Can weaving be detected and penalized?

786
United States
United States
The bane of public rooms (or, one of many, anyway!) is the game's inability to detect and punish aggressive weaving from a leading car on a trailing car. Do you think the game could be changed to achieve it?

I am particularly talking about straights. There's probably just too much difficulty in detecting a change of line in a corner whether it is deliberate or a loss of grip, but straights, that's another story!

The thing that occurs to me is the AI... They seem to be line aware, presumably there's some mechanism in the code to hold them to a line when two or three abreast. Could this same line be used to detect when a player car veers from one line to another repeatedly? And, does it need to know proximity to a trailing car to penalize it, as opposed to a trailing car that can make as many moves as it wants to on a leading car on a straight?

Online is awash in players that weave like madmen, or force a clean passing car off track on straights. Cutting is reasonably dealt with, divebombing too, if contact engenders a pass, but weaving and force-offs on straights seems consequence-less. Do you think it could be addressed?
 
"weaving" can just be guys on pads struggling to keep it straight or a dodgy ping @ times :( the other night a guy joined chatted all friendly I had a bad start but caught him up and had a real good race for 3 laps but as I passed him on the finale straight he weaved into me and blamed "adrenaline" no system can account for tools like that after all michael Schumacher got 7 world champs driving like that....
 
Deliberate weaving is pretty obvious to spot. Guy doesn't start weaving and maintains good car control until you are behind him. I already mentioned that this is for straights only. You saying you can't drive a straight line?!

And yes, actually, I do think a system could detect a sudden change in direction on a straight when a car is overlapped. It detects overlapped cars for the AI and prevents them hammering each other unless in really crowded conditions. You don't see the AI crowding each other off when just the two of them. Comparing expected AI input with driver input on straights shouldn't be a huge stretch...
 
Most likely scenario for autodetecting something like that(rough estimate numbers):
Jerks identified and penalized 10% of the time
People driving normally and properly in a good battle penalized 30% of the time
Victims getting attacked by jerks penalized 50% of the time


Weaving is a textbook case of behavior that is going to be incredibly difficult for the computer to detect on its own. Even if you do come up with a method to try to stop it, people will learn what it does and doesn't see and adjust their jerk strategies around it.
 
You sound like Clarkson "how hard can it be" well it turns out its pretty dang hard. Now lets see your best effort!
I'm going to say it's pretty much this. It would have to track both the car in front's positioning and car behind's as well as judging how many moves is too far and what a move actually is. That's before you link them all together in a racing situation which will then cause him to make a thread "Weaving can be detected. Why doesn't it work?"
 
Repeated left to right steering input on a straight when NOT within a certain distance of the leading car, AND within a certain distance of a trailing car. Not exactly rocket science. The game already seems to understand about not having an overlap when a dive is made with contact, so some kind of positional awareness seems to be in the game already.

You guys probably have an opinion about guitarists, or football players. I guess you would be content to be jeered at with 'Well, let's see YOU do it!' when you offer that opinion.

Tell you what, you guys know so much (enough to tell me it can't be done), show me YOUR work that proves it can't.

This game has already shown that many things we USED to think couldn't be achieved in a racing game actually can. I am just a bit more positive than you that those wizards can pull off something YOU obviously don't have the ability to.
 
Too many calculations to put on an already over worked console.

Both consoles have 8 cores, the Box runs @1.75 GHz, PS4 around 1.60 GHz there just simply isn't enough processing power.

Livetrack 3.0 probably takes half the cores by itself which doesn't leave alot of ceiling for physics, graphics etc.

I do like any idea that improves multiplayer, it's not that it cannot be done it's just the imagination is limited by hardware.
 
You guys probably have an opinion about guitarists, or football players.
I couldn't care less about guitarists. Sorry to rain on your parade. That wasn't the point I was making though. You're making it seem incredibly simple. If it were, every single racing game would implement it. But no. There are a number of constantly changing variables.

Repeated left to right steering input on a straight when NOT within a certain distance of the leading car, AND within a certain distance of a trailing car. Not exactly rocket science. The game already seems to understand about not having an overlap when a dive is made with contact, so some kind of positional awareness seems to be in the game already.
Nope. It warns both drivers about contact and doesn't apportion blame. Because it can't.

Ok. As it's so easy, what would be your distance in metres? How much movement would constitute a change of direction for a weave and not just a controller movement tap? How many of these movements would constitute a penalty? What's the time limit that would differentiate between a weave and just movements?

Tell you what, you guys know so much (enough to tell me it can't be done), show me YOUR work that proves it can't.
I never said it can't. I said it would require a lot of work/is very difficult and so did @FOG-Smokebeer.

Also, you can't prove a negative hypothesis.

I am just a bit more positive than you
That's a good one.
 
Tell you what, you guys know so much (enough to tell me it can't be done), show me YOUR work that proves it can't.

That's not how proof works. And that's without getting into the strawman that that's not what was said. It can be done, but it's WAY harder than you're making it sound.

If it was as easy as you make it sound, someone would have knocked it out over an afternoon. Presumably it wouldn't be that hard for you to knock up some pseudocode, if it was that easy. But I think when you do, you'll see just how many factors and dependencies there are, and how much changing variables within such a system drastically changes the behaviour.

You could make A system that does something along these lines, but one that actually relatively consistently detects actions that a human would consider weaving with limited numbers of false positives? Really quite hard. Not impossible, just time consuming for what is a pretty limited benefit.

Personally, I think you'd have better success trying to train a neural network to recognise the appropriate object behaviours, but good luck getting the spare compute power to run that in real time with a large field of cars.
 
I am not trying to assume that people with little coding skill are those going to attempt this. I can listen to all of you all day long if I want THAT opinion! I think that people with great coding skill (like those that have made this game do things I am quite sure you would all have said was 'very difficult' before they did it) could achieve this...

Oh ye of little faith!
 
What's even your point anymore?

To demonstrate your obtuseness...

But, I guess if I have to explain it in small words and simple sentences, let it not be said I didn't try...

You say you know of no game at the moment that can detect weaving, therefore it probably can't be done. You would have said, before SMS came out with Livetrack, that because no game had it, it probably couldn't be done.

Therefore, your ability to decide what can and can't be done is pretty flawed. To be polite about it.

Need any help with that, you are on your own!
 
therefore it probably can't be done
How many times do I have to tell you? I never said that.
it probably couldn't be done.
Or that.
Therefore, your ability to decide what can and can't be done is pretty flawed.
And your ability to understand what I think is even more flawed.

I agreed with somebody who said it was difficult. Not impossible.
You sound like Clarkson "how hard can it be" well it turns out its pretty dang hard. Now lets see your best effort!

I'm going to say it's pretty much this. It would have to track both the car in front's positioning and car behind's as well as judging how many moves is too far and what a move actually is. That's before you link them all together in a racing situation which will then cause him to make a thread "Weaving can be detected. Why doesn't it work?"

I said it would require a lot of work/is very difficult and so did @FOG-Smokebeer.

This is the only time I said can't.
Nope. It warns both drivers about contact and doesn't apportion blame. Because it can't.
And it was in relation to the current system as it's neither designed nor programmed.
 
Back