The TORCS Racing Board
Username: Password: Remember Me?
Lost Password Register
Author: wdbee | Created: 2017-08-28 19:15:35
Subject: Qualifying - Lap times - TORCS issue "first starter" and "rest of the starters"
Hi All.

Looking to the raw results of the qualifying, I see that the teammates have same lap times except for:

1. The first starter team has differences at all submittors.
2. My Team has one difference (at phi).

So the question is why?

For 1:
There is a TORCS bug, providing a different situation to the starter with index 0 (Resulting in the many differences of team Axiom). This is same for Windows and Linux (The differences are there before any command was returned by a bot). The result is same for all other teams (except 2.), so it is index 0 only that has another situation at start of qualifying.

For 2:
May I ask, what system is used for the phi result set?

Does anybody have an idea where to search for the TORCS bug?

Wolf-Dieter

Last Edited: 2017-08-28 19:16:32 by wdbee
    Author: phi | Created: 2017-08-28 21:18:04
    Subject: Re: Qualifying - Lap times - TORCS issue "first starter" and "rest of the starters"
    Win7 32 bits
    P4 Dual Core 3 GHz
    2 GB RAM
    Last Edited: 2017-08-28 21:18:04 by phi
    Author: berniw | Created: 2017-08-28 23:05:15
    Subject: Re: Qualifying - Lap times - TORCS issue "first starter" and "rest of the starters"
    Hi

    Is this not only caused by the different start location as already discussed via e-mail (starting grid positions applied in qualifying)?

    Kind regards

    Bernhard
    Last Edited: 2017-08-28 23:05:15 by berniw
      Author: wdbee | Created: 2017-08-29 10:22:00
      Subject: Re: Qualifying - Lap times - TORCS issue "first starter" and "rest of the starters"
      Well you answered to my first question, that the starting position is different, but as I noticed, that only the first result (index = 0) is different while all others result in the same laptime (Index > 0) I checked it and found, that there is no different starting position (Was a fake news ;) ).

      For me it looks like the initialisation of the situation is different (may be incomplete?) for the first (or for all other) starter(s).

      We have this bug since we use the tyre model.



      Last Edited: 2017-08-29 10:22:00 by wdbee
    Author: phi | Created: 2017-08-31 10:22:31
    Subject: Re: Qualifying - Lap times - TORCS issue "first starter" and "rest of the starters"
    I did 23 tests. While Lupo's time remains the same (1:06.390), Lobo's time varies:

    1:06.390 ==> 15
    1:06.324 ==> 8

    I do not know if this difference is significant (not a random one) or not, but at "naked eye" does not seem to me.
    Last Edited: 2017-08-31 10:22:31 by phi
      Author: wdbee | Created: 2017-08-31 12:29:56
      Subject: Re: Qualifying - Lap times - TORCS issue "first starter" and "rest of the starters"
      This bug blocks the learning. While learning the car is always running with index 0. But except the first starter in qualifying all cars have an index > 0.

      If I try the best results I get from learning and I setup two cars for qualifying, the first reproduces the learend lap time, but the second differes and most times there is a large gap in between.

      So I have to use not the best qualifying setup from learning but one, that results in small gaps between the cars. And to be honest, this means that the qualifying time is some thenth slower (more than the difference to the next opponent!).

      And here as well the time is not a random one. If I retry, the both lap times I get (for first and second car) are same for each test. This means there is a reason and not just an uninitialized variable.

      Your system is the only one, that does not show always the same laptime for Index > 0. All the others (Linux and Windows) do show same lap time for both cars (having both Index > 0) and show different laptime for Index = 0 and Index > 0.

      Last Edited: 2017-08-31 12:29:56 by wdbee
        Author: phi | Created: 2017-09-01 17:32:47
        Subject: Re: Qualifying - Lap times - TORCS issue "first starter" and "rest of the starters"
        A) I hope this can be resolved. In my short tests the car 'index = 0' does better time than it is in position 'index > 0'. It does not really make much sense... I think.

        B) I think you missed E-Track 6: the DummyDrivers are different in 3 machines, and the DanDroid are different in 2 machines. It's not just my machine generating different times, and, as you can see from my rudimentary data, it only generates different times in 35% of cases; I suppose the other machines also have a probability associated with different times, but, like mine, they tend to equal times.
        Last Edited: 2017-09-01 17:32:47 by phi