Reader Submission #146 – The (dis)Creation of a Sim Car

Are developers claims of using “real data” to build their sim cars merely hot air? That’s the question today’s Reader Submission seeks to answer, as an amateur rFactor modder behind the wonderful CART88 package from the Historic Sim Racing Organization has outlined a few different situations from mainstream sim developers that have left him scratching his head. With cars shipping with incorrect differential models, or popular prototypes turning lap times six seconds faster than the real vehicles, it’s a bit hard to have faith that the marketing department is on the same page as the physics gurus in charge of each respective development team. Compounding the issue is the fact that some of the information that may lead to a more accurately produced sim car is actually available for the general public to consume, so it’s a bit strange that major teams are unable to visit Amazon.com and order a $20 book in the pursuit of authenticity, when passionate modders have no problem doing so for the sheer love of sim racing.

Hello avid PRC readers, it’s your favorite cynical modder again!

As you all know, I am responsible for the CART88 project in terms of most of its physics development, as well as sounds. Now, first of all, the Automobilista conversion is still coming along, don’t get desperate! In fact, I am writing to the sim community here today to tell you how creating a mod or a car is not as clear cut and as clean as the modders, and even major sim developers might tell you; research, time, testing, and critical thinking are oh so very important. This submission comes in the sequence of two separate events that shocked me, considering we are talking about developers that made you pay for these cars.

The first portion comes from basically how it seems even official developers fail badly on the “research” part. Every, and I do mean EVERY 1960’s or 1970’s Formula One mod, from the vanilla content of Grand Prix Legends through all the iterations of the Lotus 49 in every simulator, to Niels’ Formula Vintage, is wrong. And what’s more shocking is not the fact that they are wrong, it’s instead how simple it would have been to avoid this particular error. The differential is an often overlooked component in sim racing, but when we have a car without big aero or tire grip, its role on the general car behavior can be huge. The approach everybody had, was that those cars were supposed to come with a highly adjustable Salisbury type differential variations, or simplifications of this being implemented in other sims, like rFactor.

Now what if I told you that those cars never used this differential? Well this is the truth. Back then they were all using a Clam and Paw type, which was a very simple type that was first used in the Auto Union cars of the 1930’s Grand Prix Era. This is a non-adjustable differential that is basically completely open on Coast, and locks from 50% to 75% on power. Now how do I know this amazing and crucial piece of info? Simple, I got it from a book about the Lotus 72 that sells on Amazon for 20 Euros or less. Hardly top secret or hard to get knowledge, especially for developers who supposedly have access to “real data” and other such buzzwords.

The second event that shocked me, was seeing lap times six seconds faster than the real thing coming from iRacing’s Nissan GTP around Road America. Now, we could forgive this in a non-laser scanned variant of Elkhart Lake, but in the pristine Road America iRacing has, it’s unacceptable. But what really rustles my jimmies is not that they got it wrong. After all, it’s really hard to get lap times right across many different circuits, especially if you don’t have absolute precise data from the vehicle, and assuming your physics engine does everything right. Which it probably won’t. What gets me going is how they just ignore it and don’t fix it.

I am telling you this now; a mod, a car, any piece of content the community creates, is typically not finished with release version 1.0. Modders don’t have an army of testers to go around running all the circuits and report back competitive times, or trying to break the mod with exploits. That’s why the CART88 mod is still receiving patches, because I am effectively using the current CART88 season in HSO to also test the mod in a competitive environment, which is how it should be done. This assures the end product is actually bullet proof and performs as much as the real thing in all possible scenarios, if driven to its limits.I noticed the times are off at, for example, Mid-Ohio, so expect a big patch to rectify this, based on my observations, and on me questioning some of the data I researched earlier. But this is ME doing a whole mod for free! How can people at iRacing, Kunos, SimBin, Slightly Mad Studios, etc, sleep at night knowing that they got things so wrong, and they don’t do anything to fix them, despite the users paying for this?

You see, a basic tenure of simulating a race car is what kind of apex speeds, top speeds, and braking distances a car achieves. This in a laser scanned track should at least amount to a lap time more or less within the ballpark. Sure, thousands of laps of practice in a perfect virtual track will usually allow a slightly faster lap time, but not six seconds’ worth. I find it funny the whole “Formula One 2017 is simcade” debate, when the “serious” sims can’t even get something as basic as these elements correct, and worst of all, are unable to admit that they got it wrong and begin working on a fix. So much for “listening to the needs of the community…”

Unfortunately, I think a lot of it has to do with developers knowing that the community will make excuses for their sloppiness, which is partially why simulators of the early 2000’s are objectively more accurate that the software on the market today. Most developers, and this extends all the way from Kunos, to iRacing, to Slightly Mad Studios, are all extremely passionate about sim racing in the same way that we are, which is why they still continue to push out software and content for what is a very niche genre that rarely generates a reasonable return on investment. However, because each game’s respective community will now actively work as an extension of the marketing department – whether this was intended or not – the incentive for these teams to exhibit precision and accuracy in their work just isn’t there anymore. If you’re a sim developer and you knowingly half-ass a car, only to boot up the message boards and find people still praising your work and calling it one of the best sim cars ever, subconsciously this is going to re-wire your work ethic a bit.

