1. Hello Guest,

    We wanted to let you know that the NXS17 templates and viewers have been released!

    Grab them HERE
  2. Hello Guest,

    We wanted to let you know that besides the NXS17 templates and viewers, now we have released the NXS17 Mod as well!

    Grab the latest files HERE

2017 AI Balancing Project

Discussion in 'NR2003 General Discussion' started by RacerXero84, Apr 12, 2017.

  1. RacerXero84

    RacerXero84 Grease Monkey

    Messages:
    2,415
    Likes Received:
    2,344
    Trophy Points:
    113
    Even on my PC (which is no slouch) the FPS take a nose dive if it's the one I'm thinking of. I may revisit it, although last time I ran it, my frames bombed to like 25-35 FPS from my usual 60+ everywhere else.
     
  2. Mitchroberts96

    Mitchroberts96 Member

    Messages:
    7
    Likes Received:
    1
    Trophy Points:
    3
    I'm making an offline game folder with some different tracks. Would you be open to working on those too? Also, I'd love to help in any way possible.
     
  3. RacerXero84

    RacerXero84 Grease Monkey

    Messages:
    2,415
    Likes Received:
    2,344
    Trophy Points:
    113
    I need people to test the tracks, preferably with the carsets, on 100% difficulty so that I know they are where they need to be.

    As for working on additional tracks outside of what I have planned, it depends on how this project goes; if it works out, I will be open to other tracks after I complete the primary venues.
     
  4. Mitchroberts96

    Mitchroberts96 Member

    Messages:
    7
    Likes Received:
    1
    Trophy Points:
    3
    Awesome! I've also got a live updated 2017 MENCS car set. Want me to change their ratings so you can have those cars to use? Feel free to message me I've got lots of ways I think I can help you out.
     
    Chachow88 likes this.
  5. RacerXero84

    RacerXero84 Grease Monkey

    Messages:
    2,415
    Likes Received:
    2,344
    Trophy Points:
    113
    Quickie update: mid-terms bogging me down a bit, but I'm still plugging away at tracks. Added Pocono, Milwaukee, and Mansfield for testing. Pocono times/speed are estimated based on projected speed reduction that we have seen in some places so far.

    Slowly working on Daytona Rising. I will likely do a version that is "unrestricted" (Chassis=0), as that seems to emulate real world speeds and car behavior to mimic the 2017 Daytona 500, as the nr2003 SS chassis is simply too sluggish and doesn't act like what we see in real world racing anymore.
     
  6. Darren Ingram

    Darren Ingram Moderator Moderator

    Messages:
    2,009
    Likes Received:
    958
    Trophy Points:
    113
    Just did a 100 lapper at Bristol with the NXS17 cars you posted.

    100 percent difficulty, I think it might be a tad on the easier side. I qualified on pole with a 14.901, second qualified with a 15.129. 40th was a 15.430.

    During the race, I caught the rear of the field after about 15 laps, and got through lapping them pretty well. I stayed about 2.5 seconds ahead of second place most of the time. Lapped up until about 20th when I got caught up in a wreck on lap 51.

    From watching the replay, and being in the lapped traffic, the racing looked pretty good throughout the field. There was passing going on, and the AI weren't doing anything stupid.

    I enjoyed it though, definitely not a bad starting spot. I'm not the best driver, so given my results, I think this a tad on the easy side. I think the AI is almost spot on, so maybe lowering the user grip, or however you might slow down the user.

    I did just use the default Fast setup as well. Realistic damage and weather.

    Hope this helps!
     
    RacerXero84 likes this.
  7. RacerXero84

    RacerXero84 Grease Monkey

    Messages:
    2,415
    Likes Received:
    2,344
    Trophy Points:
    113
    Exactly the feedback I was looking for, thank you! I did leave the qualifying settings alone (for now) because that is a singular value that a user can modify based on their skill and starting preference; some people like starting mid pack, whereas others like to be able to compete for the pole, so that really is personal preference.

    If it is on the easier side of things (which I thought they might be), I can decrease tack grip, and increase AI grip, accel modifier, drag, and/or decel modifier based on where you are beating them based on:
    - beating them corner-off: increase accel
    - beating them corner entry: increase decel
    - beating them on the straights: decrease drag
    - beating them corner-center: increase their grip/decreasing track grip (depending on the user lap times, if they are realistic, I'd go for increasing AI grip)

    I'll make a note of it, and revisit the track this week. Again, I really appreciate the feedback!
     
    roostmotor87 and Darren Ingram like this.
  8. Lookey2000

    Lookey2000 Well-Known Member

    Messages:
    182
    Likes Received:
    119
    Trophy Points:
    43
    I agree with him, most of the tracks do seem pretty easy for being 100 percent difficulty. But other than that it is fun as hell
     
    RacerXero84 likes this.
  9. Darren Ingram

    Darren Ingram Moderator Moderator

    Messages:
    2,009
    Likes Received:
    958
    Trophy Points:
    113
    Just ran a few laps, and, in my case at least, the AI and myself are about even everywhere I can tell, except for corner exit. I'm gaining most of my time there it seems.
     
