PC Sim Racing

[Bigbazz]

Most advanced physics is pretty simple when you take into account all of the different physical occurances it simulates. Things such as tyre deformations and flatspotting for example are something that no other tyre model has. Like I said in my post the most advanced physics (which Rfactor 2 currently is) doesn't always relate to the best driving experience because it is related to how well those physics are utilised. Though in many cases I would say they have achieved that.

As for revamping, SRW has done a lot more to modify the base engine than Reiza. Talking about Reiza and SRW they are running on a dated version of ISI's engine (aka gmotor 2), it stands to reason that the new heavily updated engine with many new features and aspects being simulated is by default more advanced than either of those titles for they are limited by the very dated engine. The same goes for NetKar, Kunos himself states how dated and limited in certain areas the engine is, though Assetto Corsa is far more advanced and capable/accurate in these areas. Yet there are still people who prefer NetKar to Assetto Corsa. There are detailed descriptions and explanations about the advances of the Rfactor 2 engine available online, go take a read for yourself rather than take my word for it.



The work is ongoing in SRW, GSC, iRacing, Assetto Corsa too, Rfactor 2 is not alone in this, but it is already at a point though where some cars are beyond the reach of what any of the gmotor2 engine games can achieve, simply because they are limited by certain factors where Rfactor 2 is not.

 

[guybo]

90% of what is more advanced about rF2 over other sims is in the tires. It's all in where the rubber meets the road. Sidewall deformation, tread flex, flat spotting (this is a basic part of racing not an "extra" LOL) the way that the tires wear, etc. The updated graphics are OK, and the sound of the ISI cars is very accurate- but the way that the tires interface with the road surface is what sets rF2 apart.

 

[Bigbazz]

90% of what is more advanced about rF2 over other sims is in the tires. It's all in where the rubber meets the road. Sidewall deformation, tread flex, flat spotting (this is a basic part of racing not an "extra" LOL) the way that the tires wear, etc. The updated graphics are OK, and the sound of the ISI cars is very accurate- but the way that the tires interface with the road surface is what sets rF2 apart.

I think flatspotting is an extra, its something that happens in an unfortunate situation, just like having a damage model is only really relevant when you're crashing into something. The normal driving experience of the car for 99% of the time does not have a lot to do with flatspotting tyres. If you have the situation where you have the tyres that do everything real tyres do but the car drives nothing like a real car then that flatspotting means nothing in the grand schemes, on the other hand if you have a car that drives and reacts very realistically then you can easily forgive the fact that flatspotting is not modelled, in best case scenario you will never deal with flatspotting in any case.

You get my point?




This is why some people still prefer GSC to Rfactor 2, because some of their car models simply drive more realistically or provide more satisfaction to that use, regardless of the fact that the tyre model or whole physics engine is less advanced it can still provide the better overall experience, even better overall realism.

 

[guybo]

LOL, no I don't. Nevermind, you don't understand. It's OK- agree to disagree.

 

[MikeSaints]

Most advanced physics is pretty simple when you take into account all of the different physical occurances it simulates. Things such as tyre deformations and flatspotting for example are something that no other tyre model has. Like I said in my post the most advanced physics (which Rfactor 2 currently is) doesn't always relate to the best driving experience because it is related to how well those physics are utilised. Though in many cases I would say they have achieved that.

No, wrong.

As I said earlier, fans get all excited about "new features" just because they're new (and obviously fuel better the tirades "my sim is better than yours because it has this and that and that").

I am sure most people have a very basic understanding of what those things are and in what way they impact tire performance.

Several Semi-empiric tire models deal with flex, different modes of vibration, heat, pressure, velocity and accurate enough to be taken seriously.

Nothing done thus far to rF2, CARS, AC or iRacing surpasses what has been done by the tire industry and testing facilities for years. It is, however, overly hyped up.

Back to basics: talk to tire engineers and racing engineers so that you understand what the "dated" version of ISIMotor2 has accomplished.

As for revamping, SRW has done a lot more to modify the base engine than Reiza. Talking about Reiza and SRW they are running on a dated version of ISI's engine (aka gmotor 2), it stands to reason that the new heavily updated engine with many new features and aspects being simulated is by default more advanced than either of those titles for they are limited by the very dated engine.

No, it doesn't stand to reason. You are extrapolating what devs are telling you and taking conclusions without data to do so.

You can throw in a million features (flatspotting, visual deformation, carcass effects, wear, degradation) and you still won't inevitably have as a result a more "advanced" anything. The results speak for themselves, accuracy of these models will tell you (if you're willing to dig in instead of just saying "Amen"). Compare the output of the sim or its telemetry readings to RL readings to readings from the dated "ISIMotor2" - then come back and report.

As for "dated"...

You still don't understand (or didn't read the whole post).

ISIMotor2 has several versions. The last version available for rFactor was 1.255F (not counting the latest C).

Reiza, not SRW, picked a 2009 version.

Then, something you obviously didn't want to comment on, a lot of work was done by Reiza, just as SIMBIN did for race07.

