How accurate is Assetto Corsa?

[Bruno Iwanczuk]

If i'm 5 seconds slower than a real life F1 lap world record. Am i really 5 seconds slower?

 

[Bruno Iwanczuk]

Tberg in my view is correct, assuming you just got straight into an f1 car you would almost
certainly be 10 seconds off pace. But if you had a spare £10,000,000 and proceeded though
all the minor and major formula cars you then would be maybe 1 second to somewhere on
the pace but as you are an “paid” driver and not an “ability” driver that may also be a bridge
too far.

unfortunately skill is a free gift so buying it is not an option.

Also if you have the f1 world championship in mind i would like to remind you
that there are approximately 7,500,000,000 people on this planet and normally
only one seat available to win it.

Come to think of it if i were a betting man, my money would definitely be on Tberg

Obviously he is right, and how do you assume i'm thinking in any possibility to be a f1 driver? Just because im comparing my times with F1 driver times?

The point is that he said i would be slower in a real car as if im the only simracer in the world who would be slower.

 

[Trezoitao38]

Well, even for my own cars, only one that’s been released can be classified as “very accurate” (Oreca - and even then, I barely had any aero data for that car). Some private work and some upcoming public work (delays, delays, delays) of mine also fits that category though.

David (RSS, etc.) does very good work, but hasn’t *technically* done any real cars as of late. Jason (aphidgod) also does nice work, though he works more with road cars and therefore doesn’t really need to bother with aero or large tire load variations (two areas where Kunos cars often suffer). As far as Kunos cars...the less possible load variation, the better. Hard to pick out the specifics of which are the best as I don’t have data for everything. But it’s easy to generalize the worst of the bunch: high downforce cars.

And it’s not usually a question of “not bothering” to get them right...they work on a limited timeframe and can’t spend a lot of time on each car. Couple that with some (IMO) judgment flaws in educated guessing and you get inaccuracies.

Edit: It’s also worth noting that lap times are not a particularly good measure of accuracy. The Kunos F1 in the video linked in the post above, for instance, has roughly 75% of the downforce of the real car (and that might even be on the high side, I don’t recall the exact value. A low-end estimate would be about 65%).

So, they put more mechanical grip on car like SF70-H and F2004 because of the things you mentioned?

I understand all your points and I cant comment any of them. But Assetto Corsa seems to me, a ordinary guy that never sit on a F1 cockpit, the best F1 simulator available on the market for ordinary people like us. I didnt drive too much on Rfactor 2, but driving the F2004 on Assetto Corsa or Lotus 98t is a blast experience. The various mods, from you and others are very good too.

So, Assetto Corsa cant handle precisily aerodinamic grip, but seems to me that the other variations are so good that is still a good sim to make F1 cars. Is this the case?

 

[mclarenf1papa]

So, they put more mechanical grip on car like SF70-H and F2004 because of the things you mentioned?

I understand all your points and I cant comment any of them. But Assetto Corsa seems to me, a ordinary guy that never sit on a F1 cockpit, the best F1 simulator available on the market for ordinary people like us. I didnt drive too much on Rfactor 2, but driving the F2004 on Assetto Corsa or Lotus 98t is a blast experience. The various mods, from you and others are very good too.

So, Assetto Corsa cant handle precisily aerodinamic grip, but seems to me that the other variations are so good that is still a good sim to make F1 cars. Is this the case?

First point: sort of, yes.
Second paragraph: Just because it's fun doesn't mean it's (at all) accurate. Kunos LMP1s are pretty fun...they're also rather different than the real cars. Just as lap times aren't necessarily a good indication of accuracy, so is feel (unless something is clearly wrong - the same goes for lap times).

To comment on your last point, no. The Kunos F1 and LMP1 (and GT3 and GT2...etc...especially LMP1) cars are only inaccurate because of the numbers in their physics files. That, however, doesn't mean the physics engine itself is at all bad, in fact:

It's a good sim to make any car...if you do it right. The Kunos tire load sensitivity formula yields too high of grip at low/high loads when its medium loads are set to be correct. Thus, if you use correct downforce, your car will be far too fast in high speed corners (e.g. Eau Rouge with the GT2/3s) and slightly too fast in low speed corners. There's the possibility to use lookup tables for the load sensitivity, which is what I and top physics guys (David, etc) do (and is almost certainly what is done in ACC, though they surely have more for different tire pressures etc.). This allows you to use correct downforce, which means you can use correct spring stiffnesses, which means your car behaves correctly over bumps, etc.

