For a while now I’ve been trying to work out a way to set AI speeds for offline racing, and I see some discussions about global hype scaling elsewhere on the forum. So I have an idea that works for me, that others may wish to try.
I don’t like to use the basic GPL feature that adjusts the AI to how fast I’m going , instead I prefer to set a target for each race where I know that if I’m at my very best I just have a chance of pole position. I started out a long time ago by setting the AI so that pole position would be the Rank target. But the problem here is that for some circuits the Rank is extremely difficult, and at others it is much easier. There are also too many car and track combinations that don’t have Ranks, so that was limiting.
Then I tried setting the AI lap times as a percentage offset to the world record for that Mod, but again the difficulty varied a lot from track to track. Even in percentage terms sometimes my best ever lap is less than 3% slower than the WR but sometimes it is over 6%. So again it was difficult to set a consistent target for the AI.
After doing some thinking I realised the problem is that the amount of influence the driver has on the lap time varies from track to track. At Monza there are not many corners, and so it is easier to be close to the WR. At the Nordschleife there are very many corners and so the WR is difficult to get close to, because the driver who holds the WR is gaining time against me at every corner – and they all add up!
So recently I developed a new method. This compares not just the WR lap time, but the WR lap adjusted by the number of corners per minute of lap time. The attached spreadsheet uses this measure as a modifier to the WR lap time to create an adjusted value I’ve called ‘Delta’ (couldn’t think of anything else to call it.) If Delta is 100% then that is the WR time.
What I find is that the Delta values are much less variable between tracks than just the WR %. My best laps usually have a Delta of 99.3% to 99.6%. I can enter a target Delta into the spreadsheet and it calculates the target lap time. If I set the target Delta to 99.7%, I know I really have to drive well to get pole.
I change dlong_speed_adj_coeff in the track.ini to get the fastest AI drivers qualifying at the Delta-target lap times, and I’m ready to go. I now find I have much better racing against the AI, as their speed is consistent across all different types of track and I can keep moving the goalposts so I keep improving.
Now, I realise not everyone will agree with the number of corners I have put down for each track. I have based this on the number of individual times the driver’s skill can affect the lap time. This isn’t perfect, so change them if you want, but these numbers work for me. Some WRs may also be out of date since I did this.
I have left the 1965 Mod section populated as an example.
Maybe some of you will find this tool useful. If anyone sees any problems or improvements, please let me know!
(Sorry for this very long post.)