ISIMotor2 is a game engine, a dev platform. Dev studios have used it and some (SIMBIN and Reiza) improved it markedly, corrected a few things, even added a feature or two. Do ask Reiza and SIMBIN what they have done so that you understand the process.

As for SRW: the italian professional racing driver serving as a vehicle dynamicist said that everything is in the physics files (well...we do know that since before 2005). So, what have they done with the physics files? What sort of physics calibration have they done?

I took a look at it some 4 months ago and I was totally dismayed by it. Several errors (misspelled variables, wrong values, missing "=", etc) and a terribly done physics calibration meant most cars performed distantly from their real counterparts. I sent emails to and chatted with the prtogrammers and support guys, explained what was wrong, they said they were looking into it...but nothing.

There's no comparison between Reiza's work (or SIMBIN's) with SRW. One is professional, the other sadly sloppy.

The same goes for NetKar, Kunos himself states how dated and limited in certain areas the engine is, though Assetto Corsa is far more advanced and capable/accurate in these areas. Yet there are still people who prefer NetKar to Assetto Corsa. There are detailed descriptions and explanations about the advances of the Rfactor 2 engine available online, go take a read for yourself rather than take my word for it.

What he, as a programmer, would like to do, and what tire modelling is, are two totally different things. Kunos knows this. For you to actually understand this you have to be into and follow the discussions at the gamedev forums (Kunos still writes posts there, explaining his vision and what the physics engine is supposed to achieve).

As for "detailed descriptions and explanations" of rF2: I read them, mate. I know the model Gjon is basing is work on. And I know very well that neither Gjon nor Kunos have enough official, reliable tire data to make it the final model accurate and reliable.

Probe around, ask the right people the right questions and you will understand.

The work is ongoing in SRW, GSC, iRacing, Assetto Corsa too, Rfactor 2 is not alone in this, but it is already at a point though where some cars are beyond the reach of what any of the gmotor2 engine games can achieve, simply because they are limited by certain factors where Rfactor 2 is not.

Wrong again.

More features and what a physics engine can reach are two distinct things. The one area where ISIMotor2.5 will no doubt surpass ISIMotor2 is in suspensions - more suspensions are modelled. The same thing applies to CARS, for instance (which has a different game engine).

It's more of a quantity thing and surgical interventions to the engine (Gjon explained this, go look fior yourself) making it a bit more accurate.

With the exception of suspensions (ISIMotor2 allows for several types, but not all), I can model any car and present this to professional racing teams and have them compare their own RL telemetry and my cars telemetry - mostly spot on. I can (so do Niels and Bristow) capture any car's performance envelope and simulate it accurately enough with ISIMotor2. To my experience (and from what I know of Niels's), no cars "are beyond the reach of" ISIMotor2. But you believe the myths you wish.

By the way, your use of "gmotor2" tells me you are not knowledgeable enough of what these engines are (from ISIMotor2, to the old Madness Engine, to Lizard to ISIMotor2.5). gMotor2 is the graphical part of the engine. gMotor2 is not meant to do any physics (check ISI).

If you want physics, talk about pMotor2. If you want the whole engine, talk about ISIMotor2.

This is why some people still prefer GSC to Rfactor 2, because some of their car models simply drive more realistically or provide more satisfaction to that use, regardless of the fact that the tyre model or whole physics engine is less advanced it can still provide the better overall experience, even better overall realism.

Yes and you don't get why.

Physical Realism and/or accuracy have nothing to do with a boatload of features and sophistication.

people are waking up to the fact that GSC is truly astonishing because the physics engine underneath is being well/properly used. You don't need a mountain of features (which satisfy the eye-candy addicts), you just need something that accurately represents behaviour of cars and tires.

The one area where ISIMotor2 is lacking is in the graphics area, and still SRPL proves that a lot can be done via the use of complex shaders. Reiza is very much aware of this and lo and behold their shaders reflect a lot of work.

But for the rest, Reiza proves ISIMotor2 is really fantastic.

 

[Bigbazz]

LOL, no I don't. Nevermind, you don't understand. It's OK- agree to disagree.

It isn't anything to do with agreeing to disagreeing, the post has seemingly gone straight over your head completely.

No, wrong.

So what you're saying is that ISI's old engine is so awesome that it's better than their newer, more advanced updated version. And that their new version is not more advanced because the old version is awesome? Have I got that right? A sort of "it's not more advanced because none of that means anything and ISI2 is the best", never mind that we're talking about the same developer here for both engines.

Race 07 feels dated and it feels bland it is innaccurate and lacks detail in the FFB, Race 07 is arguably the least accurate proper "sim" out there, and GTR2 which again is by todays standards one of the lesser accurate sims out there , even SMS (the guys who developed GTR2) speak of the innaccuracies and the limitations of the engine, neither of these games are even in the same league as either Simraceway or GameStockCar in terms of accurate physics or detailed FFB. Then you have SimRaceway, guys who have completely overhauled the engine, turning it into their own game with many physics updates, but its still limited, Reiza GSC that I've heard as being called a "pay mod" again have done great things with it but it is still limited by the gmotor 2 engine, with a lot less under the hood changes than Simraceway have done.


