Myth: Gran Turismo Sport and thus GT7 have unrealistic Front Wheel Drive Understeer. Status: BUSTED!

  • Thread starter Magog
  • 228 comments
  • 33,874 views
To me, Gran Turismo is a sim, it aims for realism, it's just not the most accurate sim out there and has a few glaring issues which I'd be very happy if they fixed.
See, I'm not sure that's fair to Gran Turismo or Polyphony. If we assume that they're making the game as realistic as they can but that GTS is the result, then the conclusion is that they're just not very good at their jobs.

The alternative is that they're not aiming for absolute realism. If the intention was always to make a game with a physics system that's realistic-ish but simplified to cater to less experienced drivers then what Polyphony produced is exactly what they intended to produce, and the conclusion is that the devs are really quite good at their jobs.

I think it's worthwhile being clear about the design intention, and separating the design aspects from the developmental execution. It isn't the same people doing both, and I don't like blaming people for things that aren't their fault. I assume that the devs did what they were told to do. There is still some evidence to suggest that they didn't necessarily do a great job however, stuff like reversed ride heights wouldn't be desirable even in a simplified system.

There's a separate discussion to be had around the best ways to simplify a "realistic" physics system in order to make it accessible while keeping as much of the depth and driving pleasure as possible. I'd argue that the design choices that Gran Turismo made to try and achieve that goal are not the best, but that starts by assuming that the goal is more complex than "aim for realism".

Judging Gran Turismo solely on how realistic the physics are feels like it ignores a large portion of what makes the physics system in Gran Turismo attractive at all. People get caught up in the "Real Driving Simulator" tagline and the marketing BS that Kaz spouts and miss that the game is very much not about absolute realism.
I don't particularly like the term simcade, it doesn't really have a set defintion and I tend to think more of titles much less realistic than Gran Turimo when I hear that term.
It doesn't have a hard definition because it's defined by being the combination of two terms that are absolutes. Telling me that a drink is a rum and coke tells me nothing about how much rum or how much coke is in there - but it's a useful starting point for a conversation. I know the drink is neither straight rum nor straight coke, and then we can talk about how ****ed up I'm going to get if I drink it.

I've seen people try to organise games on a scale, like so.

1641859780477.png


Or you can just rate them out of ten or something. Let's say ACC is 9/10 sim points. Wipeout is 0/10. Let's call FH5 a 5/10. We have a baseline to talk about. We can put GT on this, and depending on who you ask it's going to be somewhere between a 6 and an 8. Where you actually put it isn't hugely important, because like any tier list it's then more about the discussion around why it's where it is that's interesting.
 
Yes, the OP still has a lot to learn, but most of us have been there at one point. :lol:

It's all good though, sometimes you need people that are less knowledgeable on a topic to get a discussion moving. An exchange of ideas takes place, and sometimes you get some new and interesting information out of it in the process. Humanity would be quite stagnant and dysfunctional if no one ever disagreed on anything.
 
Last edited:
@Imari that's all very sound logic, I just personally don't like the term. And no, I wouldn't say Gran Turismo is as realisitc as it can be, just that it's aiming for realism. Ultimately if you break down any sim you'll find flaws, none of them are as realistic as they can be beause they all have to make compromises in one way or another.

The more hardcore PC sims for example have to work on a variety of processors and GPU's that aren't all the most powerful options available. Of course, GT Sport isn't in their category, but my point is simply that on a technical level, none of them are as good as they could be if you aimed squarely for getting the most out of the best hardware.

I understand why the term exists, I just don't like to use it, personally the scale would look pretty different if I were to categorise what I felt fell in the sim, simcase and arcade categories, there's already one that's hurting my eyes to see so far to the right, iRacing. I wouldn't neccessaarily shuffle the order of many games, but where sim and simcade seperate would probably be further to the left. But then I guess on that scale GT Sport is already almost on the border, so my thinking that in my view it's not quite simcade isn't too far removed from that chart.

