Exaggerated crash physics are far from exclusive to Automobilista 2. What I hope to highlight is the ease that regular contact between vehicles can lead to exaggerated crashes. For example, in AMS1, I have to go out of my way to force an opponent's wheels to leave the ground. In AMS2 with the same car, a basic rear-end collision dooms them to lift. And it's not caused by the AJR's slanted front-end. I was able to recreate this in '86 Opala's before reaching Turn 1. It's hard to describe, but currently, there's this feeling that cars are simultaneously too light upon being hit, and far too heavy when hitting. Obviously, this is unrealistic, but more importantly, it negatively impacts the gameplay's enjoyment. Close wheel to wheel racing is discouraged, small touches can morph into race-ending crashes, netcode errors are exacerbated, disciplines like Oval & Rallycross are more vulnerable, the realistic damage model's immersion gets diminished by bouncing cars, it gives a lot power to cause incidents (accidental or intentional), etc. I'm a total layman, I have no idea what the first steps would be to addressing this or its complexity. I've just seen this topic pop up over the years and wanted to give it a dedicated post, and do believe that improving the contact physics would greatly help in polishing Automobilista 2. Making it both more realistic and enjoyable.
The physics of a car nudging the other (turning into a corner) isn't realistic as well. When going into a right turn, for example, and being nudged at the left side at the back, will not turn right (altough there is nothing blocking it at the right side). The same goes for turning left and being nudged at the right side. I hope this behaviour can be improved to more realistic physics.
AMS2 has a huge problem in this area, and it’s the first thing that needs to be addressed (physics-related). The resulting vector of a collision is not calculated correctly, and/or the effect on the vehicle is not proportional to the collision forces. I’ve been hammering this point for a long time because, as it stands, this is the single biggest immersion killer in the sim.
If I had to make a guess is that it is probably not easy to fix, likely combined with not being a priority. The worst combination when you'd really like something to be fixed. Also, with time, it disappears from the radar and people forget about it because there is something always more urgent to do. A bit like the proverbial bug that is there since forever, got forgotten and is never fixed.
You put it very politely. but if we’re right, that’s a big problem for AMS2’s future. It’s reached a point where I can’t deal with it anymore - perhaps it’s time to move on. It's a sim, it needs to get the basics right or not call itself a sim. Just watch the red car (4 cars ahead) in this 22yo free game. That's 'right' from my perspective, both the collition and car physics.
No sim is perfect. The physics have to be heavily faked to be able to run on household PCs. If the complaints about the physics fakeness surface mostly around the collisions then it's not a bad state of things And I'm not saying everything else in AMS2 is peachy but it reached the nice state of reasonable believability in most areas.
I would say that, a simpler but important improvement might be at least the following - at the moment if I get hit by the AI, it basically is able to return to the circuit IMMEDIATELY; but this is something that is not really realistic in my opinion. In modern racing you often see a car go off, take a few seconds (or more!) for the driver to get their bearings, get into gear, etc. Adding something like this would be very immersive for what I _hope_ would be a simple-ish effect in the AI. It at least means that, if I am hit by the AI, they don't immediately return to their place like they often seem to now.
That’s why I mentioned a 22-year-old game doing it right (check out those rear bump physics and compare them to AMS2). Those were the days of the Pentium III @ 1.13 GHz (<30W TDP ). I disagree, because this completely turns me off. I’m not talking about refinement or nuance, I’m talking about outright impossible physics (e.g., two cars colliding on a perpendicular trajectory and one of them simply staying at the crash site). Along with my second-worst aspect of AMS2 (see @wegreenall post above, Super Mario Kart territory), these are real immersion killers for me.
