Feds Told Tesla To Stop Making 'Misleading Statements' On Model 3 Safety (arstechnica.com) 116
Last October, Tesla said that its Model 3 had the "lowest probability of injury of any vehicle ever tested by the National Highway Traffic Safety Administration (NHTSA)." Two days later, the agency responded without naming Tesla by saying that its "5-star rating is the highest safety rating a vehicle can achieve," but that it "does not distinguish safety performance beyond that rating, thus there is no 'safest' vehicle among those vehicles achieving 5-star ratings." Now, according to Ars Technica, citing documents obtained by the website Plainsite, we have learned that the NHTSA told Tesla to stop making "misleading statements" about the Model 3's safety, as the company's use of NHTSA's 5-star ratings and associated data "is inconsistent with the NHTSA's guidelines." From the report: The NHTSA conducts a number of different crash tests for each vehicle and then issues a series of ratings ranging from one to five stars for different aspects of vehicle safety. There's no disputing that the Model 3 performed well on these tests, achieving five stars -- the agency's highest rating -- across the board. The NHTSA would have liked Tesla to stop there. Instead, Tesla dug into the NHTSA's data and spotted an opportunity to further toot its own horn. As part of its evaluation process, the NHTSA calculates a number called a vehicle safety score, which the agency has characterized as "relative risk of injury." The agency then awards each vehicle a star rating based on VSS ranges.
Tesla noticed that the Model 3 had a better VSS score than any other vehicle on the market. That, in Tesla's view, means that a Model 3 driver is less likely to be injured in a crash than a driver of any other vehicle. But the NHTSA argues that this is statistical malpractice because it doesn't take into account vehicle weight. In a vehicle-to-vehicle crash, the occupant of the heavier vehicle is less likely to be injured. The NHTSA's tests, which involve crashing a car into fixed objects, don't necessarily account for this difference. Tesla says it stands by the statements it made last October.
Tesla noticed that the Model 3 had a better VSS score than any other vehicle on the market. That, in Tesla's view, means that a Model 3 driver is less likely to be injured in a crash than a driver of any other vehicle. But the NHTSA argues that this is statistical malpractice because it doesn't take into account vehicle weight. In a vehicle-to-vehicle crash, the occupant of the heavier vehicle is less likely to be injured. The NHTSA's tests, which involve crashing a car into fixed objects, don't necessarily account for this difference. Tesla says it stands by the statements it made last October.
Tesla results would mislead by weight category (Score:3, Interesting)
the occupant of the heavier vehicle is less likely to be injured.
I don't see how that would be true considering a Tesla and another vehicle anyway, because generally the reason behind that weight protection differential in the past has been the mass of surrounding car components like the frame or the engine.
With a Tesla a large about of the weight is in the battery, which is underneath driver and passengers and so I don't think adds much to your protection? Tesla does have all very good modern safety aspects like great crumple zones, but it's a different case than for other kinds of cars.
So in the end I do think it's fair to say you have to judge relative ability to withstand accidents more by actual crash results than things like weight consideration....
On a side note, I have wondered if Tesla's lack of an engine mass meant that it was better in a front end collision, because it could use the front end asa much larger crumple space, or was worse because it couldn't use the engine block as a shield. But that mass would also be sliding back towards the car occupants so probably lower mass is still better in front assuming the front end is well designed for crash absorption.
Re:Tesla results would mislead by weight category (Score:5, Insightful)
Re:Tesla results would mislead by weight category (Score:5, Informative)
That's exactly the point.
1) Tesla made no claims related to weight at all. By definition, if a vehicle has the lowest combined probability of injury - the very stat that the NHTSA uses to assess star ratings (and without further weight restrictions, I should add) - then it's the lowest in any weight category. Narrow down the list to any subset of vehicles, and the Model 3 will always have the lowest score in said subset.
2) The specific vehicle in question - the Model 3 LR RWD - weighs 3885lbs, while the NHTSA average vehicle was 4035lbs. So even if you ignore #1, it is within 250lbs of the average.
3) Not only are the NHTSA's own star categories based on the VSS score without respect to weight, but the NHTSA's own 2015 recommendations for NCAP revisions maintain the focus on VSS and made no new proposals to make it weight dependent.