And then there is Rfactor 2, designed by ISI, the people who made the Gmotor 2 engine that runs on all these games. The people who have updated the engine so that it simulates more and more accurately, it could not possibly result in anything less than what Gmotor 2 delivered because it is made by the same people trying to improve on and remove the limitations of the old system, it is the latest version of the ISI engine.

Why you're so defiantly defending the old Gmotor 2 compared to the newer engines that are simply more advanced and more accurate I have no idea. I've never heard the Gmotor 2 being defended so valiantly before in my life because it isn't generally thought that it is such a great physics engine. It is definitely good, and solid but people have been asking for more, for better for years.

Finally you fail to understand my use of the word "advanced" and that is the issue here. The Rfactor 2 engine (made by ISI!) simulates more than the old gmotor 2 does, you can't dispute this because it is a fact, you can argue all you want but by advanced I am talking about how far the engine goes into simulating different aspects and the Rfactor 2 engine simply takes it further, there is no arguing about that.




Lastly, I think that the Gmotor 2 does produce a great experience, I rate both Simraceway and GameStockCar about equal in physics, they have both done a great job in showing what can really be done with a good sim and with the Gmotor 2 engine. Sadly I still think the best cars at Rfactor 2 (370z for a start) outclasses both games, and based on the Assetto Corsa tech preview I'm inclined to believe the full product will also surpass both.

 

[MikeSaints]

So what you're saying is that ISI's old engine is so awesome that it's better than their newer, more advanced updated version. And that their new version is not more advanced because the old version is awesome?

You got it wrong, and deliberately so (perhaps trying to be funny).

ISIMotor2 is not perfect, but it does the job right. It could be better? Yep. But as it is, it is just right and does the right things well.

Next gen physics engines may have a boatload of features, but their sophistication says ZERO about their physical realism.

You got it wrong, you didn't even bother to read what I wrote and deliberately ignored valuable information that might help you understand what physics engine development is about and in particular what tire modelling entails.

It is clear debating this further with you is simply running around in circles, so for the last time...

Have I got that right? A sort of "it's not more advanced because none of that means anything and ISI2 is the best", never mind that we're talking about the same developer here for both engines.

I say again, you are badly, badly misinformed about ISI, ISIMotor2 (your "gmotor2") and are distorting what I said.

Leave it at that.

Race 07 feels dated and it feels bland it is innaccurate and lacks detail in the FFB, Race 07 is arguably the least accurate proper "sim" out there, and GTR2 which again is by todays standards one of the lesser accurate sims out there , even SMS (the guys who developed GTR2) speak of the innaccuracies and the limitations of the engine, neither of these games are even in the same league as either Simraceway or GameStockCar in terms of accurate physics or detailed FFB.

This is the internet and people believe they can say whatever they think.

You are misinformed.

GTR2 is based upon an earlier version of ISIMotor2 - around v1.07.

Race07 is based on a latter version and benefited from a significant overhaul of tire physics (done in-house). Mark Reynolds, head of AI and physics at SIMBIN at the time, explained this thoroughly. Do realize that Blimey! never commented on race07 being bad or having problems - which speaks volumes.

Race07, sir, is still being used by several professional racing drivers (some of which former or current F1 drivers). These drivers and investor groups have set a number of private leagues. As I worked for one of these groups, I can confirm that.

Race07 is accurate, though not perfect. Same thing with rFactor or even NetKar Pro (which, with a different physics engine and in spite of its dynamic features, still has uses a mathematical based model for its tire physics, just as ISIMotor2).

As for GTR2: you know perfectly well (if you don't, then you are thoroughly and irrevocably misinformed) that SMS promoted the first Shift "sim" at the expense of GTR2. Ian Bell and SIMBIN had recently parted ways, things got pretty heated between the two companies (actually, 10tacle was deemed insolvent in 2008, and a year later Blimey! went down with it, leaving the business ties with SIMBIN behind them).

Still, from an insiders perspective, GTR2 had some issues with grip estimation which Blimey! were unable to solve. SIMBIN did it though with Race07.

As I said, you are thoroughly misinformed.

From a racing engineer's perspective (and I have worked or co-operated with several), rFactor v1.255f, Race07 and GSC 2012 do the job right. Can these sims be regarded as engineering tools? No. But they're the closest thing to it.

As for FFB: I see you reflect the usual misinformed notion in regards to FFB. FFB does not necessarily reflect a sim's physics. FFB (have you ever read the FFB code of any simulation?) is, shall we say, a layer on top of the physics engine. You can have a good physics engine and a poorly done FFB code. Kunos himself has stated NKP'S FFB code was surprisingly simple and short (as far as I know, under 100 lines of code).

Still, Race07 is regarded as having a rich FFB - perhaps as good as NKP's (though that is a matter of debate).

