Physics engine: GP4 vs. F12002

First of all, concerning all the small wheel movements, what I see is the driver reacting to what he is feeling in the wheel whether it's bumps or momentary loss of grip, without even having to think about it. These reactions become almost hard-wired in a driver's brain after years of driving. He doesn't consciously decide exactly how much correction he needs for each situation. The bump moves the wheel and he sub-consciously reacts and corrects by the right amount. This physical reaction added to visual judgement guides him along the race line. That's where improved force-feedback would be required to obtain this in a race sim.

Real F1 drivers also have the advantage of feeling the car through G-forces which could never be recreated using only graphics and force-feedback. They can feel the car losing grip in more ways than we can in a sim. So if everything else in the sim was totally realistic, the lack of this feel (G-forces and all) would put the sim racer at a disadvantage compared to a real F1 driver. So to make a game playable and produce normal laptimes, they have to compensate in other ways, one of them being more grip.

This is exactly the reason why I - even after driving cars for 14 years now - always had problems with manual shifting in games like the NFS series or similar. Normally I do not care about the rev counter or the speed when I drive my private car. I simply feel when it's time to shift. Same goes for braking or acceleration. You feel it through your feet. I never really understood why they do not put FF effects on the pedals as well, or try to build pedal sets with a brake that really feels like a brake. Even if I strongly dislike agreeing with MS, there is one thing he is right about. He always says that the most important instrument a driver has to make good setups is the popo-meter (butt-o-meter), since it is the bum you feel the road with. This butt-o-meter, however, can (still) not be simulated.

---

REHAB IS FOR QUITTERS

Let's force a preliminary conclusion here:

1) The more parameters a physics engine has, the better. It's only logical, because you can build more mathematical functions and therefore the chances of getting closer to the reality are higher.

2) Even if there are more parameters and more functions, this does not necessarily mean that the model is accurate. If the functions are wrong (more or less) the behaviour will be wrong.

3) There are parameters that just cannot be modelled, using the current technology. The butt-o-metter, brake feeling, G-forces etc. are wishfull thinking, at least for now. This means that, no matter how good a model is, it can never be accurate. I will say that it's not even close to accurate. However, this doen not exclude the possibility for one engine to be better than another.

4) Under/Over modelling is necessary because we are talking about games. There is no way that a gamer can even dream that it's capable to control an F1 car. This is not for the gamers, this is for the elite of 22 drivers and some test drivers. Even the elite drivers have difficulties handling such monster cars (take a look at Takuma Sato).

These are generic conclusions, without going into a specific game. We shall do that later.



maniak

Oh, my public! How they love me! (Looney Tunes)

I would like to comment a little about "what you see on TV". Morbid had a good argument by saying that if you judge a simulation model by what you see on TV you are doing a reversed analysis, from effect to cause, which can lead you to erroneous conclusions. In theory I fully agree with Morbid.

However, from a practical view point, I think that the previous statement is not good enough. I would like to discuss (actually re-discuss) the correction a driver must apply to the steering, or at a more generic level, the hands movement.

We have all seen the comparative images between real F1 and GP4 driving. I think it's not that difficult to judge why the drivers make a particular movement. The main reasons are:

1) bumbs in the road - it translates in small constant left-right movements which are rather independent of the drivers' will.

2) tyre grip - it translates in violent variable left of right movements that are meant to correct the sliding of the car because of loosing the tyregrip. The movement is dependant on the drivers' will based on the informations it can gather from the car.

3) optimal line seeking - it translates in gentle left or right movements that are meant to position the car on the best possible trajectory througha curve. The movement is dependant on the drivers' will.

4) friction force, G-force, etc - it is similar to the movement triggered by the bumps, but one can tell the difference, within reasonable limits. It is noticeable in curves when the driver must defeat the above mentioned forces. The movements are more of a reflex, rather than rational.

Anybody wants to add anything?



maniak

Oh, my public! How they love me! (Looney Tunes)

There was an article in Autosport last week in which Allan McNish talked about driving an F1 car for the first time - it covered G forces (including the fact that Kimi Raikonnen's neck muscles could barely support his head after his first F1 drive) - the effect of the hitting the brakes flat out for the first time brakes etc. etc.