As to the order of the games, I guess it starts to become subjective when you start comparing sims with different strengths and weaknesses. That's why you get the Forza 7 is more realistic than GT Sport v GT Sport is more realistic than Forza 7 arguments, they aren't all that different overall, they just have different strengths and weaknesss.

@Tornado Shox is lower down than RFactor, I think that means it has worse graphics.
 
Last edited:
GT will never be a "full" sim like the PC stuff because it would alienate a whole group of casual players.

Nothing in the GT universe is particularly "hard" to drive, or drive very quickly per se, compared to how they'd act in real life or in the more hardcore sims.

Hell, I WISH GT would add in a host of front drive touring cars from the 90s with the proper physics, those things are a blast and relatively all over the place and you need to commit to go fast but the casual player which makes up most of the fanbase would hate stuff like that and would call the game "too hard" and it would cause a ton of would be players to avoid it.

Stuff like GT is merely the stepping stone towards to full, hardcore sims that exists and that's totally okay that it occupies that position.
 
Last edited:
Nothing in the GT universe is particularly "hard" to drive, or drive very quickly per se, compared to how they'd act in real life or in the more hardcore sims.
I disagree, as techniques I have developed in reality I can't use in GT, which makes adapting to GT unnatural and harder than they should be, particularly if it's an area that PD pick to switch back and forward between, such as understeer FFB rattle.

Hell, I WISH GT would add in a host of front drive touring cars from the 90s with the proper physics, those things are a blast and relatively all over the place and you need to commit to go fast but the casual player which makes up most of the fanbase would hate stuff like that and would call the game "too hard" and it would cause a ton of would be players to avoid it.
I often see this claimed, yet without any data to back it up at all. Driveclub modelled list-off oversteer in FWD cars to a minor degree and that had a shed-load of players and no-one called it 'too-hard' PD could even reduce the degree of lift-off oversteer that FWD cars produce to make it 'easier', it would still be the correct balance for the cars in question and would addressed the (claimed) issue with lift-off oversteer. Nor does such a claim slack-up with the issues that MR and RR cars have, as they don't understeer when they should, and move straight to oversteer all the damn time.

A default bias towards exaggerated oversteer has also never hurt the Forza series either, so without some actual citations I don't buy this 'reason' at all.
Stuff like GT is merely the stepping stone towards to full, hardcore sims that exists and that's totally okay that it occupies that position.
The issue is that it fails in its attempt to do that if you then have to re-learn how some of the drivetrains act, how understeer reacts via FFB, etc.

The majority of titles around I can jump into and apply real world techniques and the cars react generally how they should and the FFB that's delivered does what it should; however some titles don't, such as FM and GT.

This is one of the reasons why Driveclub was actually a really interesting title, as it was a firmly arcade title with solid physics underpinning it, the cars react as they should do, but in a simplified manner. DC didn't resort to having some drivetrains react utterly counter to how they should, Wreckfest is also similar in this regard, with both being evidence that its perfectly possible to use a simplified, but generally accurate physics model.
 
Last edited:
I've often wondered, and don't recall ever reading anything in depth on the subject, how PD and other game developers model how a car 'handles'.

I assume, at least these days, that it's data-led in that they take the raw data of a IRL car - wheel base, track, spring rates, weight distribution, centre of gravity, which wheels are driven etc etc, and dial it all into its game physics model to come out with a digital representation of that particular car that should be fairly accurate.

But i wonder how much, if any of it, is tuned by actual human 'feel'?
 
I've often wondered, and don't recall ever reading anything in depth on the subject, how PD and other game developers model how a car 'handles'.

I assume, at least these days, that it's data-led in that they take the raw data of a IRL car - wheel base, track, spring rates, weight distribution, centre of gravity, which wheels are driven etc etc, and dial it all into its game physics model to come out with a digital representation of that particular car that should be fairly accurate.

But i wonder how much, if any of it, is tuned by actual human 'feel'?
From what I know it varies massively, not just from title to title, but within a title as well.

Certainly a number of titles have used human 'feel' to fine tune some of the cars, which is why many have professional drivers as part of the dev team, SMS made a big marketing play of this during PC and PC2's development (and tellingly didn't say a thing about it during PC3s development), but it's always going to be difficult to know exactly how involved they are.