    RacerXero84 likes this.
  10. RacerXero84

    RacerXero84 Grease Monkey

    Messages:
    2,415
    Likes Received:
    2,344
    Trophy Points:
    113
    So I'm going back and looking and I'm seeing what you're talking about @Darren Ingram , at least with Bristol under gns physics. I'm decimating them corner off, how I missed that, I have no idea. So I went ahead and fixed that to a point, as I am noticing that the way the AI runs the track, it seems like they slow right before launching off the corner and into the straight, which is either something to do with the lateral settings or the LP file. Something I'll look into further prior to updating the INI on the OP.

    Thanks for the input! (this is what I need to hear from people)
    I'm benchmarking the racing where the player (for right now) is able to race in the lead, so long as they can beat the AI equally throughout the track, on each track. From there is just a matter then of reducing the track grip and increasing the AI grip/accel/decel/etc, in small increments to get the desired level of difficulty for 100%. Of course, it's not going to be perfect for everyone (varying skill levels) but if more people say it's too easy overall, then that is a quick and easy adjustment across the board for me to make. The tedious stuff is what I'm doing now.

    Bottom line is I have a vision of this being something where a user boots the game up, slaps it on 100%, and, without a good setup and track understanding, is, at best, a mid-pack or lower half car, no matter which track they attempt.
     
    Darren Ingram likes this.
  11. sabres123fan

    sabres123fan Active Member

    Messages:
    40
    Likes Received:
    15
    Trophy Points:
    8
    Mansfield piqued my interest since I was going to delete that track due to rough AI so I decided to give that one a try. Ran Cup there, on 100% difficulty my fast lap was about 4 tenths faster than second and during the race or practice sessions on an average lap I was 2 or 3 tenths faster than the next guy. I wouldn't exactly call myself a pro at this game either, much less on short tracks, so I would say this is probably too easy. Big issue I noticed was the AI would get off way sooner going into the corners than I would, by a good half second or even a full second sometimes. I was running into their rear ends in practice constantly but I'd be pretty even with them on the straightaways. In two 25 lap races I lapped up to 30th and 27th starting from the pole both times. Last thing I would note is twice a car washed up out of turn 4 when three-wide and crash into the hole in the wall coming to the frontstretch, I don't know how much of an .ini problem that is if at all but I felt it was worth mentioning. I'll test GNS and CTS in a minute and edit this with whatever I find.

    GNS/Xfinity: Only did a brief practice session but this was pretty good, same report as the Cup cars. Corner entry is where I gain the most time, it's not as pronounced in this mod though. Actually turned my fastest lap of the day in an Xfinity car lol. Anyway they're also a bit slower overall than the Cup guys, I'd say probably 5 to 6 tenths is the difference between me and 10th on an average lap.

    CTS/Trucks: Also did a brief practice session here. They're much slower overall, there's probably a second between me and about 5th place. They're just way slow overall, especially on corner entry, but the drag definitely needs to be messed with here. Not nearly as crash-happy as the old truck AI was, thank god.

    Overall I'd say the aggressiveness/raciness of all three mods/physics was pretty good. AI would probably push you if it was higher but they wouldn't be on the verge of wrecking you. For the most part the only accidents being caused are by me when I slam into cars braking earlier than I expected.
     
    Last edited: Apr 17, 2017
  12. RacerXero84

    RacerXero84 Grease Monkey

    Messages:
    2,415
    Likes Received:
    2,344
    Trophy Points:
    113
    I'm starting to wonder if it could be a difference in the papyai.ini that is causing such disparity between users.... maybe I should also upload my papyai.ini as well?

    Also, which carsets are you using?

    EDIT: added the PAPY AI INI file to the original post. If the tracks are WAY too easy it might be worth trying to test it with that as well
     
    Last edited: Apr 17, 2017
  13. sabres123fan

    sabres123fan Active Member

    Messages:
    40
    Likes Received:
    15
    Trophy Points:
    8
    Using the carsets you provided in the original message. I'll try out your papy_ai file and see if there's a big difference.
     
    RacerXero84 likes this.
  14. ThatOneShredder

    ThatOneShredder Well-Known Member

    Messages:
    1,769
    Likes Received:
    734
    Trophy Points:
    113
    Seeing as I've never been historically good at Phoenix (across all platforms), I decided to give your file a shot with the BR15 mod. I'm able to consistently run 27.1. While I race my carsets with a slightly modified ratings formula from Masgrafx, I feel that my feedback is still valuable. I ran about seven-tenths faster than second in practice, only a tenth faster in qualifying. During my 63-lap race (I crashed out on lap 24), here's what I noticed: the user's tire wear is there, but there's no speed falloff, at least from what I saw. The last lap before I crashed, I ran a 26.8. That's 23 laps into a run. The only reason I crashed was because I overdrove turn three, and drilled the wall with my rear end. In those 23 laps, I amassed a three second lead over second.

