Is there some sort of AI catchup/slowdown feature

Discussion in 'General Discussion' started by Sky8lade, May 7, 2021.

Tags:
  1. Sky8lade

    Sky8lade Registered

    Joined:
    Apr 29, 2021
    Messages:
    10
    Likes Received:
    0
    I am new to rFactor2. I done a couple of races at Spa after a lot of practice and getting used to everything in the game. I wanted to know if this game has some kind of AI catchup/slowdown feature. Basically I was leading the race and had a few seconds gap lead. I noticed that the AI never fell too far behind even when I set a faster lap. Then I made a mistake and fell to fifth position. However I caught the AI drivers in no time at all. After the race looked at the lap times and saw that the AI go faster if I'm in front and slow down dramatically if they are leading. Is this a "thing" in RFactor2 and if so, is there a way to turn it off?
     
  2. pilAUTO

    pilAUTO Registered

    Joined:
    Nov 7, 2020
    Messages:
    674
    Likes Received:
    618
    Sky8lade likes this.
  3. Sky8lade

    Sky8lade Registered

    Joined:
    Apr 29, 2021
    Messages:
    10
    Likes Received:
    0
    Interesting thanks, seems like it's not possible to turn off and not something that is going to be developed anytime soon. For offline racing it's quite important as in my case playing online is a bad experience due to high server ping rates as I don't stay in a country which is close to any servers.
     
  4. Lazza

    Lazza Registered

    Joined:
    Oct 5, 2010
    Messages:
    12,476
    Likes Received:
    6,651
    @Sky8lade pilAUTO is (was) DrivingFast. I think he has misunderstood you here, and his wishlist item isn't about your issue.

    I'm pretty sure AI rubberbanding can be switched off, just can't remember the details right now. I'll have a look later if no one else responds.
     
  5. marvelharvey

    marvelharvey Registered

    Joined:
    Apr 25, 2013
    Messages:
    410
    Likes Received:
    283
    I forget the name, but it's the bottom most option in the AI settings, under Number of Opponents, Strength and Aggression. Set it to 0% and there should be no catch up.
     
  6. DucMan888

    DucMan888 Registered

    Joined:
    May 5, 2017
    Messages:
    62
    Likes Received:
    73
    I believe you are referring to AI Limiter, this is the description as found in the json file. Just a note that setting the value in game will read a whole numbers as a percentage, while in the json file it is a decimal (5 in game will be 0.05 in json)

    "AI Limiter#":"Range: 0.0 (no limiting) - 1.0 (limiting used to make racing closer but also make more driver differences on flat-out tracks)",
     
    Last edited: May 8, 2021
  7. Lazza

    Lazza Registered

    Joined:
    Oct 5, 2010
    Messages:
    12,476
    Likes Received:
    6,651
    @DucMan888 and @marvelharvey I'm not sure this is the same thing.

    As far as I know AI limiter on 0 will allow each AI driver (via the talent files) to drive to their own speeds fully. 1 basically makes them all the same 'speed' (talent). I believe neither option has anything to do with matching the player's laptimes, unless that's an underlying 'feature'.

    And now I think about it, the small amount of testing I've done (in devmode, but should be the same) with AI showed no variation in their laptimes whether I was driving fast or slow.

    I'd be interested to hear from people who routinely race against AI to confirm whether this is a thing in rF2.

    I have vague memories of an explicit setting for this in rF1, but can only see the same AI limiter parameter from a quick look. So I'm not sure at all.
     
  8. pilAUTO

    pilAUTO Registered

    Joined:
    Nov 7, 2020
    Messages:
    674
    Likes Received:
    618
    Indeed I misunderstood the original question. I wish the feature I'm looking for was sought after by other people.

    Regarding the AI LIMITER, here is what I understood :

    - For each mod there are "strengh" values FOR EACH AI which are defined. So some AIs are stronger than others.

    - The AI LIMITER, the higher it is, reduces the difference between the strength of the AIs predefined in the mod, until MAYBE make all the AIs in the mod have the same force during the race, for the value 100% of AI LIMITER.

    - Setting a high value of AI LIMITER will therefore ALSO have the effect of reducing the spread of AIs on the runway. - Moreover, but I'm not sure, I was told that AI lap times were more variable and therefore more "amateur" and less "professional" when the value of AI LIMITER increases.
    - Finally, as the AI of rF2 is not very good, I suppose that setting an AI LIMITER value too high could (MAY BE) reduce the quality of the AI because the different AIs would be too close to each other, and could hardly be doubled.


    Also, the AI MISTAKES in the player.json file set to 1.0 gives very good results, in the sense that the AI is more "human like" even if certainly less "Professional".
     