It's ridiculous that the NHTSA tried to ban Tesla from talking about the fact that its combined probability of injury was assessed by the NHTSA to be less than half the requirement needed to get a 5-star rating. From mentioning the NHTSA's own public data, and in doing so in an accurate description of the NHTSA's evaluation metric. It's also IMHO ridiculous that the NHTSA thinks it's just fine to have its rating system stop at 5 stars, when 40% of all new cars get 5-star ratings. Great way to make all stalks of wheat equal by cutting down the tallest ones.
I'll add "tried", because the NHTSA sent its letter nearly a year ago (this was already public news back then, we're just rehashing it now), Tesla replied with a polite "go pound sand" letter, and nothing else came of it.
Re: (Score:1)
Teslas also are the most frequent cars to steer themselves into jersey barriers because the car manufacturer claims to autopilot the vehicle.
I get that you make money on Tesla shares, but the shilling is rather ridiculous.
Re: Tesla results would mislead by weight category (Score:2)
No, not by a long shot. The number of impacts to jersey barriers by Teslas are insignificant next to other cars. What metric you are looking for is number of impacts to
Jersey barriers that are news-worthy. Tesla has had so few on autopilot you could count them with your fingers.
Wrong again (Score:2)
Teslas also are the most frequent cars to steer themselves into jersey barriers
Nope. Teslas are the only car where if you let go of the wheel for an extended period you are probably not going to run into barriers. Just because it sometimes does does not mean existing cars that will certainly crash if you fall asleep driving them are vastly worse safety risks.
At any moment of intention in any other car, you are at great risk. In a Tesla if you feel something like a tight chest or other health issue coming
Re: HI REI! Sooo glad you're back! (Score:4, Informative)
Normally I don't bother responding to your AC trolling (I have no interest in chatting with teenagers), but for the record, I wrote that I had no issue with the notion of them expecting to be feature complete by the end of the year - e.g. being able to handle all situations (incl. stop lights, stop signs, traffic circles, etc) at some level of confidence. I am not a particular optimist as to the time for no driver at the wheel supervising, however - there's a long process of "chasing the nines" before I'll be comfortable with that. The key point, however, was that nor has Tesla claimed that there will be no driver behind the wheel by the end of the year - the specific claim was feature complete. And, as I wrote, I have no issue with that being the expectation.
Re: (Score:3)
As for SAE driving levels, however, they do not specify a minimum "number of nines". They simply require specific featuresets, without specifying how reliable those features have to be (either in absolute terms or relative to humans). So being feature complete would, yes, be rated as level 5. If the vehicle A) does not require a human monitor or backup driver, B) performs all aspects of the driving task, and C) can do so under all roadway and environmental conditions, then it is by definition level 5.
One
Re: (Score:2)
Re: (Score:3, Insightful)
They literally named their stat "Vehicle Safety Score" (VSS), literally describe it as the combined probability of injury of all forms (not just, as you put it, "crashing into a wall"), and it's literally the one stat that they use to grant star safety ratings. If anything is broken, it's the NHTSA.
Re: (Score:2)
Re: (Score:2)
It's also IMHO ridiculous that the NHTSA thinks it's just fine to have its rating system stop at 5 stars, when 40% of all new cars get 5-star ratings.
This.
We see this problem everywhere and don't learn. I do a lot of risk analysis work, and it's the same thing. If every risk worth mentioning at all is rated "high" or "red", the whole system has lost its meaning. (which, btw., is why I advocate quantitative risk analysis every opportunity I get and consider your risk management broken if its based on three colours).
Let's be honest here: If your product beat the standard so much that you're unhappy that they don't have a higher category, wouldn't you want
Re: (Score:2)
People in general are actually really bad at cardinal rating; but this is a scientific process, which takes the "people" factor out and replaces it with process and metrics and numbers. The problem we see here is one of resolution: the broad, complex data is distilled down to a scale of 1 to 5.
Re: (Score:2)
Resolution is only part of it. Someone mentioned that the VSS score was 50% higher than it needed to be for a 5 star rating. So the stars don't even represent even slices of the point spread. If they intended for the stars to be representative of the score distribution then it fails there as well because 40% of new cars get a 5 star rating.
