AC GT3 @ Red Bull Ring - Sunday 21st February 2021

[demetri]

I'd like to share some interesting information about setups.

I tried to test what difference a good setup makes using AI driver (for the sake of consistency) in the Audi at RBR. This is a track where AI has a decent fast line and doesn't do lots of stupid things in GT3 cars like it does at Spa for example.

First, I made AI run with 100% stock setup (except for fuel set to 20L). AI did 1:31.752 lap. Slow, but expected since the tires were stone cold.
Then I changed the tires from default mediums to softs and AI did 1:31.016. Three quarters of a second faster already.
Then I also changed the final gear ratio to the lowest available and AI posted 1:30.876 lap. Another tenth and a half.
Finally, I loaded my setup with the same 20L and AI did 1:30.464.

So, the difference between bone stock and custom is ~1.3 seconds, but more than half of that is simply selecting correct tires. And if we compare the minimally tweaked (just softer tires and shorter gears, which is an absolute no-brainer for this track) default setup with 100% custom setup (changed suspension, dampers, aero, alignment, gears, diff, tire pressures - everything) the difference is just 4 tenths.

I also ran that tweaked default setup myself and did time which was whole one second slower than my best time with the custom setup. I guess AI has better setup adaptability than me . That setup felt way too understeery and stable (probably due to rear wing at 6 degrees), but was very easy to drive.

I have to say that these AI times are somewhat slow even with the custom setup, I think the most of this slowness comes from AI hesitating to use curbs and upshifting too early in this car. Still, I think it is a useful illustration of the setup importance (not as big as most people think, the biggest improvement comes from running better lines and choosing correct tires).

 

[Fiberoptix]

Hey Chris, can I change to the 911 please. Thank you

 

[JamKart22]

I'd like to share some interesting information about setups.

I tried to test what difference a good setup makes using AI driver (for the sake of consistency) in the Audi at RBR. This is a track where AI has a decent fast line and doesn't do lots of stupid things in GT3 cars like it does at Spa for example.

First, I made AI run with 100% stock setup (except for fuel set to 20L). AI did 1:31.752 lap. Slow, but expected since the tires were stone cold.
Then I changed the tires from default mediums to softs and AI did 1:31.016. Three quarters of a second faster already.
Then I also changed the final gear ratio to the lowest available and AI posted 1:30.876 lap. Another tenth and a half.
Finally, I loaded my setup with the same 20L and AI did 1:30.464.

So, the difference between bone stock and custom is ~1.3 seconds, but more than half of that is simply selecting correct tires. And if we compare the minimally tweaked (just softer tires and shorter gears, which is an absolute no-brainer for this track) default setup with 100% custom setup (changed suspension, dampers, aero, alignment, gears, diff, tire pressures - everything) the difference is just 4 tenths.

I also ran that tweaked default setup myself and did time which was whole one second slower than my best time with the custom setup. I guess AI has better setup adaptability than me . That setup felt way too understeery and stable (probably due to rear wing at 6 degrees), but was very easy to drive.

I have to say that these AI times are somewhat slow even with the custom setup, I think the most of this slowness comes from AI hesitating to use curbs and upshifting too early in this car. Still, I think it is a useful illustration of the setup importance (not as big as most people think, the biggest improvement comes from running better lines and choosing correct tires).

Interesting finding. I'm surprised a good setup makes so little difference to AI.

If you have a race against AI then do they all use your own setup and fuel load? If you need a pitstop then would they always change tyres?

It would be interesting to know what AI car setup you're racing against as I use these races as a bit of training for the real event.

 

[demetri]

Interesting finding. I'm surprised a good setup makes so little difference to AI.

If you have a race against AI then do they all use your own setup and fuel load? If you need a pitstop then would they always change tyres?

It would be interesting to know what AI car setup you're racing against as I use these races as a bit of training for the real event.

It does a relatively small difference (just a second) for me as well. Maybe default settings for the Audi are decent enough, it could be different for other cars.

You can name your setup "ai_default" and they will sort of take it (at least the log file shows something about AI loading the setup), but they will still adjust tire compound to their liking (always mediums for GT3 cars, they seem to use softs in quali only), final gear ratio and fuel load, making it almost pointless if you want to race against them. I could not force AIs to use softs tires in a race. Also, in a long race, AIs never change the tires or fix the damage, they just refuel on pitstops. I remember running a long race against AI a few years ago where they could barely stay on the track because their tires were absolutely shot by the end of the race

P.S. I did these AI tests in hotlap mode by pressing Ctrl+C which makes AI to take control of your car and run laps with whatever setup you got loaded.

 

[VictorIS]

in pleas Z4
Victor Isikov or VictorIS not cyber0punk I jast dont understand how change that.

 

[jackdyson31]

I'd like to share some interesting information about setups.

I tried to test what difference a good setup makes using AI driver (for the sake of consistency) in the Audi at RBR. This is a track where AI has a decent fast line and doesn't do lots of stupid things in GT3 cars like it does at Spa for example.

First, I made AI run with 100% stock setup (except for fuel set to 20L). AI did 1:31.752 lap. Slow, but expected since the tires were stone cold.
Then I changed the tires from default mediums to softs and AI did 1:31.016. Three quarters of a second faster already.
Then I also changed the final gear ratio to the lowest available and AI posted 1:30.876 lap. Another tenth and a half.
Finally, I loaded my setup with the same 20L and AI did 1:30.464.

So, the difference between bone stock and custom is ~1.3 seconds, but more than half of that is simply selecting correct tires. And if we compare the minimally tweaked (just softer tires and shorter gears, which is an absolute no-brainer for this track) default setup with 100% custom setup (changed suspension, dampers, aero, alignment, gears, diff, tire pressures - everything) the difference is just 4 tenths.