By the way, regarding your "even SMS (the guys who developed GTR2) ":

wrong again. SMS didn't code GTR2. SMS bought the assets from 10tacle and Blimey!. The core team that worked on GTR2 was largely Blimey! and part SIMBIN. The core team of GTR2 is not the same as the one SMS now have with C.A.R.S., nor was it the same when they developed Shift (which, again, they promoted at the expense of GTR2).

Inform yourself first.

Then you have SimRaceway, guys who have completely overhauled the engine, turning it into their own game with many physics updates, but its still limited, Reiza GSC that I've heard as being called a "pay mod" again have done great things with it but it is still limited by the gmotor 2 engine, with a lot less under the hood changes than Simraceway have done.

SRW completely overhauled the engine? Total news to me. Never, ever, did the programmers and support team admitted that to me and others, but here you are proclaiming that...Wow. And what engine? The game engine? The physics engine? The graphics engine? Which engine?

GSC, as I said, is not a "pay mod". They purchased a license to develop a sim based on ISIMotor2. Again, you are ignoring what I told you: they built GSC and GSC2012 (not sure about 2013) upon a newer version of ISIMotor2.

And regarding the limitations: you refuse to be informed about these matters. So be it.

And then there is Rfactor 2, designed by ISI, the people who made the Gmotor 2 engine that runs on all these games. The people who have updated the engine so that it simulates more and more accurately, it could not possibly result in anything less than what Gmotor 2 delivered because it is made by the same people trying to improve on and remove the limitations of the old system, it is the latest version of the ISI engine.

Wrong. ISIMotor2 runs on all these games. gMotor2 is the graphics component.

Limitations? You don't know what you are talking about. You blurt out things from hearsay and press releases. I am informing you from experience and factual knowledge of the platform at various levels (from code, to physics modelling, to physics calibration and motion simulation tie-ins).

compared to the newer engines that are simply more advanced and more accurate I have no idea. I've never heard the Gmotor 2 being defended so valiantly before in my life because it isn't generally thought that it is such a great physics engine. It is definitely good, and solid but people have been asking for more, for better for years.

Well, it is a great physics engine. Period.

SMS had their reasons ($$$) for belittling it.

Kunos has always preferred NR2003 and later iRacing to ISIMotor2, comes as no surprise his opinion on ISIMotor2. The one opinion that actually counts for many of us is that from real life pro racing drivers and racing engineers who hold rFactor and Race07 as "realistic".

Is it perfect? No. Reiza and SIMBIN proved beyond doubt that a lot can still be done to improve it further than ISI ever did.

Does it favour modding? Yes. Much easier to mod with ISIMotor2 (regardless of being rFactor, Race07 or GSC2012) tha it is with ISIMotor2.5.

Does it benefit from more tire data than physical based models (such as the ones coded by AJ in CARS and Dave Kaemmer for iRacing)? No doubt. Each set of pacejka based data is expensive but developers can benefit from having reliable (with the noise removed) data from acceptably reliable tire tests (not all tests commissioned target a large range of temperature, humidity, wear/degradation, but some very expensive tests provide important and extensive data). Ask Kazunori what he thinks about this, btw.

You believe whatever myths and press releases you want.

Finally you fail to understand my use of the word "advanced" and that is the issue here. The Rfactor 2 engine (made by ISI!) simulates more than the old gmotor 2 does, you can't dispute this because it is a fact, you can argue all you want but by advanced I am talking about how far the engine goes into simulating different aspects and the Rfactor 2 engine simply takes it further, there is no arguing about that.

What you fail to understand (and you haven't not even once mentioned some of things meant to help you get a clear picture) is this: a big pack of novel features means nothing in terms of physical realism, which is what really counts.

Your "simulates more than the old" is a fallacy. Tell you something: I presented all the info provided by Kunos, Gjon and DK to a former Williams engineer (and currently involved with tire testing companies and ELM series teams) and his response was the same as I got recently from a former F3000 engineer friend of mine: "what are the assumptions for all these different aspects of tire modelling"? When I explained, he wasn't convinced. As I and others aren't also.

Simulates more of what? In what way? Based on what data? You should not simply accept that they're simulating this and that new feature; you should look into it and see if, with the data available, those aspects/features can actually be simulated without a multi-million dollar hardware. If you look into this more deeply you'll understand the more complexity you add, the potential for less reliability and stability rises significantly; finally, how do you know if those new features/aspects are being properly represented? Take flatspotting, for instance, do you even know how many more sampling points LFS requires for flatspotting than NetKar Pro? Do you even know the sampling points threshold for qualifying flatspotting as significant or realistic?

I understand people's cheerleading of this or that dev studio and sim, but there should come a time when people sit and sift through all the hype, and really really understand what is happening.

 

[Bigbazz]

You're saying I'm deliberately ignoring what you're saying. While you're deliberately fobbing off Rfactor 2's without even giving thought to what I'm saying. You're religiously defending the old version of the ISI engine while completely facepalming the new one. For what reason? Based on what?