The real problem is that the star system is effectively meaningless but the NHTSA wants to tell people that they aren't allowed to advertise with anything else.
Re: (Score:2)
Re: (Score:2)
I might be peculiar, but I do see a difference between 1% chance of death and 0.1% chance of death, even if some pencil pusher has put them into the same arbitrary category.
Re: (Score:2)
Slashdot shouldn't cap scores at +5 (Score:2)
It's also IMHO ridiculous that the NHTSA thinks it's just fine to have its rating system stop at 5 stars, when 40% of all new cars get 5-star ratings. Great way to make all stalks of wheat equal by cutting down the tallest ones.
For the same reason, Slashdot shouldn't cap scores at +5. Not all +5 posts are created equal. Once in a while, a really profound post comes along, with far more merit than the other hundred-or-so posts that attain a +5 score each day.
Why Rei, I bet many of your posts would score +10 or +15 if they were permitted to do so!
Re: (Score:2)
Which is why statistically (in the UK at least) you are more likely to be killed driving in an SUV than a "normal" car.
Hell just changing the colour of your car can have a huge impact in the likely hood of being involved in an accident. The difference between a black vehicle (the most dangerous) and a yellow/orange (the safest) is quite astounding. Even against a plain white vehicle you would be mad to drive in a black or similarly dark coloured car.
Re: (Score:2)
I wish more people were aware of this, especially pre-dawn walkers, joggers and bikers. I've had several occasions where I've had to suddenly avoid someone in the street wearing dark clothing and no reflectors/reflective materials early in the morning. I often wish we didn't protect these idiots from Darwining themselves.
Re: (Score:2)
Re: (Score:1)
An SUV-SUV collision is more dangerous than a basic sedan-sedan.
This invokes one of those game theory grids where everyone plays the "well screw everyone else" card, the net result being everyone ultimately screws themselves. One could try to construe to the four-word statement "SUVs are more dangerous" but that would be a harder hill to defend, and devolve into conversations where everyone is nitpicks and pedants.
As for me, I like going through a roundabout faster than 3mph.
Re: (Score:2)
The amount of velocity change imparted depends on the mass too. As an extreme example, a bus hits a car, the car gets punted while the bus just slows down a little.
That only applies if the vehicles are modeled as a solid inflexible mass, so the energy transfer is a simple kinetic equation.
The crash between a big old 4x4 tank of an SUV and a Tesla would have a lesser version of that, but the person in the Tesla would still probably get more forces applied to them than the person in the SUV.
Sort of, but probably not. The Tesla vehicle would have more acceleration applied to it than the massive SUV, but the person likely wouldn't, because the car structure would collapse instead, dissipating a lot of the energy without transferring it to the person.
Modern vehicles in collisions are more like sandbags. They'll deform and deflect a lot of the impact before letting the occu
Re: (Score:2)
I don't see how that would be true considering a Tesla and another vehicle anyway, because generally the reason behind that weight protection differential in the past has been the mass of surrounding car components like the frame or the engine. With a Tesla a large about of the weight is in the battery, which is underneath driver and passengers and so I don't think adds much to your protection? Tesla does have all very good modern safety aspects like great crumple zones, but it's a different case than for other kinds of cars.
...have you ever calculated momentum transfer between two inelastically colliding bodies?
Have you thought about the whole system? (Score:1)
have you ever calculated momentum transfer between two inelastically colliding bodies?
Yes, as have most Slashdot readers I would assume.
But here's the thing, unlike a physics class these are not rigid bodies, it makes the whole system much more complex to think about, because the distribution of mass matters too.
Not to mention there is a lot of torsion force at work as well, because of the angle of impact and the distribution of mass and the design of the crumple zones...
As someone pointed out, just the fac
Re: (Score:2)
But here's the thing, unlike a physics class these are not rigid bodies
Well if you have inelastic collisions, you don't usually assume a rigid body, do you? My point was largely the one of the heavier object in a collision undergoing the lower velocity change, which seems to be relevant for possible injuries.
Re: (Score:3)
A head-on crash with a wall is about the same as a same-speed head-on crash with a vehicle of the same weight.
A head-on crash with a heavier vehicle is more equivalent to a higher-speed head-on crash with a wall.