This is also a community where the majority of participants have zero mechanical knowledge, and in some cases don’t even possess a valid driver’s license, so for every sim racer picking apart inaccuracies in a sim car (or giving it the thumbs up), there are at least twenty five others on the forum counteracting the useful feedback with outright disinformation. This right here is actually the source of iRacing’s endless tire woes, as back in 2011 the original variant of the New Tire Model generated rave reviews from the amateur race car drivers on the service, but this positive feedback was outweighed by teenagers and bus riders in the Peak Anti-Freeze Series complaining that their unrealistic setups and driving styles no longer worked. Because the volume of complaints outweighed the number of amateur drivers kicking ass on the service and actually enjoying the brand of racing, take a guess who iRacing listened to.

Unfortunately, the only option to rectify this is to either keep building third party mods that do pay close attention to detail, or venturing down another route and tweaking vanilla content. This isn’t really possible in iRacing because it’ll get your ass banned in a hurry for a legitimate reason, but given that it’s not hard to find the ACD converters for Assetto Corsa stuff, nor is it difficult to use unpacking software for Slightly Mad Studios content, I envision a future where mod teams specialize in “Community Patches” for first-party content, in which a team like Kunos would release a DLC car pack, only for the mod team to come out with a “Redux” patch that fixes some of their flubs in terms of suspension geometry or tire behavior. It’s not ideal, but this is more or less what’s happened with Richard Burns Rally and Need for Speed: Shift 2 Unleashed, both of which spawned pretty big add-on communities centered around a few groups of guys essentially digging through the internals of each simulator and fixing the issues.

 

Advertisements

68 thoughts on “Reader Submission #146 – The (dis)Creation of a Sim Car

  1. Well Bethesda has been doing the whole “here’s sort-of-a-game and we’ll just let the community fix it” thing for a good ten years and people keep eating it up. Guess its a proven model on PC at this point!

    Like

              1. Speaking of italians,Ferrari F355 Challenge on Dreamcast was the first time i saw Sepang in a video game and even back then (1999) i had a mixed feelings for this track, Exotic but at the same time kind bland surroundings,very wide track with devilish turns. This one from Raceroom looks way better of course and more importantly without a ton of post effects a trend that plague games & sims these days. however imo it still feels a bit game-ish hopefully we are going to see some improvements on the graphical front with the upcoming move to Unreal Engine 4.

                Like

                1. I must say that driving at Long Beach and at Sugo this week made me think of F355 once again. I played it so much. It was kinda hardcore in its own way, forcing you to drive from the driver’s seat and managing the shifting somewhat. If you substitute Atlanta with Texas, Project CARS 2 is only missing Sepang.

                  I think Sepang made an impression on me because of the faster sweeping bends that you can not get right on your first few laps, and it’s always an ordeal with each new car. The last time I played R3E, I was using an alternate MOOD file that changed the graphical style a bit. Maybe look into that?

                  Like

  2. Project CARS 2 also has the Nissan GTP, and in this video its qualifying time for the 1990 Daytona 24 is beaten by almost two seconds. He seems like a bad driver, so getting another four seconds out of it should be very easy. Odd that you didn’t mention the Project CARS 2 version, since you posted a screenshot of your Nordschleife time in it today on 4chan.

    1989: 1:39.182
    1990: 1:37.963
    Some guy: 1:36.120

    Like

    1. Well, there’s hope. A (former ???) SMS employee just wrote “…nor is it difficult to use unpacking software for Slightly Mad Studios content, I envision a future where mod teams specialize in ‘Community Patches’ for first-party content”.

      Like

    2. Racing sims should always produce slightly higher laptimes in real life due to the lack of fear from dying or serious injury that real life has, even in the best drivers.

      Like

        1. Yeah those pussy GTLM drivers, only one of them was able to qualify with a 1:21 at Laguna Seca last weekend, if only they had the balls of the average iRacing driver they would be able to beat the Prototypes with 1:16 laptimes.

          Like

      1. Yes that’s the case.. but sim-racers and armchair experts like the guy above who just wants to piss on pCars 2 never take details into account. Kind of ironic considering the actual article in question.

        Like

    3. May I be damned for defending the trash that is pCARS2, but in 1989-90 the layout was different compared to the modern one.