Saying "my mate isn't convinced" does not qualify you to completely ignore the Rfactor 2 engine. If you look back on the history of Rfactor 2, you will see that it started off with a lot of people not convinced, myself one of those. As they implemented improvements and refinements things started to fall into place.

I'l say that I don't think Race 07 is very good, I don't care if Rubens Barrichello has a laugh on the weekends racing his buddies on there, that doesn't make it accurate just because Rubens enjoys playing it. Amazingly I can like millions around the world walk outside and drive a real car, so it doesn't take a lot for me to be able have a comparison. Race 07 feels bland to me, it is missing so many of the real characteristics, the fine details in those characteristics of driving a car.. Comparing it to NetKar Pro, Assetto Corsa, Rfactor 2, GameStockCar or Simraceway is just a joke, it isn't even in the same league as any of those games, and no "accurate results from tyre temp testing" or similar is going to change my opinion on that when I have a car outside and a track within 15miles of my house.

I fail to see what it is you miss about Rfactor 2, have you tried it? I didn't form my opinion based on marketting, hell if I did then I'd be the first to tell you how amazing iRacing is. My opinion on Rfactor 2 is based on real world comparisons, it's not based on tyre temp testing, and it's not formed based on the opinion of my mate the engineer who read about it but hasn't tried it.

I don't cheerlead a dev or sim, but rather I like all of them. But I have an opinion of which I think is the best for certain cars, whether right or wrong. NetKar Pro was a very simple code but turned out very very good, but Kunos himself also stated that this created limitations when trying to make Ferrari Virtual Academy accurate, the result was that the 458 felt overly understeery and did not match expectations.




Finally, you miss the point of the importance of extras like flatspotting. These are not simulations made to develop cars, these are simulations made as games, and a simulator that can implement flatspotting is simply more realistic a racing simulator than one that doesn't. If I go race a car at a track and I have a massive lockup I will flatspot a tyre, if the simulator does not simulate this (just as gmotor2 or ISI2 as you correct me) then it is lacking something.

The important thing is that these are games, if 2 games have equally good physics an one is on the old ISI engine, one on the new (Rfactor 2) engine, then the Rfactor 2 is already by default more realistic. because of the inclusion of flatspotting as part of the racing experience.

 

[MikeSaints]

While you're deliberately fobbing off Rfactor 2's without even giving thought to what I'm saying. You're religiously defending the old version of the ISI engine while completely facepalming the new one. For what reason? Based on what?

I'l say that I don't think Race 07 is very good, I don't care if Rubens Barrichello has a laugh on the weekends

Talk is over.

Believe what you want, it is sad though that people like you, with little or no factual knowledge of these matters, spread their ill-informed opinion and a set of pseudo-facts.

 

[Bigbazz]

Talk is over.

Believe what you want, it is sad though that people like you, with little or no factual knowledge of these matters, spread their ill-informed opinion and a set of pseudo-facts.

And you're spreading your opinion based on ignorance, you have an almost religious defence over the old ISI engine while near enough refusing to acknowledge the existance of the new one, or any of the advances/improvements it has made. Infact from your posts it indicates that you have no hands on experience with Rfactor 2?

You using the fact that real drivers use one game or another for entertainment as if it is some kind of confirmation of a sims accuracy is just laughable. How many professional racing drivers use iRacing? A lot more than any other simulation game, does that make iRacing the most accurate sim? I don't think so, and neither do a great many other people.


None of these games even come close to what a professional simulator that racing teams use can do. That is the difference, these are games made for entertainment purposes.

 

[CC570]

I'll be the first to admit I have no knowledge of how any race/driving sim actually goes about the task of simulating a car. I would guess that there is more than one fundamental approach, much like the difference in flight sims eg. X-plane and FSX. One can argue, and many do, that one approach is inherently more accurate or realistic than the other or vice-versa. That may be what is going on in this thread. One game engine may work on one level actually calculating the simulation in real time based on physical parameters provided by the model and physical principals. Another might rely more on extensive lookup tables that are pre-calculated by the modeler. Both can give good results and bad depending on how they are utilized by the modder or developer. Again I have no knowledge of how any race/driving sim actually works. I do feel that defining what is meant by "realistic" in the context of simulators and simulations is neither simple nor universally agree to.
Anyway there has been some great posts and information in this thread. If anyone wishes to continue discussing these ideas PM me.

 

[MikeSaints]

I'll be the first to admit I have no knowledge of how any race/driving sim actually goes about the task of simulating a car. I would guess that there is more than one fundamental approach, much like the difference in flight sims eg. X-plane and FSX. One can argue, and many do, that one approach is inherently more accurate or realistic than the other or vice-versa.

Interesting post, CC570.

However, I have been arguing this point over and over at RD and NG for years now: there is not one single way to achieve physical realism. Not because all chassis/suspension/aero/tire models are incomplete or flawed, but because there's more than one or two ways to express the same system. In particle physics we have different models studying and expressing a system, and the same thing happens at a macroscopic level (cosmology). With car/tire physics we are also subject to the same rules - with a nuance here, and that is the many limitations forced upon us by programming languages (and machine language) and the complex paradigms behind the current software development environment.