  9. Roninho

    Roninho Registered

    Joined:
    May 9, 2021
    Messages:
    62
    Likes Received:
    30
    I have tested the AI LIMITER function. These are my findings:

    Test setttings
    Mod: CART 98
    This mod has rcd files (talent files) which work. Difference between the fastest guys and slowest guys is 1.5-1.7 seconds in practice, with 1 guy always being a few tenths slower at 2.0-2.1 seconds of the pace (PJ Jones in the underpowered AAR Toyota)
    Track: Indy 500 oval
    I tested it on an oval since according to the json file it's effect is more for flatout tracks)
    AI limiter settings: 0.0, 0.10, 0.22
    I started at 0.22 because in another topic on AI performance someone suggested some player setting for great ai racing including the AI limiter setting of 0.22 . For clarity, i had 0.22 in the player.json and changed the setting in game to 0.0 or 0.10. I did not do this in the player file (no idea if this would have had a different outcome).
    Other settings:
    AI strength at 83%, AI aggression at 52%
    Test runs: Multiple
    I ran multiple test runs. A test run was a race with 27 AI and myself. I started at the back and parked the car in the pits after lap 1 and then watched the race with accelerated time on. Longest runs where full 200 laps, shortest were around 50. I probably did a dozen test runs.

    My findings:
    AI limiter at 0.22:
    This was the value i started at. What i noticed was that the AI had great fights, with a bunch of passes for the lead and also a lot of 3 wide/3 car fights, while still respecting each other (no crashes). So that was nice.

    However the bad thing (and also the reason to start testing with AI limiter) was that there was no clear seperation between the cars. Essentially before the 1st pitstops you'd have a long line of 27 cars staying together, like a train with a bunch of wagons. As i said with a lot of passing, but after 35 laps the leader and the 27th placed guy were within a couple of seconds. So even PJ Jones, the guy constantly 2 seconds of the pace in practice could keep up with everyone.
    Some parts of this train broke up after the pitstops due to guys who lost a few seconds in the pits lost the 'tow' and could not keep up and they formed their own minitrains.

    All in all a setting which for me was keeping the guys to close together and probably a bit to much passing as well.

    AI limiter at 0.0:
    Using AI limiter setting of 0.0 resulted in much clearer difference between the cars, and quickly 'minitrains' would start to form. It was always Andretti, Fittipaldi and Unser Jr. who would be up front on their own (all spread out) and then 3 mini trains. Those mini trains would have the same subtopguys in train 1, midpack guys in train 2 and backmaker guys in train 3. And PJ Jones would run by himself at the back of the field. The difference between these groups would slowly increase, with later in the race mini trains breaking up due to pitstops.

    The negative of this setting was that there was no passing. At all. As in no passes, 0, nada, niente. You can see the positions per lap of the guys in the UI and those would just be straight lines with only the pitstops have an effect.

    Basically, this setting would be useless for offline racing because there would be no passing at all.


    AI limiter at 0.10:
    This setting produced a nice mix of both worlds: You would see guys losing contact with the train and get a couple of different groups, but still also see passing. And still have the guys perform close to what you expect from their RCD/talent file and car performance combo.


    So in short:
    AI limiter at 0.0: Spread out according to the RCD files & vehicle performance, but no passing. 0 passes in 35 laps.

    AI limiter at high setting: Racing is much, much closer with a lot of passing, but no breaks/cars stay together. After 35 laps just 5 seconds between p1 and p27

    AI limiter somewhere in between: Mix of both, so spread out somewhat according to talent file, but close enough to actually race and pass. After 35 laps probably somewhere between 50 and 100 passes in total, and also gaps between the fastests and slowest cars. Only the subtop to midpack guys stayed together as long as they kept into the 'tow'.

    Hope this helps.
     
    shovas and Hazi like this.
  10. hitm4k3r

    hitm4k3r Registered

    Joined:
    Jun 15, 2016
    Messages:
    1,324
    Likes Received:
    3,130
    There are some points to consider with your test though. If you run a qualy session before the race at Indy Oval, there is a good chance that the field is "sorted" by car strength and talent once you go into the race, so overtaking with AI Limiter at 0 % is obviously alot less likely. To really test this it might be good to use a spec series mod with talent files, where each AI car has the same physics file. As you have noticed, even PJ Jones in the worst car could keep up, as Oval racing is a completely unique topic for itself. The importance of pack racing and staying in the tow has much more influence than on a high speed roadcourse like Monza with a considerable amount of room for error compared to a Superspeedway.

    Other than that, some good observations. It's allways a matter of a combination of factors. You used a very high agression setting with 53 %, wich will obviously have a big affect once you start to play with the AI limiter. Then we get factors like dirty air in high DF cars compared to Stockcars etc. The list is basicly endless
     

Share This Page