Author Topic: Stunts goes Dtm  (Read 1303 times)

Overdrijf

  • Joe's Friend
  • **
  • Posts: 481
  • Professional idiot
    • View Profile
Re: Stunts goes Dtm
« Reply #15 on: November 26, 2018, 06:21:02 PM »
Demmm.

Only now, after a month of Race for Kicks, copying the scoreboard to save it, I notice I called the M3 the M30.

That's it, I'm doing a version 1.1. And I'm doing it before Dreadnaut has a chance to select the "M30" for next year.

dreadnaut

  • Skid`s Nightmare
  • **
  • Posts: 717
    • View Profile
    • chezDreadnaut
Re: Stunts goes Dtm
« Reply #16 on: November 27, 2018, 08:19:37 PM »
That's it, I'm doing a version 1.1. And I'm doing it before Dreadnaut has a chance to select the "M30" for next year.

Oh, M3 doesn't sound fast enough, we'll use the Audi then :P

Cas

  • Bernie's Brother
  • **
  • Posts: 263
    • View Profile
    • Dimioca Labs
Re: Stunts goes Dtm
« Reply #17 on: November 28, 2018, 09:02:25 PM »
Man, after this time on the community, I should have at least one car of my own design. I've never sat to analyse exactly what each thing does. The 3D design, I already understand. It's still a lot of work to do, but I can see how it's done. It's the car functioning that gets me confused... mostly. I'll be taking a deep read in the Wiki. Is there any important thing that I need to know about creating my own car that are not there?  I figure I should also read this subforum thoroughly!  Where to start! XD
« Last Edit: November 28, 2018, 09:04:36 PM by Cas »
Earth is my country. Science is my religion.

Overdrijf

  • Joe's Friend
  • **
  • Posts: 481
  • Professional idiot
    • View Profile
Re: Stunts goes Dtm
« Reply #18 on: November 29, 2018, 09:16:58 AM »
I think there's a few small holes in the wiki, but I couldn't really point them out. It's mostly in the 3D model stuff I think, the description of how the coordinates of a wheel work is for instance wrong, as far as I can tell. RES-file wise I've also got some maybe new insights in length and handling I've yet to post. I'll get around to that one of these days.

Most of the .RES stuff is actually pretty straightforward. Torque curve interacts only with gear ratio's, mass and air resistance. If you scale all of them to exactly twice their value you've got the exact same car (assuming the mass falls into the same power gear category.) The handling number interacts with car length (not sure about width) and surface grip modifiers. These actually don't scale the same way, which is what I need to post about, a short car with low handling number is different from a long car with a high handling number. But because speed and cornering are separated systems you're not going to screw up your cornering by working on your speed curve or the other way around.

The torque curve itself has some weird features, it's hard to design to an exact top speed, and if the curve goes down too fast at the end you open it up to slowdowns where you would expect speedups. But you'll notice that happening as you mod and test.
« Last Edit: November 29, 2018, 09:24:52 AM by Overdrijf »

Duplode

  • Breaths Stunts
  • ******
  • Posts: 3516
  • Through the astral door - to soar
    • View Profile
    • The Southern Cross Stunts Trophy
Re: Stunts goes Dtm
« Reply #19 on: November 30, 2018, 05:02:26 AM »
After racing with the three cars, I can confirm each of them has its own distinctive temperament. The Audi is smooth as butter, IMSA like as you say, and definitely the easiest of the three. The BMW is very nimble, and I'm fairly sure it beats the other two on a twisty track. All the responsiveness, however, means you end up being very close to the limit at all times. Another complication is the occasional curveball thrown by the bug-susceptible nature of the car. To me, the BMW is the most challenging of the set. Lastly, the Mercedes feels more like a typical "slow car", and is probably the one for which the conventional driving techniques translate more easily.



I think there's a few small holes in the wiki, but I couldn't really point them out. It's mostly in the 3D model stuff I think, the description of how the coordinates of a wheel work is for instance wrong, as far as I can tell. RES-file wise I've also got some maybe new insights in length and handling I've yet to post. I'll get around to that one of these days.

Looking forward to it  :)

The torque curve itself has some weird features, it's hard to design to an exact top speed [...]

Do you mean flat track top speed, or real top speed, or both?



