FM Vs GT - Discussion Thread (read the first post before you post)

  • Thread starter Scaff
  • 8,743 comments
  • 501,258 views
First, we don't live in a "theoretical" world and second, like you said yourself, outside of the actual game. How the real time game engine interprets the given information is a completely different story.
No, we don't live in a theoretical world, but, for example, that's pretty much what T10 is doing with the Pirelly data for FM4's tyre model: Work out a way to incorporate the raw data into the game's physics engine without having to fiddle with the data itself and therefore, to alter it unnecessarily.

Of course, it IS, otherwise people wouldn't even argue about which game is more realistic. You can do a lot of things regarding this with minimal subjective impressions though, but that needs a lot of time and testing.
Which would mean driving all (or nearly all) of the cars in the game on all (or nearly all) of the games tracks at race speeds, and by different people to avoid subjeticivtiy. I doubt any developer is going to go to such lengths to provide an audience with the most realistic experience possible, especially since most people in the audience wouldn't even be able to tell the difference.

I know people who think Shift 1 is as realistic as iRacing...
You don't even have to go that far. Some article about FM4 mentioned that the 599 GTO understeered into a corner before transitioning to oversteer - and a few people instantly assumed that to indicate false physics, because "OMG, RWD carz don't understeer, y0!"

I'm partially kidding, of course, but it just shows how much the expectation of what's realistic and reality can differ.
 
No, we don't live in a theoretical world, but, for example, that's pretty much what T10 is doing with the Pirelly data for FM4's tyre model: Work out a way to incorporate the raw data into the game's physics engine without having to fiddle with the data itself and therefore, to alter it unnecessarily.
Well, they are "working out a way" to incorparate "the raw data", that is fiddling around. They look for the best trade-off, which is exactly what I meant with the physics stuff.

Which would mean driving all (or nearly all) of the cars in the game on all (or nearly all) of the games tracks at race speeds, and by different people to avoid subjeticivtiy. I doubt any developer is going to go to such lengths to provide an audience with the most realistic experience possible, especially since most people in the audience wouldn't even be able to tell the difference.
Why do you think it takes so long to develope a proper race sim. and why good mods for PC games take ages until they're done? Every time a car wasn't tested this way, you will feel something's off (of course only if you know how it should be). Of course you don't test it on every track, but you have to do at least some testing for each car. That also explains why some cars in games like GT5 or FM3 drive more like their real-life counterparts than other cars.

You don't even have to go that far. Some article about FM4 mentioned that the 599 GTO understeered into a corner before transitioning to oversteer - and a few people instantly assumed that to indicate false physics, because "OMG, RWD carz don't understeer, y0!"

I'm partially kidding, of course, but it just shows how much the expectation of what's realistic and reality can differ.

Yeah that's annoying.
 
From the FM2 booklet:

Basically, we were developing a math model for
tuning the cars. The hardest thing is, you can’t find, for example, spring rate
data for all 300 cars. You’re lucky if you can get it for 20% of them, from the
manufacturers and from research. Sometimes you can get that kind of data
on fan Web sites, which can be freaky that way. Sometimes we contacted
the spring manufacturers, not the car manufacturers. The manufacturers we
contacted were based on the continuum of all of the cars, so we could get
a really even spread of data on low-end cars to high-end cars to race cars,
finding out about progressive springs versus linear springs, and so on.

Then we looked at ride height, the weight of the car, and what we call the
“goodness rating,” from reading reviews and learning a lot about each car.
We ranked the cars based on their “goodness,” and arranged those values
into buckets. For example, we might put a Ford Focus SVT a little lower than
a Subaru WRX STi. When we set the spring rates for those cars, and we don’t
have the actual spring rates for them, we use a formula, a mathematical model,
to automatically tune those numbers for us. Then we put in critical damping,
and offset damping with “goodness,” ranking cars by region. A lot of this we
call “automagic”—it’s our voodoo magic that we do in the game, and it’s the
only thing that makes it possible to tune 9,000 numbers on 300 cars.

Proving that this automagic model could work to tune the cars was a big part
of our getting the green light for version one. And there’s a ton of testing that
goes on. We list and graph the numbers we get, looking for anomalies, and
then we test the cars by hand. For example, one may come out with a really
loose spring rate. Sometimes we find that our formula isn’t taking weight
distribution into account as well as it should. Then we rework the formula,
reexport all the cars’ values, and retest. That’s a lot of systems—engine systems,
such as turbo pressure, rpm, inertia, and so on. Some things we get to research
a lot, like all of our stock turbo pressures. We went through and tested them,
and made sure that they were in the right stack rank relative to each other and
that their results in the game matched our research and knowledgebase.

After the automagic has done its thing, then you go in and hand tune.
For instance, getting the nuances of the suspension right is less about oversteer
and understeer and more about controlling the car with the throttle through
a turn; how getting onto, off of, and back onto the brakes creates oversteer;
how braking and throttle techniques affect the car. We’ve got a really good
team—guys who are rally drivers, guys who have driven all kinds of cars. So
we start tuning that way. But inevitably there are cars no one on the team has
driven, like the Lancia Stratos. We had trouble finding reviews on it; we just
knew it was a famous rally car, but our physics got a ton of that stuff right.
You start putting in the weight distribution, the size of the car, its moment of
inertia, and it starts getting better and better.
A lot of simplification and manual tweaking is done. Forza has not a magical formula to make cars.. physics holes aside the process explain why some cars feels and perform very inaccurate.
 
