
• Type of session when the issue was observed (practice / quali / race): Race
  
  
  
  When fighting other AI cars (this was a race where I did not take part in purpose to evaluate AI performance), when not being close enough to overtake but trying to upset rival car, AI will bite the yellow kerb too much. When that happens, it will put two wheels up in the air, and sometimes end up turning itself over in a very awkward way. Clips below illustrating every instance I found on the subject in one single race:
  
  
  
  
  
  
  
  
  
  
  
  
  
  

 

Car: Formula Vintage Gen 1 and 2 (worse in Gen 1)
Track: Spa 1970 1000km
Settings: 111% Strength
Aggression: Medium
Report: The AI still wants to go through the Masta Kink side by side sometimes which leads to disaster. Especially in the first lap it's a gamble. It's better than it was a couple of month back but still happens too often. Rest of the lap is fine. They could be faster at Malmedy

 

Hi Reiza Guys

We need to talk abou Max speed, It’s pretty obvius that we have this problem with maximum speed. Even with different race skills, the cars reach the same final speed on straights (it doesn't matter if the race_skill is 0.1 or 1.0), it's actually quite annoying to see that Sena takes several laps to get past Alex Cafffi (formula classic in Jacarepagua for example, but it is clear that the game already has this slider, as we have different final speeds between classes, or even better, between models of the same class! We just need Reiza to make them available to edit. Race_skills would remain the same (as explained on page 1), and we would have a new slider (which already exists internally) max_speed.

If Reiza lets me dream this far, another slider that would be a game changer would be acceleration, it would be directly linked to max_speed, and it is also an internal slider (But not editable, as explained above).

My last dream slider would be brakes, I believe it would be a little more difficult as we would have different lines during the race, with cars braking later or earlier depending on the situation, but you already do that too, don't you?

We would certainly need to make some changes to the race_skill, if these sliders are implemented to prevent the lap times between the best and the poorest driver on the grid from being very discrepant (perhaps in an opponent's skill of 90% it would have to be adjusted to 92.5% and 87.5% instead of 95% and 85%, but it will be worth it)

I've seen other posts on this forum talking about this issue of final speed, in other words, it's not just my demand, it's the entire community's.

I believe it would be the biggest update since you let us update the pilot sliders, but now that we no longer have the season pass it would be a good recipe for you too (you don't even need to do much, just let us edit).

I'm a fan of this company, I bought 3 copies of the game, the way you take care of updates and listen to the community is very special, congratulations!

read with affection, and please help with my dreams, you are the best!

 

• Car / Track used: Ultimate Gen2/Montral GP
• AI Strength setting: 107
• Corner (if issue is corner specific): 3/4; 8/9; 13/14; 10
• AI Aggression setting: Low
• Session Distance: Single Race (35 laps) and Championship Mode 50% race distance (35 laps)
• Weather when the issue was observed: Light cloud
• Player Tyre: Soft
• AI Tyre (if known to be different from player): Soft
• Type of session when the issue was observed (practice / quali / race): Race

AI cars are slow in the chicanes 3/4; 8/9 and 13/14. My lap times however were comparable, which means the AI is faster everywhere else. I found them particular fast coming out of turn 10.

On a side note, I did the race twice, single race (35 laps, 2x tire wear) and in championship mode (50% race distance, authentic tire wear), no mandatory pitstops. In the single race, AI pitted three times, always on soft. I pitted twice because I used a set of mediums. Tire wear seemed to match between AI and me. I found it difficult to overtake coming out of the hairpin with DRS. In championship mode the AI cars didn't pit at all and had very consistent lap times from the 2nd to the last lap of 1:15.8 to 1.16.2. They don't seem to suffer from any tire wear in that mode. I had comparable tire were to the single race. Oddly, I found it much easier to overtake coming out of turn 10 with DRS.

 

It really is very difficult for AI to overtake on straights, even long ones, as cars of the same class/model have the same engine performance.

