AI "cheating" with qualification time

Discussion in 'Bug Reports' started by Redglyph, Nov 21, 2015.

  1. Redglyph

    Redglyph Registered

    Joined:
    Aug 28, 2015
    Messages:
    43
    Likes Received:
    10
    Hello,

    I frequently see that, while I'm ahead when driving early in the qualifications, when I skip to the end of the session the AI has suddenly registered much quicker laps. Then when I start the race, I'm at the end of the grid or in the middle, it's pretty easy to catch up and they are obviously not up to their claimed lap times.

    For example, it happened at Silverstone, with best AI 5 seconds slower. In the qualifications they got better, and again consistently in the race they were much slower. 5 seconds, a big difference.

    I don't mind increasing the AI settings to have a fair game, but I don't like the idea of giving a random time in qualifications when skipping the rest of the session, there must be a fault in the code generating the final AI qualification time.

    Not sure this problem has been addressed, I saw old threads reporting similar issues but the replies were unclear.
     
  2. redapg

    redapg Registered

    Joined:
    Jan 16, 2012
    Messages:
    3,958
    Likes Received:
    2,808
    I don't know how it is programmed but maybe it has to do with the increasing grip, the longer the sessions lasts.
    So if you skip to the end, this will maybe included into the final laptime calculations.
     
  3. MarcG

    MarcG Registered

    Joined:
    Oct 4, 2010
    Messages:
    6,854
    Likes Received:
    2,234
    Don't EVER click "Finish Session" as their times are seemingly randomly generated, either finish the session in real time or time acceleration (Which is fine) and they'll qualify as normal, it's always been this way FYI.
     
  4. Redglyph

    Redglyph Registered

    Joined:
    Aug 28, 2015
    Messages:
    43
    Likes Received:
    10
    No, they already had the time to do at least 3 laps, and the session wasn't that long, 30 minutes I think, the track can't have changed significantly. Not 5 seconds anyway.

    Ah thanks, I should try that! :)

    It'd be much better if they fixed the random part though.
     
  5. Guimengo

    Guimengo Guest

    It's something happening since rFactor and in the rFactor engine-based titles.
     
  6. redapg

    redapg Registered

    Joined:
    Jan 16, 2012
    Messages:
    3,958
    Likes Received:
    2,808
    That depends on your session settings and how the Real Road gets generated and the Multiplier that you choose.
    But as said before, i don't know how it gets calculated.
    And MarcG's suggestion should work better.
     
  7. Lazza

    Lazza Registered

    Joined:
    Oct 5, 2010
    Messages:
    12,345
    Likes Received:
    6,572
    The only way to properly work out how fast simulated cars can go around a track is to 'drive' them around the track. Any sort of skipping will lead to unrealistic results. At least rF/2 gives you the chance to do it properly, instead of some games that never do it properly ;)
     
  8. MarcG

    MarcG Registered

    Joined:
    Oct 4, 2010
    Messages:
    6,854
    Likes Received:
    2,234
    Probably not as easy as it sounds and it's fairly null & void when all you have to to is Time Acceleration anyway.
     
  9. Redglyph

    Redglyph Registered

    Joined:
    Aug 28, 2015
    Messages:
    43
    Likes Received:
    10
    I've tried time acceleration and it's not more than an annoying work-around. You still have to sit there for a *very* long time waiting for the session to finish, it doesn't accelerate time that much.

    I can see it would be difficult to obtain the precise results the AI would achieve by driving the whole session, but maybe that's not necessary. When the user skips, they take the risk to see some discrepancies, but in this case they are just too large. Obviously they have thought about a level-based random result, it just needs tweaking.

    Or if it were possible to accelerate more, at the limit without the graphics that must take a big part, and parallelizing the multiple opponents, but that would require extra development, not sure it's worth it.
     
  10. Lazza

    Lazza Registered

    Joined:
    Oct 5, 2010
    Messages:
    12,345
    Likes Received:
    6,572
    Try doing a shorter session if you don't want to drive many laps in it. Then you don't have to wait so long.

    It's true the estimate could probably be made closer, but it wouldn't stop the issue. Even half a second could drastically change the results in a tight field.

    I don't believe the graphics are playing much of a part with the limited time acceleration. Apart from some minor simplification, each AI is running around with rF1-player-physics (but unlike a player, the game also has to work out what the AI driver is doing) so it can only be sped up so much. The more you try to shortcut the simulation of AI cars the more they will vary from track to track (and between different conditions) relative to a human driver.
     
