The complete "simulation" aspect is very arbitrary and subjective. While some will praise one game for its physics and nothing else will be important to them, others will highlight overall immersion of some other game making attention to details more important than just physics. That debate can never be solved.
What any driving game in the current state of time need to do is try to find a way to cater all types of public. Gran Turismo, Forza Motorsport, pCARS.. all driving games bring different people from around the world with different views on all imaginable fields of what constitutes a perfect driving game.
That is the first point where game developer faces with the first "prisoner's dilemma": in order to make game accessible - thus making more money from potential sales - game needs to be approachable to as many players possible.
On the other hand, plethora of us here on GTPlanet - and other communities - are vocal in intention to GT series provide us with everything we ever wanted from some driving game. And there we have a problem.
I was writing about this particular issue more than a two years ago here on GTPlanet (regarding then incoming Shift 2) and later on pCARS WMD-development forum. I call this point in time the "Paradox Point".
The overall driving genre has come to the Paradox Point and it currently represents the main problem of the overall genre - as far as development and focusing the potential players/buyers is concerned - and in future it will be even more and more evident. Of course, the Paradox Point exists only for titles that tends to become popular to the wider-demographics and to exist on consoles, it is a non-issue for PC-specific *hard-corest* simulations, so they're excluded from the following breakdown.
Console-specific driving games also have something that PC simulations do not have: unlockable structure of the in-game progress and various collectibles (either prize cars, credits, XP system, whatever) as an prize for that same progress. Both of those are also very important for the second part of the Paradox Point.
On one side, you have something that no other genre has in cumulative: ability to portray the essence of your genre up to almost being "real" experience. Force feedback technology, processor power for graphics/sounds/AI and online racing pretty much gives unprecedented package where adequate implementation of "real" details can produce most immersive and accurate experience of all genres.
On the other side you have market, where people who actually wants to have "real" experience are distant minority when numbers needed to make the title profitable are in context.
As a developer / publisher I have all tools and tech on my disposal to make perfect simulation. But who will actually drive it? How can I make it accessible to majority/casuals? Unfortunately, the answer is not "make a scalable difficulty". Why? Because of the psychology of the man.
That is the exact place of birth of the Driving Game Paradox Point.
If you make game with option of "Normal Physics" and "Real Physics", you have a problem. Average Joe wants to play it on "Real". He doesn't want to play it on "Normal". Because Average Joe knows he's the best driver out there. He's the "Real Deal Joe". Then Average Joe dives into "Real" and in first corner he realise he's not Real Deal Joe. Average Joe turns furious and shovels the game. Because that game can't give him the feel of being the Real Deal. He doesn't want to drive on "Normal" or "Standard" or whatever.
And there is the Paradox - you can make the game real as "Real", but making it so will alienate majority of players. If your game become popular, you will have high expectations form the future buyers. But once the game releases and they realise how they are not able to play that game without someone holding their hand (assists, etc..), you will 100% suffer from the community backslash on various forums - and never forget how vocal minority is the most important factor unfortunately - thus hurting but current sales and potential sales of the next title.
And it even goes complex than pure simulation aspects of the physics, it also goes deep into structure of potential rewarding.
That problem is evident on example of the Gran Turismo 5 Seasonal Events (specific online events made for earning loads of XP/CR/prizes). When Seasonal Events were initially introduced, they came with great HP/weight/tire-type restrictions. However, the casual majority of players were shouting their mouths about Events being "hard", "impossible", "rigged", "bugged", "broken", etc. - usual internet whine. But unfortunately, those casual majority is the main population that plays the game. Also, that same casual majority is the main demographic that will determine the future success of some game on all platforms with exception of PC.
Two weeks later when original Seasonal Events were introduced, the restrictions were gone. Praising of casuals begun big-time (fueled with exploiting the dupe-system to Jupiter) while hard-core population suddenly felt cheated. Than came the proposal "make Events scalable - give less XP/Cr for usage of assists and driving without restrictions, but award those who drive without assists and comply the restrictions". And that is the second place of Paradox Point.
If you reward the hard-core more, you again make casuals to feel incompetent. And again you're making yourself a problem because Average Joe doesn't want to get only 200,000 Cr of his drive to 6th position. He wants full 1.500,000 Cr, Gold Trophy, 1st place everywhere and feeling he's Real Deal Joe. If the game can't provide him with the appropriate feeling of being competent, he will almost certainly not buy the next instalment. Which is a problem, real problem as long as long-term success of some IP is concerned.
You can now argue above with highlighting FM4 system for example, where usage of assists leads to removal of Cr/XP bonuses - but since the actual physics model was catered to favor the controller - and all game modes are united without option to filter anything - actual Paradox Point is removed due to catering all types of players under the same umbrella. However, the more serious players are then somewhat betrayed because they can't be competitive among each-others due to (probably well-planned for the reasons of making the game more approachable) lack of filtering based on assists used, input device used (although game recognises different input devices), etc.
I see no easy way to resolve the Paradox Point as long as game wants to cater all types of players. Introducing a plethora of driving assists is a nice way for start, but in order to have the equal amount of challenge for all types of players you need to introduce a very detailed filtering. However, more filtering would result with splitting of the userbase (depending of the game structure of course) and inevitable "Real Deal Joe" problems (just take a look at the GTPlanet at threads discussing usage of assists for example). And Paradox Point is there.
You can split the physics at the start - GT5: Prologue had that for example, where simplified physics was called "Normal" and full-physics was called "Professional" - but than you have to build the overall driving experience and game-design around that (testing of particular challenges with both physics, building the online options with that in mind, etc.). Again, you're splitting the userbase. FM4 dodged that bullet by unifying both physics type (Standard and Simulation) under the same umbrella (only determination factor in FM4 is the lap-time being "clean"), but that alienated the more serious players, especially those who are using both Simulation physics and the wheel. Again, Paradox Point.
Current state of gaming in total has abandoned any traces of old-school where difficulty and skill were determining the actual accomplishment. In my driving community we have numerous discussions about what we want from driving games but we're aware that times where needs of serious players were important are long gone.
To conclude without going into further explanations, I think GT series finds a good way to solve the Paradox Point with smart implementation of game-design choices. During my "Paradox Point Theory" development I've spent numerous hours in playing all driving games I have in my collection in order to determine differences in approach to "carrot and stick" problem and complexity that arises from Paradox Point - especially how it was done in multi-million selling franchises that somehow managed to break the Paradox barrier.
Path GT series is doing it is smart, strategic and very multi-level. I am very certain it will continue to be that way because of the sheer imagination that reflects in very different ways Polyphony is approaching the meta-development of Gran Turismo existence.
).
- PD still finding that balance 
Hopefully GT6 will please all of us.
.