The momentum and energy imparted onto a vehicle is proportional to the mass of the vehicle crashing into it. A 60-mph crash with an 18-wheeler would be much more devastating than a 60-mph crash with a motorcycle. Just think about the ending speeds; with a motorcycle crash, the car will end the colli
Re: (Score:2)
I mean, that's entirely true. But, traveling at the same speed, a higher-mass vehicle will impart more energy on a car than a lower-mass vehicle. The acceleration experienced by a passenger during a crash is inversely proportional to the mass of their own vehicle.
All else being equal, a heavier vehicle is safer than a lighter one in a collision. That's the point the NHTSA was making. Their relative protection index doesn't take vehicle weight into account, so you can't say a Tesla is the safest vehicle base
Re: (Score:1)
"A Chevy Suburban may well be significantly safer when colliding with any given vehicle, thanks to the protective nature of its mass and other factors." - Not without airbags and crumplezones, which do all the work there.
The mass is not nearly as consequential as controlling the speed with which the energy is imparted - crumplezones and airbags delay and absorb the shock which causes mortal effects.
Even a much larger masses' energy is absorbed "more" by this than actually transfered elastically to the "hi
Re: (Score:3, Informative)
The NHTSA tests used a fixed barrier -- in effect, an object with infinite inertia. That allows an apples-to-apples comparison of how vehicles fare in a collision, because they are all equally disadvantaged when they hit an object that cannot move.
But the fact is that many accidents don't involve hitting a fixed barrier. If a Tesla Model 3 hit a larger vehicle with higher inertia, then all other things being equal, the people in the Tesla would undergo more rapid deceleration, and therefore have a higher ri
Re:Tesla results would mislead by weight category (Score:5, Insightful)
Re: (Score:3)
This! If NHTSA is basing it's star rating on the score, but claiming the score doesn't mean the actual level of safety, then they have a problem.
Re: (Score:2)
It does mean the same as any other brain-dead scoring system.
If your 4-star range is from (hypothetical) 100 to 150 points and your 5-star range from 151 to 200 points then it means that vehicle A (score: 148) and vehicle B (score 155) are very different, but vehicle B and vehicle C (score 189) are identical.
Every range-based scoring system has this problem, which is why they're all fundamentally flawed.
Re: (Score:3)
If it's anything like the European tests then the stars are not a direct indication of how likely you are to be injured in the event of a crash. They account for that to some degree, but also indicate what safety features the car has. For example, a car might get fewer stars because it either doesn't have an automatic emergency braking (AEB) system, or that system failed the test.
Even that isn't completely clear though. Some emergency braking systems are prone to phantom braking (including the Tesla one, FW
They are not equal though (Score:1)
then all other things being equal, the people in the Tesla would undergo more rapid deceleration,
That depends on the design of crumple zones in both cars which affect the rate of deceleration for occupants, so all things are not equal. The reality of collision physics is way more complex than people are making it out to be.
Re: (Score:2)
I don't see how that would be true
Because you are a moron who doesn't understand momentum or inertia. If you hit an immovable object, then your vehicle weight doesn't matter.
But if you hit some bushes, a fence, or another vehicle and all else is equal, a heavier vehicle will safer because it will experience less change to its velocity and direction, and the change it does experince will be less sudden. Thus, the passengers have a smaller and gentler collision.
Re: (Score:2)
I don't see how that would be true...a large about of the weight is in the battery, which is underneath driver and passengers and so I don't think adds much to your protection?
You're thinking crumple zones/structural integrity/cabin intrusion. You're thinking too much. Force equals your [vehicle's] mass multiplied by your [rate of] deceleration [in the wreck]. The battery may be contributing/performing little to no reinforment of the bodyshell, but depending on the dynamics of the wreck, it could still greatly help or hinder, as the case may be.
Re: No engine in the front (Score:2)
Re: Musk is a Sociopath (Score:1)
Musk is the guy who strived to make PayPal even worse (they wouldn't let him).
Re: (Score:1)
I don't dabble at all in the stock market. If I did, it certainly wouldn't be with Tesla.
Could Potentially (Score:2)
It's all about the way you word it.