      In the ’90s the chicane on the back straight was narrower, longer and much slower (essentially a double chicane), the one that is in use today is not only shorter but also wider and faster, which means that 25+ years ago the cars needed to brake hard for the first chicane, briefly accellerate, then brake again for the second chicane and then go back to the banking while nowadays the cars need to brake less and can accelerate faster and sooner.

      http://racingcircuits.info/north-america/usa/daytona.html

      It might not look like much, but that difference alone is enough to justify a difference in laptimes with the modern track being faster than the 90s one.

      Like

    4. Not defending iRacing here but remember back in the late 80s and early 90s, at Road America, turns like T1 and Canada corner didn’t have the huge exit curbs that they do now. Back then, you couldn’t carry as much speed as you didn’t have any of these big run-offs/exits that exist now. Not 6 seconds worth but still.

      Like

  3. “As you all know, I am responsible for the CART88 project”

    No, you`re not that important. I don`t know who the fuck you are and i couldn`t care less….

    Like

  4. well shit, i could tell you most sims do fuck all in terms of diff simulation

    >drive AC miata with open diff
    >WOT dump clutch and turn
    >inside wheel spins and you make a slow circle

    >drive real miata
    >WOT and dump clutch
    >donuts for days, both wheels spin, glorious counter steer

    yes on paper an open diff will send it’s torque to the wheel with the least grip, but in the real world it’s not that simple.

    Like

    1. In 2nd gear you can drift around, but depending on grip level and angle at one point you end up losing the slide and the inside just fires up. In 1st gear, forget it.

      Guess you drove a Torsen car.

      That said, AC’s Miata does a few things a bit wrong… Balance isn’t right on and off throttle, and it’s a bit slow, especially uphill. Doesn’t have enough pick up for a 1.8L, feels more like a 1.6.

      Like

  5. I’ve posted here before about this since I work with Aircraft simulation development for level D trainers. This means that the models have to be super precise regarding reality.

    There is a right methodology to do that and I highly doubt that those “sim” developers do that

    It really pisses me off reading about use of real data. If they have real data and develop their simulations based on that, than show me the proof of match.

    It’s not simply plotting sim data x real data and saying “oh, it matches for 95% of the time”. This is not how proof of match works. Specific tests are needed with absurdly tight engineering tolerances do create a robust proof of match. And then you need to run cases not used during the pom to show the simulation robustness.

    And of course they don’t do that. Otherwise, we wouldn’t have those absurd discrepancies between real drivers and iron table toy steeering pretend drivers. I would expect the latter to be slightly quick after a huge amount of training and experience, not whooping fast right out of the box.

    It’s better to just admit that, despite using real data, you probably don’t have the right set of tests to match your model.

    Liked by 1 person

    1. If DCS can offer a bunch of highly-simulated aircraft at a premium, what’s stopping a small studio of just recreating one car and selling it for various simulators? And I would gladly buy a one-car one-make series simulator if it’s genuine and accurate.

      Of course, like James said in his review, competing with Forza makes you cut corners to add more stuff, and it’s not like the sim racing community is concerned with realism in the first place. As for the circuits, it’s time this shit becomes modular, because what’s the point in having 5 games with the same 3D version of a track made with the same laser data?

      Like

        1. I have a bunch of their stuff for rF2 and AC, but they were never satisfying enough for me. Funny enough, I think their best content is the free Indy Lights thing they give away to owners of other packs.

          Still, I was glad to support them, because they offered types of experiences that are/were not in the main product.

          Like

      1. My guess is the ever evasive tire modeling. It would seem flight sims may be a bit easier to model aircraft as CFD is widely used and available. Not only do vehicles need CFD but super dynamic tire modeling on top of it. Seems pretty difficult but I may be talking out of my ass.

        Like

    2. Oddly enough, tire physics outside of the tiny grip envelope are even less well known than aerodynamics. There just isn’t data studying what tires do over the limit. Usually you’ll get slip angle up to ~10 degrees, slip ratio up to ~20% and that’s it, recording ends. Lucky if you get data for anything other than 0 camber too.

      Like

  6. Project CARS 2 at least improved on that. They simulate different differential models unlike all other sims which simulate 2 or 3 at most.
    Some sims even cheat when it comes to rearwheels vs frontwheel vs all wheel drive. They just modify rearwheel drive so you think it would be frontwheel – iRacing is a good example.
    Sad truth, most sim devs don’t give a shit about quality, it’s all about quantity – some are worse in this area (IRacing), some are better (rFActor Pro), some we don’t exactly know or have inconsistent quality (Project CARS 2).

    Like

  7. About the GTP, my guess is that basically every sims have too much tyre grip under high loads, so when you have high downforce levels like these cars had, you end up having much shorter braking distances and too high cornering speeds. So often devs lower downforce levels, like in AC, as a band-aid.