Last edited:
Well, they are "working out a way" to incorparate "the raw data", that is fiddling around. They look for the best trade-off, which is exactly what I meant with the physics stuff.
Dan specifically said that they're not interpolating the data - which can be done by, for example, re-working parts of the engine to take in the data as it is. This, of course, says nothing about what the engine will do with the data it has, of course.

Why do you think it takes so long to develope a proper race sim. and why good mods for PC games take ages until they're done? Every time a car wasn't tested this way, you will feel something's off (of course only if you know how it should be). Of course you don't test it on every track, but you have to do at least some testing for each car. That also explains why some cars in games like GT5 or FM3 drive more like their real-life counterparts than other cars.
Do you honestly think that T10 or PD went out to test drive all of those cars? Without mentioning that publically?
That's what driver feedback is for, if at all. Testing all cars in a game would bust almost every budget there is. Personally, I'd say that creating the physics engone itself and modelling the assets the game will use are among the most time consuming things for a sim. As for mods, it usually takes so long because the people who are making them are usually having jobs and are a smaller team.

Also, at times, it's just coincidence that some cars feel more true to their real counterpart...
 
Most, yes and Zer0's post proves that actually.
Where exactly does it proof that most of the cars have been test driven? :odd:

We’ve got a really good team—guys who are rally drivers, guys who have driven all kinds of cars. So we start tuning that way. But inevitably there are cars no one on the team has driven, like the Lancia Stratos.
They had people who drove some cars, that's quite a bit different from test-driving the cars to provide the necessary data.
 
Can't you read?
Proving that this automagic model could work to tune the cars was a big part
of our getting the green light for version one. And there’s a ton of testing that
goes on. We list and graph the numbers we get, looking for anomalies, and
then we test the cars by hand.
After the automagic has done its thing, then you go in and hand tune.
For instance, getting the nuances of the suspension right is less about oversteer
and understeer and more about controlling the car with the throttle through
a turn; how getting onto, off of, and back onto the brakes creates oversteer;
how braking and throttle techniques affect the car.
That's all manual testing and tuning, without that, the cars could come out buggy and far from realistic.
 
Can't you read?

That's all manual testing and tuning, without that, the cars could come out buggy and far from realistic.
First of, manually tuning doesn't mean that the cars have been driving, and it is even stated that they've been using reviews and such to guess how the car should behave. So, for the "can't you read?" part, right back at you.

Also, asking, say, Ken Block how a rally car drives is different from test-driving it. Getting feedback does not equal test driving a car for the game. Oh, and "testing", since when does "testing" mean test driving? If someone wants to say they test drove cars, they'd say "test driving" not "testing", no?

Regarding the "testing by hand": They're specifically writing about anomalities in their numbers in the same paragraph, guess which cars they're testing? The ones in the game, of course...
 
Proving that this automagic model could work to tune the cars was a big part
of our getting the green light for version one. And there’s a ton of testing that
goes on. We list and graph the numbers we get, looking for anomalies, and
then we test the cars by hand.
And yes, of course manual tuning also implies testing, how the hell do you want to know if your tuning was correct without the testing?

Oh, and "testing", since when does "testing" mean test driving?
Oh come on, seriously, what else should it mean when trying out physics? Looking at the trunk or what?

EDIT: Slowely I get the feeling you don't even remotely know what I am talking about, of course I mean the in-game cars.
 
Two things:

One, there are several making of videos of GT4 in which they test-drive a bunch of cars(Mazda , Mitsubishi, Honda).

Second, KY is a racing driver as far as I understand, is not Vettel but he is capable of writing the game, and that is something worthy of mention, because provide feedback is not the same as experience/direct coding.
 
And yes, of course manual tuning also implies testing, how the hell do you want to know if your tuning was correct without the testing?
Proving that this automagic model could work to tune the cars was a big part of our getting the green light for version one. Andthere’s a ton of testing that goes on. We list and graph the numbers we get, looking for anomalies, and then we test the cars by hand.
Because that's them talking about the real cars, right? Is it that hard to understand that they're talking about testing the in-game cars, the virtual ones?

Two things:

One, there are several making of videos of GT4 in which they test-drive a bunch of cars(Mazda , Mitsubishi, Honda).

Second, KY is a racing driver as far as I understand, is not Vettel but he is capable of writing the game, and that is something worthy of mention, because provide feedback is not the same as experience feedback.
Nobodies saying there's no first hand experience at all, but the thing is: Is it done for the majority of the cars? I highly doubt they've driven even half of the cars that appear in GT5.
 
Dude, you are the only person talking abot real cars.
I'm talking about testing the cars in the game, which is still time consuming as hell.
 