One tiny issue I noticed is that the stripes over the hood from the BMW's paint job #9 flicker a lot. I'm not sure about what, if anything, can be done about it (I don't see anything obviously problematic when looking at the model on Stressed, though I didn't actually attempt tweaking it yet).


Overdrijf

  • Joe's Friend
  • **
  • Posts: 481
  • Professional idiot
    • View Profile
Re: Stunts goes Dtm
« Reply #20 on: November 30, 2018, 02:02:55 PM »
After racing with the three cars, I can confirm each of them has its own distinctive temperament. The Audi is smooth as butter, IMSA like as you say, and definitely the easiest of the three. The BMW is very nimble, and I'm fairly sure it beats the other two on a twisty track. All the responsiveness, however, means you end up being very close to the limit at all times. Another complication is the occasional curveball thrown by the bug-susceptible nature of the car. To me, the BMW is the most challenging of the set. Lastly, the Mercedes feels more like a typical "slow car", and is probably the one for which the conventional driving techniques translate more easily.

Thanks, good to hear feedback like that.

Quote
I think there's a few small holes in the wiki, but I couldn't really point them out. It's mostly in the 3D model stuff I think, the description of how the coordinates of a wheel work is for instance wrong, as far as I can tell. RES-file wise I've also got some maybe new insights in length and handling I've yet to post. I'll get around to that one of these days.

Looking forward to it  :)

I need to write a proper post with exact numbers and add the example cars so people can try it for themselves. The gist of it was this: I got two cars with the same handling/length, but one twice as long as the other. The short one steers a lot faster, more responsive. I figure it's related to the reason longer cars jump farther: When a Stunts car turns it doesn't turn by a certain number of degrees per 0.05 seconds, it turns by a certain distance per 0.05 seconds. For a shorter car the front end moving the same distance means it turning a larger angle, so it steers more quickly. The handling parameter seems to be in control more of how long you can keep cornering before the car starts slipping. So they still counteract each other, if you corner slower but can keep it up longer you can corner sort of around the same distance at the same speed.

I think...

Because this is where it gets a little weirder. All of the faster cars can do a single wide corner at around 150 miles an hour, despite their length or exact handling parameter. The superkart does it, the IMSA cars do it, the Indy does it. It's some sort of ceiling, a limit to cornering ability similar to the limit there is to top speed. I found this limit, or the general area of it, for a test version of the Mercedes. And that's where something weird happened. I had four versions of the car. The one with the lowest handling parameter was the slowest around a lap, by a small margin, but it felt quite good to drive. You had some room for error. The three versions above it had roughly the same lap times, but I think it was the second slowest one that usually took more tries than the others not to spin out. When you spun, you spun, no coming back from it. The car was more on edge. When the handling parameter rose further this behavior went away again. I dubbed this place "the sour spot", and I tried to use it on the DTM cars by putting all of them just below that spot. The idea is that the car with the least amount of handling gives the most warning when something is about to go wrong, creating a balance between where the edge of the car's ability is and how well a human driver can stick to that edge. The concept does not completely match your description above though. The BMW has the lowest handling and even the lowest handling/length while being the shortest. To me it feels short and responsive, maybe a little too much so, but I also get pretty good feedback, which is probably part of the reason I did my fastest NoRH lap with it. It sits the furthest from most other cars in feeling due to being so short, but that was less of an obstacle for me after all the testing I did with these. The Mercedes and the Audi have the same handling/length, so the longer Audi has the best handling stat, but it feels a little slower and smoother steering-wise than the others. The Audi to me also feels a little closer to the sour spot, it handles quite well, but I have trouble finding the edge before crossing it. I made quite a few NoRH attempts with that one.

My experience was probably tainted by the design ideas I tried to implement though. There's probably some wishful thinking involved.

That's basically it. I have no idea if the sour spot is even available for all lengths of car, or that it's a artifact of a specific unlucky combination, but I definitely feel like there's a limit to how much handling a car can handle before more handling stops doing it any good. The superkart for instance is massively over-handlinged. It would drive just as well with maybe as little as half its number or more. Have to look the proper number up to make a better estimate.

Quote
The torque curve itself has some weird features, it's hard to design to an exact top speed [...]

Do you mean flat track top speed, or real top speed, or both?