 

[Tberg]

Obviously he is right, and how do you assume i'm thinking in any possibility to be a f1 driver? Just because im comparing my times with F1 driver times?

The point is that he said i would be slower in a real car as if im the only simracer in the world who would be slower.

No, I didn't say you'd be the only one.

 

[RobertR1]

Here's some easy advice to see how fast you are. Find a local go kart track that hosts two stroke racing karts (not the normal stuff you rent). Most will have a test day where you can pay and get an hour session in the proper race karts.

Then apply all you learned in sim racing and see how fast you go. You can then compare your times to the local race weekend running those karts. As a gauge, on a 1min lap time, if you're within 2 seconds, that's great. Move the slider up/down based on that 1min example.

The two stroke karts require you to know how to drive and manage throttle control so it covers most of the elements that you need to go fast. Plus they can pull a couple of G's which will test your commitment.

 

[Ghoults]

Overall ac is on the easier side. Slides are pretty easy to catch and powerslides especially are very forgiving. Ac cars are also extremely forgiving going over bumps and sharp kerbs. At nordschleife you can accidentally drive over all the kerbs and it doesn't really unsettle the car. You lose time but not the control of the car. No secondary bounces unless the car is set on purpose to be underdamped. Not all cars suffer from this. Some cars are very good (I love the ad3r's jota diablo and the ktm has some nice bite) but in too many cars when you get into oversteer moment your only worry is the momentary understeer when the rear grips.

I've been making my own fictional track and I was really surprised how massively huge you need to make the kerbs before you get into any kind of problems. Even accelerating at 150-200kmh the direction of the kerb sawtooth doesn't make the car handle any differently. Apex kerbs less tall than 15cm tall you can drive right over. Some apex kerbs I had to make 20cm tall to make it not worthwhile drive over them.

 

[Bruno Iwanczuk]

No, I didn't say you'd be the only one.

Sometimes people don't need to say it directly to communicate something. If it wasn't the case, my bad.

The only thing that would make a simracer slower in a real car are the g forces and maybe in some cases the fear factor. The last one depends on how much you care about life. Technically is much easier to drive in real life because you can feel the grip much better.

 

[Mascot]

Don't forget skill and ability.

 

[Bruno Iwanczuk]

Don't forget skill and ability.

But what i mean is, if you're good in a simulator, you WILL be good IRL, as long as you adapt your body to g forces and not be afraid to crash.

 

[Mr Deap]

It depend how you see the driving model as a whole.

The fascinating part is probably the simulation value of the placebo & how it translate to the real thing.

Being 6sec off pace of WR or being WR is more or less the same.

 

[Kek700]

Your not that aware of g forces and extra cornering speed is factor of the car.

As previously explained in karting ( RobertR1 ) which i did a lot of. The G is noticeable
as described, but i cannot remember it effecting any thing that i did, it was just a function
of higher cornering speeds. But i do remember the bashing about and consequently
being quite badly bruised after karting.

I can only say it from my point, driving a 1980’s tunnel effect F3 car doing sprints; then 30 minuits later repeating it again in a Peugeot rallye 1.3 road car. I never remember the g forces, just being
aware of the extra capability of the F3 car and then jumping into a road car with road tyres
and adapting immediately. I can assure i have a remarkable ability to be mediocre in all things.
In fact i am quite pround of my mediocrity.

I obviously cannot speak about F1, but first chance i get i will talk to my neighbor about this
as he practices on AC and won LMP2 at Le-mans in 2006.
I have spoken to him before about this topic but cannot exactly remember the conclusion.
( never thought i would need to remember it )

 

[Amir Vodokotlic]

If you have real life telemetry and use real life data, you can get accurate results with AC.

That being said you're talking about F1/LMP1/GT3 etc where correct data is never going to be given to them no matter what game we are talking about AC/rf2/iRacing etc etc...

Car like SF70H in terms of aero is probably done on what dev feels is right. (which on aero is on low side) if you read racecar engineering magazine or sae paper.

That being said they are crunched for time (had one guy doing physics for 150+ cars) and do the best they can.

 

[mclarenf1papa]

If you have real life telemetry and use real life data, you can get accurate results with AC.

That being said you're talking about F1/LMP1/GT3 etc where correct data is never going to be given to them no matter what game we are talking about AC/rf2/iRacing etc etc...