Dude, you are the only person talking abot real cars.
I'm talking about testing the cars in the game, which is still time consuming as hell.
Well, d'oh, that wasn't what I was on about, at all. I guess we just missed each other point completely :lol: :cheers:

Anyways, my point is this: Tweaking the cars by hand is of course what's happening. All I was saying that it wasn't impossible in theory to circumvent that and is, partially, being done now.

And, more importantly, that manually tweaking a car will always be subjective - and I doubt that it can recreate the real feeling of the car without knowledge of how the real car feels. I mean, how are you going to recreate something you don't know, aside from guesswork?

It's like asking someone to paint an animal they've never seen from a discrpition...
 
So FM coding uses all the car data without tweaking?
No, they are not (read Zer0's post again), the question is how good and intense do they test and how good their "automagic" stuff is. But I guess we will never know.
 
From the FM2 booklet:


A lot of simplification and manual tweaking is done. Forza has not a magical formula to make cars.. physics holes aside the process explain why some cars feels and perform very inaccurate.

If the same method was carried over to FM3 then that explains the inaccuracies as you mentioned.


DAVE
 
No, they are not (read Zer0's post again), the question is how good and intense do they test and how good their "automagic" stuff is. But I guess we will never know.

What about algorithm complexity, the more variables taken in count, the more accurate the game will be(according to mathematical considerations for accuracy).

I have seen "moders" who takes the 3D model out of the game to make mods for other games(FM3 GT40, I have seen it in plenty of mods), is it possible to take out the scripts file too?(I haven't seen this happening in GT5 models though).
 
is it possible to take out the scripts file too?
No, you have to redo the physics for the game you want to convert the car, which is why a lot of these mods are horrible. I think it's a bit different with cars from shift, because they still use some form of the ISI engine, so it's easier to make good physics for GTR² for example.

It's the same when you convert a car from rFactor.
 
No, you have to redo the physics for the game you want to convert the car, which is why a lot of these mods are horrible. I think it's a bit different with cars from shift, because they still use some form of the ISI engine, so it's easier to make good physics for GTR² for example.

It's the same when you convert a car from rFactor.

I know that, I'm not talking about converting them, but rather reading them, is it possible to get the "script" file and then see if its got the car data, just to determinate how many data it has (like tweaking aircraft for FS).

Also, I think that this goes a bit towards AUP boundaries, so I just leave it as "such stats can be read? Y/N?"
 
Well, if you have a program which can read these files yes, but I have no idea what file formats they are using etc. The system used in GT5 seems to be even more complex, that's most likely why you don't see any GT5 conversions.
 
So the verdict is the way FM and GT create cars weren't much different, just FM has a systematic process to help them rough in the parameters?
 
They cannot be too different, because game development follows certain restrictions like everything. I bet PD also uses a system for the parameters etc. similar to that "automagic" stuff, maybe the GT5 engine does a lot of these things itself though, I have no idea.
 
So the verdict is the way FM and GT create cars weren't much different, just FM has a systematic process to help them rough in the parameters?

No. GT5 is more complex because you can't read them. :)

the full booklet for those who want know the whole history (without other's bias distortion)
http://download.microsoft.com/downl...b7-44e4-ba61-32ea6114f54c/Forza2MNL_LE_EN.pdf

I don't have the FM2 LE, so thanks for bring it to discussion. Now I'm more even convinced about the effort that Turn 10 put in the physics.
 
No. GT5 is more complex because you can't read them. :)

Not sure about that, the incapability of reading the code doesn't mean that is more complex, the only thing that it means is that is safer.

I'm also not sure about the complexity of the scripts in GT5 or in FM3, since they both are "code locked", I just know that there should be an index for the different terrain types in GT5, I know that there is an algorithm that reads or calculates tire temperature and grip in FM3, and so on. Its nice too see the considerations that go under FM2 programming, which makes me wonder about the complexity of FM3.

I guess that is just testing and testing to actually compare the different characteristics for each car, and how the physics work for each game.
 
Not sure about that, the incapability of reading the code doesn't mean that is more complex, the only thing that it means is that is safer.
I think he's being sarcastic ;)

We can't really say which is more complex, mostly because we can't compare the two. Plus, better results don't necessarily mean a more compllex engine.
 
I'm a bit late to the lift-off oversteer party, but those looking to experience Lift-off oversteer in FM3 should try these cars.

2005 Porsche #3 Lechner Racing 911 GT3 Cup
Lotus 2-Eleven
 
I think he's being sarcastic ;)

We can't really say which is more complex, mostly because we can't compare the two. Plus, better results don't necessarily mean a more compllex engine.

I sincerely hope is sarcasm, and I do agree, a more complex engine doesn't mean more accurate results, the only way to really know is by driving in both games(which is possible and can be done) and then drive their IRL counterparts(which is pretty much impossible unless you're Clarkson or an Autocar reviewer).

And, no, the power of the PS3 don't have anything to do with the complexity of games, the platform is harder to program, but the games themselves follows the basis set by the game engine, needless to say, while the PS3 is capable of processing more data, that doesn't necessarily mean that GT5 is more complex.
 
Back