But...

And this I tried to make that poster understand: while it is, from a programmatic/SW development POV, very exciting to simply add new features to existing models, it does not (and never will) necessarily follow as consequence that these "new models" are more comprehensive and realistic than old ones. Not now not ever.

You say the "there is more than one approach" is what is going on in this latter part of the thread; it is not, it is simply a matter of, from the vantage point of experience and de-facto knowledge, explaining to someone who relies on press releases and forum posts what these models entail.

There is more than one approach. The best approach is, obviously, that which brings in better results. Better results is not weighed upon "feel" and subjective opinions of fanboys; the better results I am talking here is verifiable via a host of benchmarks, from telemetry, to simple readings, to simply evaluating things we can measure easily [like the max speed at a certain corner and the consequent g-meter readings, the max compression of suspensions under a certain load (i.e., weight of the car plus/minus the downforce/lift].

Another might rely more on extensive lookup tables that are pre-calculated by the modeler. Both can give good results and bad depending on how they are utilized by the modder or developer.

Well, ISIMotor2.5 (rF2 for that poster...) uses a set of concepts/assumptions inspired by a 2005 model originated by TNO Automotive. It marries some of the empirical characteristics of Pacejka based models (such as slip and grip estimation) with physical quantities/facets of the tire typical of physical models. At first glance, it is rather close to SWIFT, though Gjon will say it is a bit removed from that (I don't think so).

So, it uses dynamic tables which must be generated by the physics calibration development tools for a certain array of tires and situations.

Stefano (Kunos) explained somewhat how NetKar Pro did more or less the same thing - though now he claims that AC, tire wise, is much more advanced than NKP (the source code is his, so he should know better than anyone else).

The thing about sim physics is that it is highly dependent on the proper calibration. For Discovery Science, iirc, McLaren Systems explained this dependency very clearly in how they set up their vehicle simulation for the F1 car and for the MP412c road-going supercar (and its GT3 version).

So, effectively, you can have a great physics engine and due to poor calibration models (the data you feed the engine: for the tires, the wings, the chassis, the masses, the suspensions) you will have a bad simulation.

There is more to all this than most people think.

Again I have no knowledge of how any race/driving sim actually works. I do feel that defining what is meant by "realistic" in the context of simulators and simulations is neither simple nor universally agree to.

I completely disagree.

If by realism you mean physical realism, we have several tools to verify the validity of a physics engine output.

If by realism you mean the opinions of people about how a car handle or should handle...that's not realism, but simply their flawed, misconception-filled opinions at stake.

Simple example: the president of Nissan (Carlos Ghosn) first tested the Nissan GT-R 35 a few years ago and the engineers explained how he should expect 2g cornering at some points on the track.

A racing/vehicle sim representing that car will have to show the simulated car cornering at the exact same lateral accelerations. If it fails to do so (I remember NFS road cars cornering at speeds which implied 4 to 5g), could we accept that as "realistic"? No. So, it's not that hard to check the validity of vehicle physics.

 

[Bigbazz]

You want to talk Data, the Mclaren MP4-12c in project cars was shown to produce cornering G-force that matched the real car, aswell as tyre temperatures that matched the real world data accross the whole tyre. In terms of cornering G the Mclaren F1 also produced those accurate results. So then does that make it realistic?

Does it make it realistic when iRacing gets these same values right but the real life racing drivers of these cars say that the car does not drive/respond properly? You talk with such detail without explaining why you have come to your conclusions. You talk about Race 07 being realistic, but then why do the cars not react in the right manner, why does the FFB lack detail? Why do the cars often have too much grip? Why does it not even remotely compare to other sims when comparing the driving characteristics and wheel feel I get in real life to the sim? Assetto Corsa has the most "road car" like feel in the wheel of all the sims, none of the ISI 2 games do this, not a single one of them actually feels like driving a road car through the tyres, nor did NetKar Pro, though it was very detailed and precise, incredibly direct and satisfying, the Fiat Abarth did not feel like a Fiat Abarth.


People drive a car in real life and if it drives and reacts like that in a game then they deem it realistic to drive, you take that further and apply it to cornering speeds/g-force, braking distances, tyre temperatures and tyre wear and it is a step further. The problem is that if you have the second part without the first part it means nothing to the end user. iRacing used this exact approach "ahh well the numbers are right" in regards to the LMP2 car, and an actual LMP2 driver who drives the car replies "It is nothing like the real car" making a big discussion pop up on their forums.

Just because the numbers are right it does not mean the simulation is realistic.

 

[AnAL0G]

No simulation is realistic without actual g-forces, so I don't understand the point of investing so much energy into an argument that misses the forest for the trees from the get-go.

 

[MikeSaints]

No simulation is realistic without actual g-forces, so I don't understand the point of investing so much energy into an argument that misses the forest for the trees from the get-go.