Even in titles and studios that claim to do it, some cars are going to be impossible, either because they are so rare and/or valuable that your just not going to get access to them in that way, that they are concepts that only a few have driven or have incomplete drivetrains.

This is a good example of a concept making it into PC3 with no 'human feel' being part of it (despite the inaccurate claims in the thread to the contrary).
 
This is one of the reasons why Driveclub was actually a really interesting title, as it was a firmly arcade title with solid physics underpinning it, the cars react as they should do, but in a simplified manner. DC didn't resort to having some drivetrains react utterly counter to how they should, Wreckfest is also similar in this regard, with both being evidence that its perfectly possible to use a simplified, but generally accurate physics model.

I imagine you can throw most of the PGR series (specifically 3 and 4) into that category as well considering most every vehicle, and even the bikes in 4, all handle differently, and require different tactics to wrangle the most performance out of them, or to have them succeed in the events that you are given.
 
I've often wondered, and don't recall ever reading anything in depth on the subject, how PD and other game developers model how a car 'handles'.

I assume, at least these days, that it's data-led in that they take the raw data of a IRL car - wheel base, track, spring rates, weight distribution, centre of gravity, which wheels are driven etc etc, and dial it all into its game physics model to come out with a digital representation of that particular car that should be fairly accurate.

But i wonder how much, if any of it, is tuned by actual human 'feel'?
A problem that games like Gran Turismo and Forza have is that they have so many cars, and it takes a lot of time to correctly set up the physics model for a car. They all have some kind of parameterized physics template that considers stuff like weight, wheelbase/track, engine layout/power/torque, gearing, etc. that allows the developers to quickly set up car physics models by just entering some numbers that they can easily get from the manufacturer. But such simple, generic models aren't very accurate.

Assetto Corsa Competizione has such great physics not only because the physics engine models a lot of important stuff in detail, such as tire pressure/temperature/wear/flex on multiple positions per wheel, but also because they have a relatively small and specific set of cars (GT4 and GT3 from 2019-2021), so the developers can get feedback from professional drivers who drive those cars (some of the devs even drive those cars occasionally themselves) and really invest in tweaking their physics models of those cars close to perfection. Even though Polyphony and Turn 10 have a lot of resources, they probably don't have the resources tweak 400+ cars to the same level of perfection that the ACC devs can do. The ACC devs probably only need one GT3 driver and one GT4 driver to get the feedback they need (OK, maybe a bit more, since not all GT cars are the same). PD/T10 would need test drivers from Mercedes, Ford, BMW, Audi, Madza, ... basically like 50+ professionals from all over the world.
 
Last edited:
I imagine you can throw most of the PGR series (specifically 3 and 4) into that category as well considering most every vehicle, and even the bikes in 4, all handle differently, and require different tactics to wrangle the most performance out of them, or to have them succeed in the events that you are given.
Not to get too off topic but I adored PGR 4 for this reason. Combine that with it's great track designs and gamemodes, felt like there was infinite replay value. I should dust off my X360...
 
Last edited:
I've often wondered, and don't recall ever reading anything in depth on the subject, how PD and other game developers model how a car 'handles'.

I assume, at least these days, that it's data-led in that they take the raw data of a IRL car - wheel base, track, spring rates, weight distribution, centre of gravity, which wheels are driven etc etc, and dial it all into its game physics model to come out with a digital representation of that particular car that should be fairly accurate.

But i wonder how much, if any of it, is tuned by actual human 'feel'?
In the end, every car is a composite of measurements. Distinctions between car models are primarily a product of their geometry, because small differences in the position or orientation of physics parameters produce outsized effects. If something must be tuned by feel, it is ultimately for a lack of a particular measurement or set of data (eg. electronic differential programming), whether it was not provided/collected or is difficult to measure.

