It's been a bit of time since I last launched the game...and it seems the AI had a bit of rubber-banding in the race that I've just ran. We were @Bathurst (yep, it is THAT week-End ), running a 23-car GT3 grid. Did the Practice and Q. AI set to 110. Tyre wear and fuel consumption turned off. I was running first, my pace was around 2:07.500/2.08.00. The first 3 AI were keeping pace - the AI ran some sub 2.08 laps. I made a mistake at some point, and was running 2D. The AI that inherited the 1st place started running around 2:09.00 / 2:09.500 Far too much a difference for stuff like Draft to be a factor (IMHO) That's a behavior I have never noticed before - and of course it's only one race, so may be a fluke, and I'd need to do some testing, but still...did anyone else noticed it?
@Robert Holm and @Marko Hartikainen are not that indulgent with us. There's nothing like that afaIk, if you screw up it's your own fault and you have to live with the consequences. How was the race configured? Do you have tyre wear enabled? Might be caused by this, or damage or smth else, but I've never seen rubber-banding in R3E.
That's is not what I'm saying at all! Quit the contrary in fact, since I shouldn't have been able to catch up the IA after my screw-up. Tire wear/fuel consumption and damages were off. Me nether - in fact, that's why I was that surprised
I was running at Bathurst the other day in the GT3's, and so many of the AI seemed to be going round at 20mph, it was like a slalom coarse, don't know if they were stuck with pit limiter on, don't know if there are similar problems on other tracks, also the AI seem very slow in free practice, nearly 5secs of my pace, really wish adaptive AI would work in free practice as well.
Yea I got that, was trying to sum up what I believe to be the coders guideline when it comes to AI behaviour. It can happen that damaged cars behave weirdly, but as damage was turned off...
This is a most likely due a well known bug or lack of refinement of both AI and damage modeling and Bathurst is as bad as it can get with this "feature". Well, Macao is probably another one, but I haven't driven there recently. Problem is that in Bathurst, AI cars are scraping and spinning on walls a lot and they can get their gear box damaged. If this happens, instead of limping to the garage nearby or retitiring immediately, AI decides that what the hell, I'm here for racing and thus decides to continue with a broken car going around 50km/h. That is most likely what you are seeing.
GTR2 had a rubber banding effect, I believe, so it wouldn't surprise me if Race07 and this also did. It wasn't huge, and (thankfully) you could eradicate it by tweaking the PLR settings file, but it was there.
I'm not sure that R3E has any rubber-banding at all - if we define rubber-banding as unrealistic lap times from AI to artificially close gaps. I have seen rubber banding in some recent F1 games where AI will be 10 seconds a lap faster after an accident so that a gap is quickly closed. I have seen nothing like that in R3E. What I have observed about R3E AI is this: (1) First point I should make is that I think the AI in R3E is among the best I've seen - I love it - its not perfect but its VERY good. (2) In selecting an AI level you are effectively setting the pace of the lead car. So, at AI=110 the AI car in first place will consistently run a certain lap time (with normal variation) throughout the race. (3) All other AI cars will be scaled according to the lead car's pace, except... (4) If an AI car is not running in first place and it finds itself with a clear track in front of it, it will tend to speed up a bit to try to close the gap. It won't speed up by much, but it's enough for it to start running fastest laps (or close to it, again with some variation). The consequence of this is that over dozens of races whose results I have kept in a spreadsheet, it is the midfielders that record fastest laps of the race. I have never seen a winning AI car get a fastest lap (I'm not saying it can't happen, I'm saying its rare). (5) A further consequence of all this is that if a human driver is leading, the AI will tend to go with the human player. As race leader the human is now the pace-setter - in first place the human player can effectively increase overall race pace. (I've not tested it but I assume if the human player is WAY better than the AI it will be possible just to disappear down the road, but if the human and the AI are closely matched, the AI will tend to speed up a little if the human is in the lead.) Its important to understand that I'm not talking absolutes here (i.e. these are not rigid rules about what happens) - I'm talking about tendencies in the AI behaviour. It's all quite clever and subtle and it makes for better racing I think. I'm sure there will be some exceptions in races you've all had - but from the data I've collected from all my races this seems to be roughly what's going on. I also concede that my observations may be complete rubbish and I'd love be set straight by the developers if that's the case!! Apologies for long post but I think the above might explain what haik experienced at Bathurst: The (partial) solution for haik would be to turn up the AI a notch, so that an AI car in first place would be doing 2:07.500/2.08.00s.
Far from 100%, but I believe that was the principle of the line in the PLR file called "Player Car Equal". I'm not sure what the default setting was, but testing this variable showed the field spread was much higher with a lower number and vice-versa, I believe matching the player's times more evenly. That said, it was a long time ago, so maybe my memory is failing me there.
I find the AI to be top notch, and even when you get fender to fender, they keep on racing. Now if youre stupid slow into a corner that should be taken at speed, then yes, the AI will have an issue with you. Ive had some thrilling races with just the single player mode.
I can't get on with the AI at all. I can't get close to them approaching a corner as they tend to pull up unrealistically right on the corner entry and I end up 9 times out of 10 rear-ending them. Not good practice for club racing.
Everyone with problems of any kind, should delete the AI adoption file first. This sometimes comes corrupt after to many updates.
C:\Users\xxxxx\Documents\My Games\SimBin\RaceRoom Racing Experience\UserData\Player1\aiadaptation If you use adaptive AI, they need 3 short races to adapt your skills.
Old Thread! In 2022, the GT3 AI still rubber band past me, like they have a turbo built in, some of their overtakes on corner exit is astounding. They seem to have a lot of grip everywhere. I'm not the fastest, even when i get a good exit they still find a way to zip past a lot of the time. Still to play tho. I do not use Adaptive AI.
Tune down the AI. Go to 90, and then correct upwards in jumps of 2, later 1. You may also find that you might prefer slightly different settings with various car classes on different tracks. Same car class on different tracks: on some tracks I fight for ranks amongst the top quarter of the field, and with same cars and AI settings on another track I am left behind in the dust. And no, RR has definitely no rubber band AI, but imo the best AI in business (although I read a lot of good things about the new AI they inject into iR, but I do not know that title). If you want to see a rubber band AI, play GT Sports or Papyrus Nascar 2003. Then you see what the term "rubber band AI" really means.
If they overtake you on exit, you're either messing up your exit or you're exiting a mid-speed/high speed corner. It's a common thing that AI are faster in high-speed corners and slower in low-speed corners.
Delete this to begin with and avoid adaptive AI Documents\My Games\SimBin\RaceRoom Racing Experience\UserData\Player1\aiadaptation.xml Maybe your AI are messed up.