By that (flawed) point of view, no simulation ever will. I have been aboard 747-400 and A320 simulators and was part of a certain european airforce combat flight sim project team and I never experienced nor know of any g-forces simulated accurately with these sims (not in magnitude, nor in directionality).

In order for fully represent the performance envelope of an F1 car, a simulator and its motion simulator platform would have to develop 4g-level longitudinal forces and up to 5g lateral accelerations. We have no equipment, no technology capable of simulating forces and inertial forces of this magnitude (other than equip a centrifuge pod with a simulator, and even that would require a complex system to instantly convey the forces and instantly augment or decrease them). That doesn't mean we have no standards by which to judge the accuracy of a sim - professional or commercial.

So, in light of that, your "trees" mean little or nothing.

 

[CC570]

Interesting post, CC570.

However, I have been arguing this point over and over at RD and NG for years now: there is not one single way to achieve physical realism. Not because all chassis/suspension/aero/tire models are incomplete or flawed, but because there's more than one or two ways to express the same system. In particle physics we have different models studying and expressing a system, and the same thing happens at a macroscopic level (cosmology). With car/tire physics we are also subject to the same rules - with a nuance here, and that is the many limitations forced upon us by programming languages (and machine language) and the complex paradigms behind the current software development environment.

But...

And this I tried to make that poster understand: while it is, from a programmatic/SW development POV, very exciting to simply add new features to existing models, it does not (and never will) necessarily follow as consequence that these "new models" are more comprehensive and realistic than old ones. Not now not ever.

I agree completely.

You say the "there is more than one approach" is what is going on in this latter part of the thread; it is not, it is simply a matter of, from the vantage point of experience and de-facto knowledge, explaining to someone who relies on press releases and forum posts what these models entail.

I said " that may be what is going on " it may very well not be.

There is more than one approach. The best approach is, obviously, that which brings in better results. Better results is not weighed upon "feel" and subjective opinions of fanboys; the better results I am talking here is verifiable via a host of benchmarks, from telemetry, to simple readings, to simply evaluating things we can measure easily [like the max speed at a certain corner and the consequent g-meter readings, the max compression of suspensions under a certain load (i.e., weight of the car plus/minus the downforce/lift].

I agree that results that are verifiable as you describe is a good way to quantify the accuracy of any given sim. But in any given discussion, this definition of "better results" is not apriori a given; but instead needs to be defined and agreed upon.

Well, ISIMotor2.5 (rF2 for that poster...) uses a set of concepts/assumptions inspired by a 2005 model originated by TNO Automotive. It marries some of the empirical characteristics of Pacejka based models (such as slip and grip estimation) with physical quantities/facets of the tire typical of physical models. At first glance, it is rather close to SWIFT, though Gjon will say it is a bit removed from that (I don't think so).

So, it uses dynamic tables which must be generated by the physics calibration development tools for a certain array of tires and situations.

As an aside, are these tools available to a modder or potential modder?

The thing about sim physics is that it is highly dependent on the proper calibration. For Discovery Science, iirc, McLaren Systems explained this dependency very clearly in how they set up their vehicle simulation for the F1 car and for the MP412c road-going supercar (and its GT3 version).

So, effectively, you can have a great physics engine and due to poor calibration models (the data you feed the engine: for the tires, the wings, the chassis, the masses, the suspensions) you will have a bad simulation.

There is more to all this than most people think.

Very well said.

If by realism you mean physical realism, we have several tools to verify the validity of a physics engine output.

If by realism you mean the opinions of people about how a car handle or should handle...that's not realism, but simply their flawed, misconception-filled opinions at stake.

Simple example: the president of Nissan (Carlos Ghosn) first tested the Nissan GT-R 35 a few years ago and the engineers explained how he should expect 2g cornering at some points on the track.

A racing/vehicle sim representing that car will have to show the simulated car cornering at the exact same lateral accelerations. If it fails to do so (I remember NFS road cars cornering at speeds which implied 4 to 5g), could we accept that as "realistic"? No. So, it's not that hard to check the validity of vehicle physics.

The bolded statement is circular. My point isn't that one persons take on what makes a simulation realistic is correct or incorrect. Rather what I am saying is that different folks can have different interpretations of what makes a simulator realistic, and that logically speaking it is something that should be defined before we can argue about conclusions.

Anyway your posts in this thread have been quite informative and interesting. Thanks for sharing.

 

[AnAL0G]

Meh, I'll save it. But...

In order for fully represent the performance envelope of an F1 car, a simulator and its motion simulator platform would have to develop 4g-level longitudinal forces and up to 5g lateral accelerations. We have no equipment, no technology capable of simulating forces and inertial forces of this magnitude (other than equip a centrifuge pod with a simulator, and even that would require a complex system to instantly convey the forces and instantly augment or decrease them). That doesn't mean we have no standards by which to judge the accuracy of a sim - professional or commercial.

First, you just restated my point, although more elaborately. Second, the bolded part is not what I said, is it?