You've mentioned the "basics" to be right but the coded physics have completely different "basics", and I'm not really sure there are physics basics in the simulation code. Some aspects may be closer to reality at the expense of something else (i.e. better contact physics could break handling calculations at grip loss or cause unnecessary calculations of the unexpected trajectory changes on impact). All I'm trying to say is that when contact physics are off, but the rest is pretty good, then that's a compromise I personally see no big deal with - it promotes clean driving We're all different - it is an immersion killer to you, it isn't for me. It's not 100% accurate but to me it's believable enough in most cases. It doesn't mean that one of us is wrong though. We can both have different opinions as what is wrong and what isn't. I'm not denying that you cannot feel disappointed. I'm only trying to shed some potential light (I'm not a Madness Engine developer but have enough software development experience to make reasonably educated guesses) that some things that were done before in 22 year old games (and LFS and its physics are still pretty good today!) may be very tricky to replicate on some engines where there was a decision made early on to "fake" contact physics a lot more to benefit something else in the physics department. I suppose the Madness Engine name may not have been accidental And I agree, better contact physics will be welcomed by myself too - but I get that from the development perspective it may be a lot trickier than it sounds.
Don't forget the Madness engine goes back at least as far as NFS Shift (2009), and Project CARS 1 was originally hoping to release on PS3. Back in those days the processing budget had to be used wisely, so from what I remember from WMD days, compromises were designed into the engine so that it prioritised the tyre physics at the expense of things like collision physics and replay fidelity, and you can still see evidence of that even today. Nothing is impossible to change and rewrite as Renato has said many times, but what we are discussing in this thread is quite deeply embedded in the game engine and would probably be a poor return on the investment of time needed to extract it and change it to what Reiza might ideally prefer. Collision physics I think is one of those legacy issues that existed originally for good reasons, but would take an inordinate amount of work to dig out and update, so something else would have to give. If you were writing an engine from scratch with 2025 hardware in mind it might be different, but the core Madness Engine is of a similar vintage to Live for Speed which lacks many things the Madness Engine does have. So as said above, don't crash into other cars, just imagine you are paying the repair bills and some mother's son is in that opponent car. Honestly you'll be happier for it
Originally, I had a paragraph about the Madness Engine but deleted it for time. You mention however it's potential presence. While it's very true 7th-gen consoles were waaaaay more limited in hardware capability, older sims on even less capable hardware have more accurate collisions; I think there's another explanation. NFS: Shift (the first game to use this engine) rewarded XP for shoving people out of the way to overtake. I think this better explains the collision physics. The player's vehicle be heavy when hitting and opponents light upon being hit, makes complete sense with this design.
By basics, I mean basic tyre behaviour, basic mass/inertia behaviour, and basic collisions. Then, it should be further refined. Just watch the accelerometer on the telemetry HUD... Is it okay? Yes, so long as you don’t count walls as part of the virtual world. A sim cannot be accurate if a wall is not treated as an almost infinite-mass, nearly undeformable object. This accelerometer is simply wrong when a wall is involved, and it’s not just a visual issue, it’s deeply rooted in the game’s physics. Madness Engine is very strong on the graphics side, but this is a sim. Physics are the foundation; visuals come later (lower on the priority list). The original developers seemed to think otherwise. I’ll take the “...then do not crash...” argument either as a joke or as evidence - I haven’t yet decided. Moderator: Sorry if what I said came across as offensive; that was not my intention.
The "do not crash" statement was clearly meant as a joke... But in a in a racing "game" we do have to live with compromises.
Yeah my 'don't crash' statement was also only semi serious, I mean obviously avoiding crashes is good racecraft, but crashes are also part of racing. I remember in the development days of PC1 lobbying for Wreckfest standard deformable Armco and tyre barriers and car collision physics, but it was clear that all the processing power was going into the tyre model at that time. Maybe with another 15 years of PC hardware progress a new game engine might address this, but I think we have to accept some compromises sometimes.
I wonder if it could be quickly improved in a way that it is 80 or 90% satisfying vs what we have, without it being overly difficult to implement such improvements. One difficulty that I can see is that any collision improvements also have to work with MP.