  11. Frenky

    Frenky Registered

    Joined:
    Jul 10, 2014
    Messages:
    383
    Likes Received:
    157
    It depends on the mod.

    In the HDV file, there is a line
    AITableParams=(39800, 19.5) // PLR file overrides for (max load, min radius) used when computing performance estimate tables - only positive values will override, see PLR for defaults

    Lower the defaults slightly and the increase in lap times by clicking finish session is less.
     
  12. peterchen

    peterchen Registered

    Joined:
    Oct 25, 2011
    Messages:
    2,099
    Likes Received:
    287
    How should this have something to do with the estimated times after skipping??
    Please explain!
     
  13. Marc Collins

    Marc Collins Registered

    Joined:
    Jan 11, 2012
    Messages:
    3,159
    Likes Received:
    162
    Get used to people suggesting annoying work-arounds for all sorts of deficiencies in rF2. It's the best sim available, technically, but full of these sorts of aggravations and bugs. Some of the old-timers here prefer to work-around things rather than incessantly badgering ISI to fix things. You'll also learn there's a reason for that. And for the record, this issue that you point out has been annoying me since 2006 when I got the original rFactor.
     
  14. Lazza

    Lazza Registered

    Joined:
    Oct 5, 2010
    Messages:
    12,345
    Likes Received:
    6,572
    You have to simulate a car going around a live track (realroad etc) in order to produce realistic results. You can't shortcut this stuff. As I said above, at least rF/2 gives you the option to accelerate the AI and make them do proper laps. Many games just fake it, and some of those cover the faking by using your own times to generate the AI ones...

    Accelerating time isn't a workaround; it's how you do it properly. As I said, if you don't want to run much of the session then make it shorter, or be prepared to give it a few minutes to finish. No estimation will ever get it right across different tracks and conditions.
     
  15. MarcG

    MarcG Registered

    Joined:
    Oct 4, 2010
    Messages:
    6,854
    Likes Received:
    2,234

    You do enough badgering of ISI for all of us ;) no game is perfect, it is the way it is, either deal with it and get on with playing or don't, the choice is yours.
     
  16. Redglyph

    Redglyph Registered

    Joined:
    Aug 28, 2015
    Messages:
    43
    Likes Received:
    10
    A 3rd option would be to improve the sim, but maybe it's just me ;)

    I have to agree with Marc Collins though. Despite all the opinions saying the AI cars needs to be simulated in order to get more precise results, I wouldn't go so far, as I said previously a user ending the session would implicitly accept slightly imprecise AI lap times, and may accept an accelerated simulation to get them (30 seconds to 1 minute should be fine). Any other sim I know (R3E, AC, pCARS) can do it, so there's obviously room for improvement, I don't want to sit idle for 30 minutes waiting for the AI and I'm pretty sure I'm not the only one.

    I'm pretty sure a pre-calculated table could do the trick, by gathering statistics over a few circuits and correlating that to others. It could be updated during actual races to improve with time.
     
  17. MarcG

    MarcG Registered

    Joined:
    Oct 4, 2010
    Messages:
    6,854
    Likes Received:
    2,234
    Of course improving the sim would help, but ISI don't have the resources of a major game developer to have one guy working on all AI aspects 24/7, that's just the way it is and we have to live with that. We all want better but we can't expect the world ;)

    Time acceleration doesn't take 30 minutes anyway, not quite sure what you're implying there, also this isn't pCars, AC or R3E - just because they can do one thing doesn't mean RF2 can and vice versa.
     
  18. Juergen-BY

    Juergen-BY Registered

    Joined:
    Jun 16, 2012
    Messages:
    3,089
    Likes Received:
    440
    As others allready pointed out, 30 minute session length could gain a hugh additional grip on track...depends on how many AI`s you have selected and on your realroad settings. Wich would make a bit difference in lap times.
     
  19. Redglyph

    Redglyph Registered

    Joined:
    Aug 28, 2015
    Messages:
    43
    Likes Received:
    10
    Sorry, but already mentioned, already answered. Besides, they obviously have another pace in the race where all parameters are equal for everyone, so such a big difference has nothing to do with slightly more grip on the track.
     

Share This Page