 

[MikeSaints]

You missed my point. You two can write pages on end back and forth, but even with the best technology you can nitpick one vs. the other because they all miss out, even F1 simulators, etc. And they're not my trees, they're yours (you all are the ones going back and forth), which was the point. But if you all want to go on and on about it then by all means. I've driven all the sims, and none of them 'feel' like driving any kind of real car.

I did get your point, AnALOG. But I do make it a point (pun intended) to show how that reasoning is flawed on a very fundamental level: you can't complain about these sims not being accurate/realistic because they don't simulate real forces, when in fact NO SIM can do it.

As you, I also believe nothing feels like the real thing. Unlike people like Mark Webber, I do see the value of these sims as both a driver training tool and engineering tool (though commercial products such as GT5/Race07/iRacing/NKP/rFactor need some extra tools to make them helpful engineering wise).

As a matter of fact...the pilots I had the fortune to work with knew exactly how to best take advantage of these "cheap" simulations, in particular those with an engineering background.

 

[AnAL0G]

You tree'd my edit.

I never complained about their realism.

you can't complain about these sims not being accurate/realistic because they don't simulate real forces, when in fact NO SIM can do it.

This right here was my initial statement to both you and Bazz, just in other words. We have no argument with each other.

Shouldn't have said anything, just seemed to me that you guys were getting too worked up and invested in something I see as a non-argument. Carry on.

 

[MikeSaints]

I agree that results that are verifiable as you describe is a good way to quantify the accuracy of any given sim. But in any given discussion, this definition of "better results" is not apriori a given; but instead needs to be defined and agreed upon.

We are talking about physics, CC. Physics require mathematical objectivity; the only definition required is that which proceeds from physics and math.

As an aside, are these tools available to a modder or potential modder?

Yes.

Sadly for me, though, I am not jumping on the bandwagon just yet. It takes hours to have a single set of tires ready for rF2 (the computation is intensive and is very time consuming). As a modder, hmmm, that puts me off a little. Hopefully, newer tools, newer cpu technology will yield faster results. I hope.

But please don't let that get in the way of trying. In my case, with my 4 year old computer, it's a no-go.

Rather what I am saying is that different folks can have different interpretations of what makes a simulator realistic, and that logically speaking it is something that should be defined before we can argue about conclusions.

That's where we disagree. If we were talking about handling, then by all means, let us discuss standards and interpretations. One man's fantastic handling ir another man's awful handling. The same pilot, with the same car, in the same race, can complain about differences in the handling of the car (mostly a result of the setup). Highly subjective, I'd say.

But physical realism? No standard, no interpretation is needed other than the one given by equations that govern our current understanding of physical systems - i.e., no subjectivity involved.

 

[AnAL0G]

One man's fantastic handling ir another man's awful handling. The same pilot, with the same car, one the same race, can complain about the handling of the car (mostly a result of the setup).

But physical realism? No standard, no interpretation is needed other than the one given by equations that govern our current understanding of physical systems - i.e., no subjectivity involved.

This I agree with 100%, but how would you respond to Bazz's accusation about iRacing real-world drivers saying the simulation is completely off despite the numbers lining up?

 

[MikeSaints]

This I agree with 100%, but how would you respond to Bazz's accusation about iRacing real-world drivers saying the simulation is completely off despite the numbers lining up?

earlier I referred my professional/personal experience with motorsports professionals (some at NG and RD know which drivers and teams I worked for). These people had no commercial/financial ties to SIMBIN, ISI or Reiza, no reason to be favourable to rFactor, Race07 or GSC. I can't say the same (who can?) about the long list of drivers who have expressed how astonishing iR is (though most speak of the laser scanned tracks).

Then there're the other drivers - Dave Basu, Tommy Milner, etc, who believe most iR cars are, and I quote, "broken", "wrong", "too unpredictable".

These cars can't be "right" by the numbers either, sorry. I saw the telemetry of the Mp412c (a known real world racer posted them at RD) simulated by iR and you can definitely see that, while it hits the right "markers" (top acceleration, max longitudinal and lateral forces, etc) it fails elsewhere - namely, slip, grip and suspension loads.

Some numbers line up, others do not. And because of this, and quite rightly, real world drivers complain that something is still off with iRacing's physics engine.

I would say this: there are two ways of simulating cars, by its effects (i.e., making sure it hits the right markers at all times, and this would require a complex array of mathematical algorithms) or from the ground up, studying the system, what makes it up and how it works. Either way can produce a realistic simulation. Either can fail completely.

For me and many others, iR will be great when ALL numbers line up: from suspensions, to aeros, to tires, to forces. That will only happen when:
- iR's tick rate is high enough across the board (i.e., not leave behind part of the suspension system)
- the tire model is finally stable
- the car calibration has finally settled

Hope you don't mind the long answer.

 

[guybo]

There is no sim without rain.

Any "sim" that does not have rain, is only a racing game. Weather cannot be separated from motorsport (well, except for oval racing and even then temp/wind is a huge factor)