The combination, kind of. The BMW has a slightly lower max rpm than the other two with the same top gear ratio, yet I could not for the life of me get its real top speed (the highest of the tree) down just one or two mph, at the very least not without altering the air drag. I could give it slowdowns or lower its flat track top speed, but I think I could only lower that by 2 mph, not one. I think one of the things that's happening is that because the game works in discrete intervals of 0.05 seconds you don't actually touch every bar in the torque curve. You often end up skipping a bunch. That's also what causes the slowdowns I figure. You skip ahead to an rpm that has a really low power (just 2 ticks or so) the game makes its new calculations with that very low torque and drops you down quite a few rpm's on the next time step. But the problem with the top speeds is that it's not one or two bars that decide a particular top speed, just the ones that don't have enough power to speed up any further. If the bars lower on the curve are too high you can skip right over those two. High air drag helps, it makes the high speed behavior of a car more predictable, and a power curve that slopes down over a longer way, giving it a big gap between flat track and real top speed, probably helps as well in that regard. Or you should just not care too much about specific numbers and exact balance.

Quote
One tiny issue I noticed is that the stripes over the hood from the BMW's paint job #9 flicker a lot. I'm not sure about what, if anything, can be done about it (I don't see anything obviously problematic when looking at the model on Stressed, though I didn't actually attempt tweaking it yet).

I had some problems with that earlier. I didn't pay a lot of attention to it post release, but I'll happily believe it. I'm not sure what can be done either. I think part of the story was interference from the wheels and the sort of wheel houses. Since the BMW has the lowest hood it's the easiest to interfere with. The final form those wheelhouses take is pretty trimmed down but I also didn't want to leave them out completely.
« Last Edit: November 30, 2018, 02:06:21 PM by Overdrijf »

Duplode

  • Breaths Stunts
  • ******
  • Posts: 3516
  • Through the astral door - to soar
    • View Profile
    • The Southern Cross Stunts Trophy
Re: Stunts goes Dtm
« Reply #21 on: November 30, 2018, 09:28:09 PM »
I got two cars with the same handling/length, but one twice as long as the other. The short one steers a lot faster, more responsive. I figure it's related to the reason longer cars jump farther: When a Stunts car turns it doesn't turn by a certain number of degrees per 0.05 seconds, it turns by a certain distance per 0.05 seconds. For a shorter car the front end moving the same distance means it turning a larger angle, so it steers more quickly.

Car length does have a major effect on responsiveness, and as you say it is indeed related to jump distances. If I remember correctly, it goes like this: the game calculates displacements for each of the four wheels, and then, roughly speaking, averages them to compute the overall position of the car. Steering is actually done by changing an angle, but that angle is the direction on which the front wheels are moving. For a longer car, the same change of direction of the front wheels has less of an effect on the car as a whole over a  time step.

The handling parameter seems to be in control more of how long you can keep cornering before the car starts slipping. So they still counteract each other, if you corner slower but can keep it up longer you can corner sort of around the same distance at the same speed.

That is the right intuition. The handling/grip parameter tells you how hard you can steer (i.e. how large the steering angle of the front wheels can get) at a certain speed before skidding or losing control. I had found the exact formula for that, but I lost my notes about it in my HD crash in '16. In any case, longer cars require harder steering for changing the direction of the car at the same rate, so they can't take corners as fast.

All of the faster cars can do a single wide corner at around 150 miles an hour, despite their length or exact handling parameter.

Not quite. The limit speed of IMSA cars following the ideal line on large corners is slightly above 170mph; for the Indy, it is somewhere around 190mph. Such speeds are, as one might expect, very hard to attain without RH.

The sour spot theory is very interesting, and deserves further investigation. I feel it might also have something to do with steering responsiveness (i.e. how the car responds to steering input by the driver).

The BMW has a slightly lower max rpm than the other two with the same top gear ratio, yet I could not for the life of me get its real top speed (the highest of the tree) down just one or two mph, at the very least not without altering the air drag. I could give it slowdowns or lower its flat track top speed, but I think I could only lower that by 2 mph, not one. I think one of the things that's happening is that because the game works in discrete intervals of 0.05 seconds you don't actually touch every bar in the torque curve. You often end up skipping a bunch.

Besides that, the torque curve is made of segments which are 128RPM wide. If you set the redline at 8192RPM, there will be 40 segments; if the ratio of the top gear is set so that the real top speed is 200mph, the resolution of the real top speed will be 5mph per step, which is indeed quite coarse.