Car like SF70H in terms of aero is probably done on what dev feels is right. (which on aero is on low side) if you read racecar engineering magazine or sae paper.

That being said they are crunched for time (had one guy doing physics for 150+ cars) and do the best they can.

I would maybe...reconsider the last group of cars you mention in that statement. And really, to a lesser degree, all of them. Pretty good data can be found or calculated for F1 and LMP1 cars. There's no excuse for the Kunos LMP1s to be 80+ hp down on power when the regulations provide a maximum fuel flow and there's thermal efficiency data online for the Toyota TS030 and TS040.

RSS has managed better accuracy with no manufacturer connections than Kunos did with a license and some data. The 2017 Mercedes F1 car, near the end of a straight, in an unknown aero config, had an ScZ (CLA) of 5.4 and an ScX of (edit)1.55. The Kunos Ferrari isn't even close to that. The aero sensitivity on the LMP1s is way less than real life, their power levels are way off, inertia is off, tires are off, etc.

And then there are the little things. There's a parameter in the tire files that controls the rolling resistance gain with slip. Kunos cars use far too high values for this number and its easy to see in any fast/near flat/flat corner where the tires are still working (you can hear it in the engine sound as much as you can see it in the telemetry). And as @Ghoults mentioned, race cars in AC are pretty forgiving at high slip. That's also just a different parameter. Stuff like that adds up.

 

[Kek700]

Interesting reading the above. You know AC is not absolutely correct when driving.
But you accept that it is a simulator with limits. Funny enough there are things in real life
driving that i am glad that they are missing.

 

[garyjpaterson]

But you accept that it is a simulator with limits.

Sure but all of the things listed above have nothing to do with limitations of the simulation.

 

[Kek700]

True. But they are still limitation of accuracy and there are many more varibles that have their
own limitations.

 

[Amir Vodokotlic]

I would maybe...reconsider the last group of cars you mention in that statement. And really, to a lesser degree, all of them. Pretty good data can be found or calculated for F1 and LMP1 cars. There's no excuse for the Kunos LMP1s to be 80+ hp down on power when the regulations provide a maximum fuel flow and there's thermal efficiency data online for the Toyota TS030 and TS040.

RSS has managed better accuracy with no manufacturer connections than Kunos did with a license and some data. The 2017 Mercedes F1 car, near the end of a straight, in an unknown aero config, had an ScZ (CLA) of 5.4 and an ScX of 1.35. The Kunos Ferrari isn't even close to that. The aero sensitivity on the LMP1s is way less than real life, their power levels are way off, inertia is off, tires are off, etc.

That's very impressive with that little drag. Considering their frontal is lot bigger now.
How accurate is that data?

Its not like that kind of info is plastered all over the internet. That kunos dev could have found.

(shouldn't have mentioned LMP1. don't follow it)

 

[mclarenf1papa]

That's very impressive with that little drag. Considering their frontal is lot bigger now.
How accurate is that data?

Its not like that kind of info is plastered all over the internet. That kunos dev could have found.

(shouldn't have mentioned LMP1. don't follow it)

It was found by Gary on the internet... It’s literally a screenshot of telemetry data. Same deal as when Acura accidentally leaked Windshear data for the ARX-05 in its promo video. It’s better to spend a week on research and a day on the physics model than a day on research and a week on the model.

Basically everything I said about the LMP1s (besides the power) is also true for the F1s. What I was getting at with the “reconsider the last group of cars” is that Kunos is currently making a GT3 game with extensive data. GT3 is generally one of the easier categories to find/get data for.

 

[The Zman]

I was just driving last Friday at the Porsche Experience Center Atlanta and they use Assetto Corsa in their simulators. Naturally driving Porsche's and using Laguna Seca when I was there.
I think that says it all for accuracy.

 

[Amir Vodokotlic]

It was found by Gary on the internet... It’s literally a screenshot of telemetry data. Same deal as when Acura accidentally leaked Windshear data for the ARX-05 in its promo video. It’s better to spend a week on research and a day on the physics model than a day on research and a week on the model.

That's crazy. 5.4 for open wheel car is something!

I know there is leak of all Mercedes engine modes from this year spa.

But still from race car engineering their cfd studies were high 3.8-3.9 range.

even f1technical estimate is high 3.

https://www.f1technical.net/features/21667?sid=5bf9d342014c4fb44104e8cbd3a5d39b

@garyjpaterson mind sharing link?