Worth a read in the context of this discussion - although the bottom line is that there is no way any of us who have not driven an F1 car can really know what it should feel like.

I have driven a variety of cars over the last 25 years - yes I'm that old! - from Sports Cars to heaps of junk and most driving sims from Nascar to GPL to F12000+ and of course the various GPxs. To me the driving experience is pretty similar (GPL is slightly different as you would expect!).

So overall, comparison appears relatively spurious - you will inevitably have EA fans and Crammond fans. I prefer GP2,3,4 because I find all of EA's sports games rather cynical and shallow - you name it FIFA, Madden whatever - and more interested in a glossy package overall than accuracy and preparing for the next annual tweaking session. That includes the latest which to me is no great improvement over F12000, F1CS2000, F12001 (crappy Jordan driving school - what is all that about?) and F12002.

As the man said earlier, until we have steering wheel sets that reproduce every facet of the driving experience whether they be through pedals pulleys, weights etc. and games that simulate all of the effects - no-one will ever resolve the physics argument.

My opinion - and that is what we are all entitled to - our opinions.

Good post Tony!

You're right in spot. There are fans on each side and they will be hard to convince that the other one is better. However, this should not prevent us to try and find rational arguments (you know, better than "that game sucks because it sucks" I think that people who wrote in this thread (even myself, every now and then came with solid arguments.

One way or the other, most of us know how a car handles. I know that there can be no match between a street car and a racing car, but still physics apply. We do not intend to have a Master degree in Physics, but I consider quite possible to qualify the "Physics accuracy" for the games in question.

And yes, we are entitled to our oppinions. Thank God is not a matter of life and death, I would have been dead by now



maniak

Oh, my public! How they love me! (Looney Tunes)

Hi maniak and all,

Sorry, can't agree with you here, but maybe we are not talking about the same over-/undermodelling properties.

Try thinking about it this way,
we have a system with inputs and outputs. The outputs are what we get on the screen, the inputs are the wheel and pedal actions. So we have, like Morbid said, a cause-effect relation here. Our "objective" is to get the feeling we are driving a real F1 car for the fun of it.

Now, how can we model the real thing?

1. Physics! Yupp, cool, possible but not always 100% sensible. The thing is that we may model a lot of effects which are not really relevant "for our objective", or which are physically impredictable, that is we only know rather imprecise parameters or model structures. So we may get "better" or "worse" results when including these effects into our game, or we may not ever see any difference!! The question is what is "better" or "worse"? I like to approach this question by thinking about our objective again. That is getting results closely resembling what we see on TV and somehow make sense to us (recall the excellent Morbid's damper setting theories!!!). Although the model may be "theoretically" more "precise" when augmenting some further physical component effects into the game, this augmentation is only sensible if is is useful for us "the players".

2. Simplifying the model? Yupp, can also be good, also when thinking about computational efforts. But also when we don't have all the "outputs" of the system available, such as neglected forces the real F1 driver can detect. The trick is however to compansate these effects in the "simpler" model, such that these missing measurement become less important. This can be done by adding black-box models into out scheme, these are models that have nothing to do with physics, but attempt to resemble what we would like to get, namely something similar to real-life behavior.

So in my opinion we can't speak of under and overmodelling, since the submodels included into the game may not alway be existent in reality. Because of this fact, it is not possible to compare the resulting model with something "real" which makes the model a stand-alone thing, only meant to improve the feeling of the game, which should actually be the main "objective" of the whole thing.

To answer Ronnie:

I guess we were talking about slightly different notions. I understand your arguments and I pretty much agree with them. However, I think that the sims are built having realistic models at the basis, but slightly modified in order to compensate the lack of information the gamer/driver experiences.

I will give you an example: If the car slides to the side, in real life you will be able to pick up information very soon, the moment you FEEL the car sliding behind you. This moment is sooner than the moment you see it sliding. Therefore, you start correcting the trajectory sooner, with greater chances to succeed. In games you will pick up the information when you SEE the car sliding, which is later, therefore you start the corrections later, when they could be useless.