In this topic: Automobilista 2 - Extended Car Info Spreadsheet we can verify that each model of each class has two crucial fields, Power and Torque, that is, the cars, within each model, are all the same, consequently, categories that have only a single model such as Formula Reiza, and Formula Ultimate Gen1 and Gen2, the cars are differentiated by the sliders for each driver, as there is no slider for power and torque ( and "race_skill" does not deal with straight line speed), we can say that all cars in the same class are the same, just with different drivers, however this is not the conception of racing cars (at least in formula racing the cars are very many different).

I've seen someone ask here why pilots with the same sliders had different performances, being from the same class (but different models) Example below of the formula classic G1, taken from the spreadsheet of the topic mentioned in the previous paragraph:

Formula Classic G1M1 - Power 1190 Torque 780
Formula Classic G1M2 - Power 1300 Torque 820
Lotus 98T - Power 1280 Torque 795

With this information it is easy to understand why the G1M2 are faster on the straights, even if the drivers have a lower "race_skill".

I believe that Reiza's conception of the game perhaps considered that the drivers' sliders would be enough to differentiate the performance of the AI on the track, but this does not always happen (like the example cited of Senna's difficulty overtaking Alex Caffi, on a long straight). Works perfectly on lap times, as "race_skill" works mainly on curves

I think this concept of the same cars and different drivers would be great at the game's launch, but with each update that becomes more realistic, players become more demanding too, in such a way that I, as an offline gamer, also hope that "Power" and "Torque" may be implemented in the near future as a pilot slider.

I'm not a programmer, but it seems simple to me, put these sliders in the file for each class, ranging from 0.1 to 1.0 and the game assigns the power and torque of each vehicle when starting the race.

I don't like playing with controllers, much less multiplayer, I like customizing as many aspects of the race as possible, abandoning it and watching the AI x AI spectacle, I know I'm in the minority, but we exist, and adding the "power" sliders and "torque" in every AI driver, would be simply phenomenal.

Please Reiza Guys, tell me if you read this here

Thank you very much

(bad english, sorry - Google translate helping)

 

• Car / Track used: Copa Classic B / Cascavel
• AI Strength setting: 120

• AI Aggression setting: high

• Weather when the issue was observed: dry

The AI is too slow. The faster AI in 120% does 1m20s2, and the leaderbord record for the class is 1m19s6.

 

• Car / Track used: Opala 86 / Interlagos - Historic
• AI Strength setting: 120

• AI Aggression setting: high

• Weather when the issue was observed: dry

The AI is too slow. I am no alien and I am 2,5s faster than the faster AI in 120%.

 

• Car / Track used: P1 Gen2 / Brasília, Interlagos.
• AI Strength setting: 118

• AI Aggression setting: high

• Weather when the issue was observed: dry
• Other settings: tyre wear and fuel consumption = x2, Race length = 2h

Not sure what is going on with the sim. The AI was getting smarter regarding fuel and tyre use/pitstops. Now they are again doing one more pitstop than needed.

 

• Car / Track used: Stock Car 2023 / Goiania
• AI Strength setting: 118

• AI Aggression setting: high

• Weather when the issue was observed: dry
• Other settings: tyre wear and fuel consumption = realistic, Race length = 40min, Mandatory pitstop

The AI strategy makes no sense. They will all pit in different laps (great!) but will all apparently fill the whole tank (not great), even when it's not needed. So the AI that pits earlier will have massive advantage over the others because it will finish refueling quicker. And if the player does the math, it will have massive advantage over all the AIs...

 

• Car / Track used: Montreal mordem / F-10 Gen 2
• AI Strength setting: 92
• AI Aggression setting: high
• Weather when the issue was observed: dry
• Other settings: tyre wear= x3 fuel consumption = realistic, Race length = 15min, NO mandatory pitstop.
  
  Issue: All AI drivers stopped on lap 9 for tires (at least thats what i got from the replays). This was clearly not a strategy choice, not only because there was very little wear for the the short distance, but also because all of the drivers stopped on the same lap. Something is going on here.

 

Where you using medium or hard tires? The AI only use softs, so the player always has an advantage when the mediums or hards are a better choice. 15m at 3x wear in Montreal is 35-39 laps, so the softs would get some considerable wear in them at 9 minutes.