An accurate center of gravity is one piece of data that may be hard to come by. It's also only a point -- it does not describe the distribution of mass throughout the vehicle's 3D form. Two cars may have similar resting weight distributions, but distinctly different moments of inertia; you need the first data point for accurate weight transfer, and the second data point for an accurate response to that shifting weight in a yaw situation. Even if manufacturers were to supply that data, I don't think it would be in a format developers could readily use. I imagine it's one that is commonly hand-tuned.
 
An accurate center of gravity is one piece of data that may be hard to come by. It's also only a point -- it does not describe the distribution of mass throughout the vehicle's 3D form. Two cars may have similar resting weight distributions, but distinctly different moments of inertia; you need the first data point for accurate weight transfer, and the second data point for an accurate response to that shifting weight in a yaw situation. Even if manufacturers were to supply that data, I don't think it would be in a format developers could readily use. I imagine it's one that is commonly hand-tuned.
And then you have to factor in roll centres as well, the yaw point, etc.

Again little if which will come from an OEM and all of which is dynamic.
 
I see it as PD simulating “real driving”. We all pull the right trigger, car go. We all pull the left trigger, car stop. Now, “real driving” as accurate as real life, with all real world finer detail? Not as good as some simulators, but better than other simulators.

I mean, why have “Arcade Mode” as a feature, if you’re simulating real driving? Why have two modes? In a real driving simulator?
I think it’s been a long line of the interpretation and translation. In the real world, there’s no “arcade mode”. If anything, real life arcade mode, would be a track day(s). Simulation Mode would be Driving instruction through to professional racing series.

I don’t use simcade. I just name the game. People pretty much know what to expect with the physics in a GT game. Mentioned before how people feel they can control the car. So, maybe, that means to them, it’s simulating real driving. Whether the physics are accurate or not, the car moves forward, side to side and reverses.
 
In the end, every car is a composite of measurements. Distinctions between car models are primarily a product of their geometry, because small differences in the position or orientation of physics parameters produce outsized effects. If something must be tuned by feel, it is ultimately for a lack of a particular measurement or set of data (eg. electronic differential programming), whether it was not provided/collected or is difficult to measure.

An accurate center of gravity is one piece of data that may be hard to come by. It's also only a point -- it does not describe the distribution of mass throughout the vehicle's 3D form. Two cars may have similar resting weight distributions, but distinctly different moments of inertia; you need the first data point for accurate weight transfer, and the second data point for an accurate response to that shifting weight in a yaw situation. Even if manufacturers were to supply that data, I don't think it would be in a format developers could readily use. I imagine it's one that is commonly hand-tuned.
Absolutely. My RX-7 has 50:50 weight distribution. My Boxster has 48:52 weight distribution. Dynamically, they feel nowhere near similar. Once things get moving around, the RX-7 yaws it's back (Imagine the pivot point is right at the front axle centerline) just like any other FR car and the Boxster just welds the rear tires to the ground (Imagine the pendulum is right next to your thigh) - it's a totally different experience, particularly on corner exit, that can't be explained just by static weight distribution.
 
Absolutely. My RX-7 has 50:50 weight distribution. My Boxster has 48:52 weight distribution. Dynamically, they feel nowhere near similar. Once things get moving around, the RX-7 yaws it's back (Imagine the pivot point is right at the front axle centerline) just like any other FR car and the Boxster just welds the rear tires to the ground (Imagine the pendulum is right next to your thigh) - it's a totally different experience, particularly on corner exit, that can't be explained just by static weight distribution.
Boxster has staggered tyre sizes that the RX7 doesn't have as well though, right? That changes the balance significantly.
 
As I've said before, if PD do their homework they can make a physics engine with enough depth and forgiveness to please the majority. I like ACC but it's very hard to drive on the limit and a good setup is required to extract the maximum performance. That can be quite tiring and I don't want that for GT7. I want to get in and drive at a decent speed out of box, but I don't want to adopt a strange technique to do so. It should be the same as real life.
 