Considering the above, by either under or over modelling some physics, the game compensates the lack of information by altering the behaviour of the car so the gamer/driver can succeed in its corrections. From this point onward, it's just a matter of taste how much you alter the model.



maniak

Oh, my public! How they love me! (Looney Tunes)

Yes, this is one way of doing it ... but in my opinion the wrong way.

Physics is something we should trust in ... so if we have a good model based of physic-laws, (and I know excellent models exist for this category) we should be happy and leave it the way it is ...

in the game however we "the drivers" are pretty lousy at "modelling" the true dynamics of the real F1 driver ... this is the case because of lacking information ... and this is where the black-box modelling should come in ...

what you call over/undermodelling (although I find the term misleading) should be extended models on the top of the realistic car model, only to compensate the missing "measurements" without (and this is the important part) changing the dynamics of the modelled car!

if this is done, we get results closely resembling what to expect from a real-life experiment ...

compensating the missing information by dropping out parts of by changing the physical model of the car is wrong and is not really necessary ... think about the driving aids you can select!

the best approach would be to have us "the players" acting as a "part" of the driver which has other dynamics also modelled in the game ...

I know this is a difficult approach but it makes more sense to me than over/undermodelling the car we are trying to get the right feeling of on the track ...

You're right. The terms under/over modelling could be misleading. That's why I prefered "alter".

Driving aids could solve your problem, but only if they are numerous enough. EA did a good job by including a larger number of aids, but still the car dances all around the track. On the other hand, the aids GP4 has for the players are only the essential ones. The only one that makes a great difference is Traction Control. And with all these the car behaves smoothly.

Can you please detail what you consider to be the better way.



maniak

Oh, my public! How they love me! (Looney Tunes)

One way that the GPx series compensates for the inevitable lack of feel without affecting the physics model is through added sound effects. Just as you start losing grip in a turn, your tires start screeching. That feedback is immediate and gives you a warning before it becomes visually noticeable. Obviously, an F1 car does not screech on impending loss of grip but it is an extremely effective and acceptable alternate sensory input that does not require screwing up the physics model to compensate for lack of feel. That's also why I always turn car skid volume to maximum in the sound options.

Good thinking, but tell me again, why don't F1 cars screech when loosing grip?



maniak

Oh, my public! How they love me! (Looney Tunes)

Why would they? :-p

Or maybe they do but very softly. :-D

LoL))



maniak

Oh, my public! How they love me! (Looney Tunes)

Actually, they do screech but at such a high pitch that only dogs can hear. That gives MS an advantage over his rivals.

Sorry, I just couldn't resist the temptation. :-p

Give dogs some credit....they can lap quicker than MS

what about F1 Racing Championship psyhics?!
ok ok you all are here for gp4 but.....try bouth of them..... and see the difference.
make a setup with asymetrical setting in f1RC an Gp4.....
When i did that F1RC turned to left slightely a slow speed and awesome at high speed.
try try....



\\Visit RenaultF1 Team homepage//
\\[www.renaultf1.com]//
\\ 100% RENAULT //

In the post that started this thread Maniak said:

What do you think guys? Is f12k2 more realistic than gp4 or not? Please don't come with "fanboy" (thanks to Morbid for the poetic license) stuff. Try to be fair and rational.

Well since then the next 36 threads have become extremely highbrow thanks in some part to our resident highly educated trainee Philosopher and Historian, Morbid. (no offence intended)

I just wanted to get back to Maniak's point with a cutting from a post I made last week:

QUOTE:
However despite all the slagging off by GPx diehards here I think F1 2002 is a VERY good game.

1 It works AND FAST out of the box.

2 FF effects 500% better

3 Car NOT ON RAILS - under/over steer, slides and shudders sideways just like I imagine a real F1 would - handles just like my Cosworth (when I could still drive it) GP4 handles like Scalextric IMHO

Yes it has things that are not so good but it hasn't got over 100 bugs - I been playing it for a week - not seen 1 yet. Graphics subjective - some like some don't but it is NOT bad and it motors

SOUND on F12002 I believe SUCKS and tracks too wide but overall IMHO it is A BETTER F1 GAMING EXPERIENCE.