    Would easily explain the huge laptime gaps, extreme cornering and super short braking distances the iR car has (and PCars’ apparently).

    Too little grip at lower loads, too much grip at high loads, sounds familiar doesn’t it ? That’s on top of the over the limit behavior being wrong everywhere. Changes dynamic a bit.

    Liked by 1 person

        1. Fixing things? Nah.

          I was thinking about the ZX Turbo, it’s planted even at slower speeds. In general, Project CARS 2 doesn’t feel that slippery at low speeds, nothing like rFactor 2 for example.

          Like

    1. Armchair simracing analysts always think its as simple as turning the “grip” variable down and tuning the “downforce” variable up. The number of variables and calculations that are combined to simulate any of things are far more complex than your peanut brain will ever be capable of understanding.

      Like

      1. It’s not really a secret that every simulation has fundamental limits on physics (due to the frequency of the physics engine, the discrete integration methods used) which makes simulating very stiff cars a problem. The devs know where these limits are, they compromise between ‘type in the correct number and get physics explosion’ and ‘type in different number that ends up playable’

        In general, once this kinda thing is set, it’s there for the life of the game (AC: 333Hz, iRacing: 360Hz, RF2: 400Hz, AMS: 500Hz), because the devs would get worse feedback from making the game unplayable on old hardware (which at launch worked ok), than from improving a few cars in ways a handful of players would have the knowledge to detect.

        Modders don’t have to take this kind of thing into account, because their content’s held to a different standard – if they release a car that simply will not run on a GTX660, even though the game’s min spec said GTX640, then that’s just the way it is. If they ignore physics limitations and type in stats for a car that self-destructs when it hits sharp bumps, that’s just the way it is.

        Like

        1. Of course it’s not about “lowering 3 grip” and “adding 5 downforce”, I’m not fucking stupid.

          Doesnt mean most sims havent figured it out, I don’t care about the reason(s).

          Like

  8. OR, games could offer training and feedback on how to drive and setup cars. Then the user base wouldn’t incorrectly complain about things which are realistic and correct.

    Like

  9. enjoyed this article, this is the sort of crit i like, evidence based & interesting although screenshot of said paragraph stating diff in the 49 would’ve been good.

    Like

    1. I’d love to see the same but I’ve been looking for years and not found it. When I was working on the 65 Mod for GPL I discovered the 65 F1 cars used the cam and pawl diff. Then when working on the 69 mod I discovered the later cars still used the cam and pawl diff. This strongly suggests the 67 Lotus 49 also used the cam and pawl, but the only 67 car I found direct evidence of using that diff was the Ferrari 312. I should also state that I never found any evidence that the 67 F1 cars used the salisbury diff.

      Like

  10. I have that book, there’s several instances where the cam and pawl is mentioned but specifically I can indicate page 103 where Eddie Dennis, Lotus F1 mechanic at the time, mentions that the cam and pawl differential wasn’t tuneable at the track, but required careful rebuilds at the shop -Which is a know treat of this self-destructing unit that wears out during the races, the more so in a powerful car.

    Like

  11. Cam & Pawl (not Clam & Paw) aka the Jack Knight diffs to those around in the late 50’s are not particularly fragile but they do wear quickly. A friend of mine had a Mini Cooper S with one in the 80’s and he’d rebuild it every couple of months of hard street use, but it never broke.

    Like

  12. Not to be pedantic, but the sentence “a basic tenure of simulating a race car…” should instead read “a basic tenet of simulating a race car”.

    Tenet = a principle or belief.
    Tenure = the holding of an office.

    Other than that, I continue to have a real problem with using lap times as the primary metric for verisimilitude. Why?

    Because Tires.

    Tires contribute more than any other single factor to lap times. And it’s nearly impossible to hold that variable constant when comparing Real vs Simulated. Real Tire Data is a closely guarded secret.

    You need to pay serious cash to get that data, and even then it would be for only a specific tire or set of tires. Try calling up Pirelli to find out what they charge: I can almost guarantee it would exceed the development budget for Assetto Corsa.

    In any case, there’s only one metric that really matters: Skill Translation. Do the skills of a real race driver translate seamlessly to driving the simulator? Can a real driver just “jump in” the simulator and be quick? This is all that matters in the end.

    Liked by 1 person

    1. I take lap times a bit more seriously now with all the laser-scanned circuits around, but when it’s used to justify the F1 games or Forza, the argument in the favour of that metric disintegrates like Kenny Bräck’s car at Texas.

      Like

      1. That’s correct. I can create a ridiculously inaccurate simulated car, then endow it with, say, tires that are way too grippy – and Bam I’ve got the “accurate” lap time time of the car IRL.

        Like

Ratio of vowels to consonants will be monitored. Post at your own discretion.

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s