As I've said before, if PD do their homework they can make a physics engine with enough depth and forgiveness to please the majority. I like ACC but it's very hard to drive on the limit and a good setup is required to extract the maximum performance. That can be quite tiring and I don't want that for GT7. I want to get in and drive at a decent speed out of box, but I don't want to adopt a strange technique to do so. It should be the same as real life.
I think we will get this good balance with GT7, the footage for me looks promising
 
As I've said before, if PD do their homework they can make a physics engine with enough depth and forgiveness to please the majority. I like ACC but it's very hard to drive on the limit and a good setup is required to extract the maximum performance. That can be quite tiring and I don't want that for GT7. I want to get in and drive at a decent speed out of box, but I don't want to adopt a strange technique to do so. It should be the same as real life.
Which one is it? Because these are mutually exclusive.

If you want it the same as real life, you have to accept that some cars are hard to drive on the limit. All cars require specific setups to go as fast as possible at a particular track. Some cars you will not be able to jump in and be fast straight away, and you will have to learn new techniques to drive them well. That's why racing drivers are professionals who take years to learn their craft.

If you want a game that is simplified so that you don't need setups, you don't need to adapt to the particular car that you're driving and that all cars are easy to drive on the limit, you're explicitly asking for the game to be made unrealistic in those areas.

To be clear, that's fine. That's a choice and not necessarily a bad one for a game of Gran Turismo's type. But let's not pretend that it's more realistic. ACC being hard to drive on the limit (I mean, it's not particularly hard, they're GT3/4 cars) and requiring setups is far more realistic than the opposite.
 
I have a feeling we'll come back to this when the problems with FWD handling nothing close to reality still haven't been fixed in GT7.
Even if they fix it im pretty sure you still will **** on it regardless so it pointless to discuss it with you...
 
Last edited:
Even if they fix it im pretty sure you still will **** on it regardless...
I'll be the first to congratulate Polyphony if they do.

But that's the biggest thing. If. And I don't exactly have the highest estimations they will, considering legacy issues still persist in the series.
 
Yeah sure...
Knowing who this is coming from, who tried to pick a fight with a guy who (correctly) pointed out that GT's vehicle dynamics weren't exactly that great even compared to the relevant competition (and put Forza on the same level, which it absolutely is in with most regards), and had better knowledge in that topic then you did, I guess I shouldn't be surprised.

But whatever.
 
Knowing who this is coming from, who tried to pick a fight with a guy who (correctly) pointed out that GT's vehicle dynamics weren't exactly that great even compared to the relevant competition (and put Forza on the same level, which it absolutely is in with most regards), and had better knowledge in that topic then you did, I guess I shouldn't be surprised.

But whatever.
I myself i pointed some of the problems of the game on the physics thread that i started some time ago way before than anyone else but anyway...again im not going to discuss with diehard forza fans.
Enjoy your game and have a good day
 
Last edited:
I see it as PD simulating “real driving”. We all pull the right trigger, car go. We all pull the left trigger, car stop. Now, “real driving” as accurate as real life, with all real world finer detail? Not as good as some simulators, but better than other simulators.

I mean, why have “Arcade Mode” as a feature, if you’re simulating real driving? Why have two modes? In a real driving simulator?
I think it’s been a long line of the interpretation and translation. In the real world, there’s no “arcade mode”. If anything, real life arcade mode, would be a track day(s). Simulation Mode would be Driving instruction through to professional racing series.

I don’t use simcade. I just name the game. People pretty much know what to expect with the physics in a GT game. Mentioned before how people feel they can control the car. So, maybe, that means to them, it’s simulating real driving. Whether the physics are accurate or not, the car moves forward, side to side and reverses.

Arcade mode doesn't mean what you think it means. It's just a way to quickly race a selection of cars you don't own.
 
Arcade mode doesn't mean what you think it means. It's just a way to quickly race a selection of cars you don't own.
You do know the “cade” in simcade refers to arcade. Yes?
By your expalination of you thinking I‘m not understanding what I posted, a simulation is the same as Arcade. A Simulation game allows you “to quickly race a selection of cars you don’t own.“ Right?

If you didn’t quite understand what I posted, that is why I don’t use the term simcade: A simulation, but more arcade style in execution. It has aspects of real driving, but the dynamics are not as true to form as a more accurate simulation of realism.
 
Back