Now I WAS a GPx DIEHARD - but GP4 killed me off and I bought F12002 - Always hated EA Sports games when I have tried the demos before BUT I think they have cracked it and are winning the F1 war."
END QUOTE.


Earlier guys were talking about bum-o-meters to feel the car. Well I think the best we have is a hand/wrist-o-meter (FF Wheel) and IMHO GPx games have the WORST FF engine of any driving game since GP2 (exactly - no FF support) and Test Drive x. If all you get is slight kerb rumble and 1 jerk when you "hit the tyrewall" what is the point.

Basically it is a waste of cpu power because it gives you no useful feedback AT ALL

In F12002 the feedback IS IMHO the best I have ever experienced in any game/sim since GPL and is better than GPL but then it should be it is 2 or 3 years newer. Why GC cannot learn from others like Papyrus and EA Sports I don't understand. Do what the Japanese do - take it apart find out how it is done and do it better. Do it like GC improving your product without looking at the competition you get left behind like British Leyland / Rover did.

I have NOT driven or been a passenger in an F1 car BUT I HAVE driven Formula Fords, Formula Vauxhalls and various racing saloon cars, (at Brands and Thruxton Racing Schools several times), also Mk1 Escort Rally car, Skoda Rally car, Mk1 Lotus Cortina and Mk1 GT Cortina Rally cars (mine in the 70's), Sierra Cosworth road car (REAL COSSIE which I still own but can't drive due to spinal injury and inability to pump clutch any more - not the "noncy" Sapphire) and a racing Truck at Brands (Now that was an enlightening experience - acceleration and handling was outstanding !) So I do have some experience with the "real thing"

So if we havn't got a bum-o-meter nor a g-force-o-meter to "feel" then we must have feel through the steering column and GPx just don't go there still. EVEN NOW in the year 2002 and FF has been out what 5 years ?

F1 2002 DOES - you even feel the shudder of those big slicks as they slip sideways when you get it wrong. The FF effects also give you the chance to correct the over/understeer or even induce them if required, and this then gets the "sawing the wheel back and forth" effect.

I think because GP4 doesn't reproduce these effects "if you get it (reasonably) right" you turn in and exit "on rails" a- la Scalextric. If you "get it wrong" you cannot correct the slide or oversteer and as I read somewhere else "you go off in a canned spin" i.e. once the sim decides you have got it wrong you cannot correct it. Opposite lock or more/less gas or brake changes nothing - YOU'RE OFF THE ISLAND in BLACK FLAG country (another badly implemented idea IMHO - as good as Refs in South Korea Matches at making decisions)

I think because F12002 DOES give you enough feedback to know what attitude the car is taking and which wheel(s) have/have not got grip it does accept input to change this - if you have oversteer you steer in and it corrects, too much and you start "sawing" get it wrong and you still go off (but don't get slowed by a stupid black flag)

My conclusion is (leaving aside graphics and tracks etc - discussed at length elsewhere) F12002 is a much more accurate and enjoyable driving experience than GP4.

I think GP4 is losing the plot - living on it's earlier laurels in the cause of Corporate $$$ and NOT really improving what we all want i.e. the simulation elements. Yes I like it to look good (and GP4 DOES THIS VERY WELL) but if it drives like Scalextric I lose interest.

Well sorry about the verbal diarhoea but that's my opinions and I have been rational.



PIII 1.0GHz Abit SA6-R MSI GEForce3 512MB SBLive5.1 Win98SE/XPPRO dual boot Mirrored IBM 7200rpm UDMA100 disks Guillemot Ferrari FF Wheel

Quote: "Well since then the next 36 threads have become extremely highbrow thanks in some part to our resident highly educated trainee Philosopher and Historian, Morbid. (no offence intended)"

Plenty taken.

I strongly, nay violently reject your placement of cause. In pure volume of text - half of my original post, and the whole point of it and my follow up posts was actually the opposite. To show the folly of the highbrow approach to discussions like these.

And I never slagged off any of the games. I slagged off the people that slag off other people, just because they don't play the same game.

---

It's only after we've lost everything, that we are free to do anything.