Car set-up wisdoms

Server was set to pretty cold weather (17 air and 22 tarmac as I can recall) so overheating street tires was not an issue.
Still not sure about cobra street tires, are they modern (radial) or old fashioned cross ply because radials have stiffer sidewalls = lower pressure have less impact on tire deflection ... some experimenting is needed.

 

Thanks @Nahkamarakatti
I watched your whole first video and a bit from the other one and actually got a setup that felt good for me with the bmw gt2 on Spa. I wasnt as fast, it actually differed a couple of seconds + which is pretty much from stock setup, but it made me feel the car more and gave me a bit more feeling for the car which will allow me to tweak it further. Unfortunately i forgot to save when i quit the game But just someone talking about it made more sense than reading about it, well at least for me. Thanks

 

I have only driven the Cobra with the default setup to date. However, since I always set the fuel load to 50% in my practice sessions (and even short online races), that translates to 80L in the Cobra because of its huge fuel tank. That probably helps stabilise the car a bit.

 

In one of my previous post I did mention things you need to know and decisions you need to make ...

In case of cobra it is where the biggest gains are to go faster.
It is not cornering, it is being able to use huge engine torque.
The car is traction limited big time!

Since there is multiple solutions, you need to choose right one that will not penalize you more than the gains are.

Bottom line is that penalties equals gains (time gains and penalties) but you feel more comfortable thus reducing a chance to make driving mistake so choose wisely.
Sometimes this is exactly the way for you to go faster.

 

Sadly we have no access to the best racing sensor available, the buttmeter.
Understanding and applying race setups in conjunction with telemetry numbers are fine as a good setup to go for a track and car combi, but to get the last percentages (thats what we're talking about I think) I'm still too much connected to my natural sensors too be able to finetune accordingly.

So what I do instead...
- 1. drive drive drive
- 2. Do not change and test more then one setting at once.
- 3. Take time to test the changes, means don't make early conclusion. It can also be because its simply not your day or round.
- 4. dont look too much on numbers
- 5. disable laptimes app when testing
- 6. Dont drink alcohol at all or too much coffee when setting up and testing setups

 

I should mention to review your telemetry. Braking on street cars requires a detailed feathering if any, gas on the AC cobra is tricky so watch out for torque drop and spikes.

The way i see it, guys who uses load cells are good with open wheel cars and pot pedals are shining for guys in the street and gt cars because pot pedals can be controlled more precise with muscle memory and pressure. Load cells on the other hand have a very little travel to modulate so it only goes by pressure, by the time you release some pedal pressure, your caliper pressure have lost by the multitude.

 

This has already been covered in an earlier post. As I said you'd have a lot of data to analyse to discover any "why"s you cared about. To claim you'd have no clue is ridiculously unimaginative.

Indeed, there are people who are creating setups that may or may not make them faster (they've probably run a few laps and convinced themselves they are) who have no more clue as to why it has made them faster.

 

Sheesh Bob, get a grip. And read the thread.

We're talking about the complexity in Assetto corsa, not curing cancer (which is actually not even possible, let alone easy) That you equate the two suggests you don't really understand.

It wouldn't require anything near "infinite processing power" to test the setups - especially in many of the cars that only have limited options (as mentioned numerous times through the thread)

The point was to show that the idea there's incredible complexity here that would take "years" to learn is simple fallacious.

Yes you did. Especially if you think a year is infinite

You think a processor that can do a few billion operations a second would, conservatively, take 5 seconds to do a lap?
Really? I'd up the length of your passwords Bob if you used the same logic there

The point here (again in the thread already) things that have millions of options are well within brute force capabilities. Billions is where you might need to be clever.

So, 5 options with 10 values. That's called "decimal" i.e you can represent each option as a 5 digit number. 00000 to 99999. Or 100000 options to test. It probably wouldn't take a human a year to test them.

It worth noting too that Stefano has, I'm sure, talked about in his videos about a program mode he has that lets him test the AI by making them run "every track/every car" - it doesn't take years to run, does it?

Hand-waving, imo. As said, physics is the same for everyone (except Eric Braithwaite) If a setup makes a car lap faster then it makes it lap faster.
Whether you can drive fast or not is really orthogonal - and I think this is even more so in a simulation.

Of course you can. The "AI" here would just be a set of finite inputs : like gas, brake and steering. The output is lap time. You don't need to "gauge" anything. Unless Assetto corsa has some randomness, it should be deterministic - i.e the same state conditions + the same control inputs should equal the same results.

Better to think of it as automation rather an AI because clearly some confuse AI as meaning "the crap drivers in racing games" and others that it's some kind of "will be sentient one day and take over the world" that Bill Gates et al are talking about. Neither are useful concepts to think about finding paths through states to search out something like 'minimum lap time'

As said though, you might hit some limits there. i.e you might need some decent processing power to start searching for optimum race lines and lap times.
However, no human drivers would get remotely close to running the number of laps you could by automating it and, as you suggest, most of them probably can't even drive the optimum setup. They'll decide some worse setup is better for them - but they'll be slower than optimum as a result.

Well, this is not really true. Of course it is possible. The only question is, what's your budget?
Most of the leading car manufacturers as well as several famous technology companies are working on automating driving. I can't imagine the lure of track will evade them forever if it hasn't already been something they've looked at.

Of course real life is more complex. There's no question there. Perhaps some saying it takes "years" to learn setups are thinking about real life. That's the point whatever you think about real life driving or conditions or setups is entirely moot. Assetto corsa isn't real life and the complexity of the setups in this game is not anywhere near as complex as real life.

 

I really recommend to have a read to @BoNI maidire camber setup guide for netkar pro, it's really worth it and explains some other stuff as the hz and %critical damper stuff that can be found using dev app called Car Engineer

Sent using Tapatalk

 

The cancer thing was just an analogy. And since you seem to already know that curing cancer isn't possible you might want to inform all of those phds trying to find a way that their research is pointless.

I know I made an error in the calculation, I corrected it in the edit. When I said the values were conservative I was referring more to the number of options than the time for a lap. I have seen Stefano's lap tester in action, and it takes longer than 5 seconds to lap. Without having access to the source or doing some reverse engineering it is impossible to get an accurate figure for that. Doing the math though if the physics engine updates at 333 hz (I have heard that it does but don't know for sure if that is accurate. It is the only number I have got though so I am using it). 1 gflops of processing power could handle a little over 250000 floating point operations per cycle (assuming a 1 minute lap is sped up to 5 seconds) and I don't think that the physics engine should need that many. I never said it was infinite. I did say that the total ends up taking longer than you will live (which I never said a year is longer than, I said the total for all cars and tracks is). With the corrected figures I put in the edit even if it takes .1 seconds per lap it will take much, much longer than any of us will live. You're the one who brought up infinite processing power.

It may be possible to do at some point in the future, but right now AI has just recently gotten the ability to drive at regular road speeds. The context of that comment is about simulating reality, which implies that it is current. Find one AI capable of making incredibly consistent and fast laps at a race track. Plus there is still the problem that the track conditions will never be exactly the same for 2 laps in the real world.


You still seem to be missing the point of setups. There are certain changes that can make a car as close to objectively faster as possible, such as gearing and aero, but small changes produce such small differences that the difference in lap times with an absolutely perfect lap is pretty much negligible. But those differences can make the car easier to control, and if the car is easier to control a human driver can get closer to perfect on each lap, which means better times. What is easier to control is subjective and up to the driver to decide.