Given the release of the Ferrari GT3 I figured I would see if Studio 397 has addressed the issues related with AI BOP across vehicles. They have not. The 2020 Bentley is still slow to the point of basically being broken, and the Radical GT3 is still ridiculously fast to the point of basically being broken. The new Ferrari is fast but not ridiculously so which makes it a bit easier to run in an AI race. In the table below the Gap indicates the gap to the Aston Martin. Gap to fastest is the gap to the Radical. AI Adjustment indicates what the AI would need to be set on in the talent files to put all the cars on a similar pace. There are a number of cars running at a similar pace: 2017 Bentley McLaren 720s Porsche 911 GT3 R Aston Martin McLaren 650s 2018 Audi R8 Mercedes AMG The 2018 Audi R8 isn't too much different. However, it is surprising that it was slower than the 2019 Audi R8 in this analysis as it has been faster in past analyses. To estimate the difference in performance per lap between cars I ran 26 AI with randomized skill for a large number of laps (typically between 30 and 300) on the following tracks: Silverstone Spa Zandvoort Paul Ricard Nurburgring Hockenheim Monza Misano Sebring After each run I recorded the fastest lap. With that data I ran a linear regression in the form of: [gap to leader] ~ [track] + [car] + [AI skill] The model explains ~80% of the variance in the gap to the leader. The coefficient for AI skill is -0.030 which would mean that a 1 point increase in AI skill equates to a -0.030 second decrease in the gap to leader. This is useful for determining how to setup AI skill files in rFactor 2. This coefficient is a bit higher than the last time I ran the analysis (-0.026), but more inline with my historical analysis (-0.032). Given the changes in cars and tracks between analyses and Studio 397 changes in default setups it makes sense that this number continues to evolve. Generally these results are in line with my expectations given past analyses (broken Radical & 2020 Bentley).
Did you run a Static Track Surface or was it Dynamic? Any Weather involved? (Guess not but just checking!)
No dynamic weather and the surface wasn’t static (good idea), but even if you ran it with those parameters you would see a difference between cars. Get a grid of 10 radicals and 10 bentleys with the same AI skill and you’ll have 10 radicals at the front and 10 Bentley’s at the back.
I can confirm this...Radicals too fast...Bentleys too slow....every time (100/50 AI Strength/Agression)
Seems like a solid test, quality input from the community to improve the product. Thanks for taking the effort. Has this issue been acknowledged by S397? Any fixes?
No to both questions as far as I know. I migrated to AMS2 shortly after this test which is imperfect for sure but they seem more responsive to community feedback. I am still waiting on an AMS2 championship mode. On the flip side I had to write my own championship mode for rf2. So I dunno. The sim world is just bad at single player.
Actually the AI BOP was released a while ago on 21st Oct see this Post for the Change Log: https://forum.studio-397.com/index....pdates-includes-bop.67278/page-7#post-1043595