I also ran that tweaked default setup myself and did time which was whole one second slower than my best time with the custom setup. I guess AI has better setup adaptability than me . That setup felt way too understeery and stable (probably due to rear wing at 6 degrees), but was very easy to drive.

I have to say that these AI times are somewhat slow even with the custom setup, I think the most of this slowness comes from AI hesitating to use curbs and upshifting too early in this car. Still, I think it is a useful illustration of the setup importance (not as big as most people think, the biggest improvement comes from running better lines and choosing correct tires).

Hi Dmitry,
I would agree with that from what I have found driving these (excellent) cars. The quick laptime is very sensitive to the racing line and percentage throttle (this is hurting a lot right now). You can't afford any sliding or you lose a lot of time and if you don't use the curbs you lose a second a lap there as well. To use the curbs all the time by habit requires a special setup, a large pair of nuts and a compliant setup.

Thanks for posting the results of such an interesting experiment.

I wanted to ask if using the experimental physics option on the ACM was ok for the server tonight ?

PS pressing ctrl-C might not be what the AI themselves use however because I have read somewhere that they don't use realistic physics apparently.

 

[thom462]

Always too slow here but for the fun please sign me in with McLaren. Thanks

 

[uvraver]

I'd like to share some interesting information about setups.

I tried to test what difference a good setup makes using AI driver (for the sake of consistency) in the Audi at RBR. This is a track where AI has a decent fast line and doesn't do lots of stupid things in GT3 cars like it does at Spa for example.

First, I made AI run with 100% stock setup (except for fuel set to 20L). AI did 1:31.752 lap. Slow, but expected since the tires were stone cold.
Then I changed the tires from default mediums to softs and AI did 1:31.016. Three quarters of a second faster already.
Then I also changed the final gear ratio to the lowest available and AI posted 1:30.876 lap. Another tenth and a half.
Finally, I loaded my setup with the same 20L and AI did 1:30.464.

So, the difference between bone stock and custom is ~1.3 seconds, but more than half of that is simply selecting correct tires. And if we compare the minimally tweaked (just softer tires and shorter gears, which is an absolute no-brainer for this track) default setup with 100% custom setup (changed suspension, dampers, aero, alignment, gears, diff, tire pressures - everything) the difference is just 4 tenths.

I also ran that tweaked default setup myself and did time which was whole one second slower than my best time with the custom setup. I guess AI has better setup adaptability than me . That setup felt way too understeery and stable (probably due to rear wing at 6 degrees), but was very easy to drive.

I have to say that these AI times are somewhat slow even with the custom setup, I think the most of this slowness comes from AI hesitating to use curbs and upshifting too early in this car. Still, I think it is a useful illustration of the setup importance (not as big as most people think, the biggest improvement comes from running better lines and choosing correct tires).

How did you do this? Make the AI run your own setup.
It's an interesting idea.

 

[Goblin616]

@Chris Down Please move me to the reserve list. I have a migraine and don't know if I'll be able to make it today.

 

[demetri]

How did you do this? Make the AI run your own setup.
It's an interesting idea.

Press Ctrl-C when you're in your car and ready to drive. Only works offline of course

Hi Dmitry,
I would agree with that from what I have found driving these (excellent) cars. The quick laptime is very sensitive to the racing line and percentage throttle (this is hurting a lot right now). You can't afford any sliding or you lose a lot of time and if you don't use the curbs you lose a second a lap there as well. To use the curbs all the time by habit requires a special setup, a large pair of nuts and a compliant setup.

Thanks for posting the results of such an interesting experiment.

I wanted to ask if using the experimental physics option on the ACM was ok for the server tonight ?

PS pressing ctrl-C might not be what the AI themselves use however because I have read somewhere that they don't use realistic physics apparently.

They post times very similar to the default setup in my experiment when you race against them and I believe that AC is one of the few sims where AI cars use the same physics. They don't use ABS or TC though

 

[thom462]

I have conflicting dates maybe I will make it to join race

 

[demetri]

in pleas Z4
Victor Isikov or VictorIS not cyber0punk I jast dont understand how change that.

If you use Content Manager it's here:

And I'm sorry for punting you yesterday, but you left your car blocking the racing line and wouldn't respond to chat, so I had no way to make you move rather than smack your car really hard

 

[Goblin616]

Press Ctrl-C when you're in your car and ready to drive. Only works offline of course


They post times very similar to the default setup in my experiment when you race against them and I believe that AC is one of the few sims where AI cars use the same physics. They don't use ABS or TC though

For me it only works on Hotlap so far

 

[jackdyson31]

Press Ctrl-C when you're in your car and ready to drive. Only works offline of course


They post times very similar to the default setup in my experiment when you race against them and I believe that AC is one of the few sims where AI cars use the same physics. They don't use ABS or TC though

Thanks mate - doesn't surprise me about AC given its quality, I had tried this same thing on rF2 and the AI gives you a time much slower than what the other AI's drive. That's why I asked as I didn't know about AC.

 

[demetri]

For me it only works on Hotlap so far

Works in quick race too

 

[Stefano La Commare]

Sorry @Chris Down, I won't be back home in time for this evening's race so I'll have to skip this one.

 

[jackdyson31]

Found it

 

[Chris Down]

Race server #13 is being configured, no more changes will be permitted.

(I'm not feeling too good, so won't be racing tonight. I'll set the server so that everything will tick over naturally.)

 

[Goblin616]

I'm feeling well enough to drive Chris. Can I still join? If not that's ok. I'll just do my thing.

 

[hape]

hi
sorry guys i have to sign out. I have pain in my left arm and can't drive the race.
hape