    Just food for thought. Maybe there could be a little more speed falloff. The wear line (F5 screen) is there, and it'll fade as the run goes on, but there wasn't any difference in speed.
     
  15. RacerXero84

    RacerXero84 Grease Monkey

    Messages:
    2,415
    Likes Received:
    2,344
    Trophy Points:
    113
    So what I am seeing generally is that on these smaller and non-cookie cutter tracks is I have to generally increase the difficulty fairly significantly (not really surprising, they are my weak point).
    I'm not seeing too many comments about the 1.5+ ovals other than 'a tad too easy' (which is a minor tweak).

    I am interested in some of the other aspects of feedback as well, like tire falloff, where the AI could be better (corner entry, etc.)

    I appreciate the feedback and I plan on getting to work and getting the areas pointed out shored up.
     
    Darren Ingram likes this.
  16. canadienhits

    canadienhits The Dominator, Cup champ. Moderator

    Messages:
    10,892
    Likes Received:
    3,715
    Trophy Points:
    113
    Just ran Auto Club. For Cup, I was tops of the test session, recording a 38.1, compaired to 2nd at 38.6, and last at 39.5. I got 203 on the frontstretch, with 165 in 1 and 2, and 195 on the backstretch, with 175 in 3 and 4. By comparison, 2nd got 199 on the frontstretch, 169 in 1 and 2, 193 on the backstretch, and 172 in 3 and 4. In Xfinity I was around 21st in my test session, with the top at 39.8, myself at 40.4, and the bottom at 41.1. I'd get 193 on the frontstretch, 151 in 1-2, 189 on the backstretch, and 165 in 3-4. The computer got 195 on the frontstretch, 163 in 1-2, 190 on the backstretch, and 175 in 3-4.

    Hope that helps some. Will work on Michigan tonight or tomorrow.
     
    RacerXero84 likes this.
  17. RacerXero84

    RacerXero84 Grease Monkey

    Messages:
    2,415
    Likes Received:
    2,344
    Trophy Points:
    113
    Little update for everyone, hope people are still testing particularly the 1.5 and 2 mile ovals... I need input on those please):

    On the shorter tracks and intermediates, I plan on essentially bumping the difference between myself and the AI to a .5 to a 1 second ADDITIONAL difference across the board based on feedback, focusing on the area where people said they were definitely beating the AI solidly. I expect to do a ZIP file bulk release of these tracks in the coming week, giving others some time to play around with the 1.5 and 2 milers so I can adjust them as well. If it goes as planned, I should have all tracks in the ZIP except for superspeedways and road courses for the moment.
     
    Darren Ingram likes this.
  18. canadienhits

    canadienhits The Dominator, Cup champ. Moderator

    Messages:
    10,892
    Likes Received:
    3,715
    Trophy Points:
    113
    Just did Michigan. For Xfinity, I was 2nd in my testing session with a 37.6, while tops was 37.1, and bottom was 39.7. I got 200 on the frontstretch, 181 in 1-2, 195 on the backstretch, and 182 in 3-4. The top AI got 198 on the frontstetch, 186 in 1-2, 196 on the backstretch, and 183 in 3-4. For Cup, I was 3rd in my session with a 35.7, with top being 35.5 and bottom being 37.3. I got 209 on the frontstretch, 190 in 1-2, 205 on the backstretch, and 194 in 3-4. The top got 208 on the frontstretch, 191 in 1-2, 205 in on the backstretch, and 194 in 3-4.

    I also noticed something odd on the Xfinity side. The AI would accelerate in the apex of the corner, but would slow down 10 mph coming out of 2 and 4. I didn't notice that on the Cup side, but it may have been there as well.
     
  19. RacerXero84

    RacerXero84 Grease Monkey

    Messages:
    2,415
    Likes Received:
    2,344
    Trophy Points:
    113
    If anyone has additional feedback, this is a good time in which to speak; I'm going through the tracks, generally making things more competitive (unless someone noted it was already tough, as was the case with @canadienhits at Auto Club using GNS physics, where I brought the AI a little bit more into line for that).

    Given some thought, if anyone else who is savvy with AI tweaking via the INI wants to make edits to the files I uploaded that they tested to get it more centered (if they notice a specific area that needs to be adjusted), feel free to upload. That would expedite the project quite a bit.

    In case anyone is curious, I am NOT giving up on this, I don't plan on dropping this project because it gets 'too time consuming', etc etc. I'm doing this for myself, something I MUST do in order to be happy with the game again. So this project is here to stay until completed to the best we can get it.

    Track progress.jpg
     
    Last edited: Apr 23, 2017 at 9:32 AM
  20. Kharvick29

    Kharvick29 Well-Known Member

    Messages:
    791
    Likes Received:
    159
    Trophy Points:
    43
    Did a 2 cup races at Bristol... didn't drive... just watched to see how the AI ran. Looks fine... the only problem I see is the AI is wrecking every 2-3 laps (every time on the start they wreck before the end of lap one). In a 100 lap race, longest green flag run was about 6 or 7 laps (until there was only about 28 cars left).
     

Share This Page