Now had Tesla made a statement like :
"Could potentially have the lowest probability of injury of any vehicle ever tested by the National Highway Traffic Safety Administration (NHTSA)."
It would read the same to consumers and no push back from the NHTSA because you aren't breaking any rules.
NHTSA should tell us 5 stars is by weight class... (Score:2)
IMHO, yeah, Tesla slightly stepped over the line w/their phrasing.
But, the bigger issue is WHY ARE ALL 5 STARS NOT THE SAME?!
I actually learned this when the Model 3 got their 5 stars, IIRC, Musk on his twitter feed or something mentioned that the X is actually safer because it's heavier and it's 5 stars is better than the Model 3's 5 stars.
Well, who knew that before?! For a long time I assumed 5 stars in a sub compact was approximately the same as 5 stars in a pickup truck. It didn't really make sense that
Re:NHTSA should tell us 5 stars is by weight class (Score:5, Insightful)
IIRC, Musk on his twitter feed or something mentioned that the X is actually safer because it's heavier and it's 5 stars is better than the Model 3's 5 stars. ... Fix the bigger system, not just Tesla's claims.
Of course, an alternative takeaway would be to not assume anything Musk says is Gospel.
Re: (Score:2)
IMHO, yeah, Tesla slightly stepped over the line w/their phrasing.
But, the bigger issue is WHY ARE ALL 5 STARS NOT THE SAME?!
I actually learned this when the Model 3 got their 5 stars, IIRC, Musk on his twitter feed or something mentioned that the X is actually safer because it's heavier and it's 5 stars is better than the Model 3's 5 stars.
Well, who knew that before?! For a long time I assumed 5 stars in a sub compact was approximately the same as 5 stars in a pickup truck. It didn't really make sense that they'd be exactly the same safety, but I kind of figured the minimum score for 5 stars was relatively low, or scores in different areas offset themselves enough to make the math work out.
Fix the bigger system, not just Tesla's claims.
It's a broad 5 star rating - if you meet certain requirements. It's as safe as they are willing to rate it.
Why?
Because if the NHTSA chose a favorite then they would be promoting a specific manufacturer.
That's a no no
Re: (Score:2)
They're not supposed to pick a favorite.
But the system seems to come from a time when seatbelts were optional. When the average standard of safety has moved up, the rating becomes meaningless because too many achieve it. If you have a 5-star rating and nobody gets 1, 2 or 3 stars anymore, you need to raise your standards.
Re: (Score:2)
Constantly moving the goal posts higher and higher would lead to cars
...becoming more and more safe. There, fixed that for you.
designed solely to score higher and higher ON THE TEST without regard to any sort of real world applicability
That's a problem of test design that can be solved.
Yet no safety evaluation gives bonus points for being a fast car.
I wonder why... no, wait, my mistake, I don't. There's scientific evidence that faster == less safe. I'm just too lazy right now to dig it out, google yourself.
If the "five star" standard is not reasonably achievable then
Nobody asked that the rating should be moved that high. You are having and argument with yourself there.
It would have failed all sorts of typical weapon tests
Absolute rubbish. The Russian military wouldn't adopt a weapon that fails its weapon tests. And it wasn't sloppily designed - it just had d
Re: (Score:2)
"Well, who knew that before?!"
Anyone who's ever taken a physics class.
Musk (Score:2)
Tesla's forte... (Score:2, Insightful)
But no, they have to embellish the truth or outright lie. Then the Tesla defence force is triggered with fanbois lea
Re: (Score:3)
I don't fully understand is why they even have to lie like this. Tesla vehicles are obviously pretty safe (as born out by crash testing), and they have advanced driving features. Why not leave it at that?
What if you were the best in the world at math, and you were up against me. Our exam was a grade 8 math quiz and we both got 10/10. Would you be happy to declare yourself an average schmoe despite being able to run rings around everyone else, or would you point to other factors, such as that you finished the test perfectly in record time without using a calculator?
A shitton of new cars get a 5-star NHTSA rating. There's no reason to stop their when the NHTSA's own published data shows that there are major d
Re: (Score:2)
A shitton of new cars get a 5-star NHTSA rating. There's no reason to stop their when the NHTSA's own published data shows that there are major differences between those cars getting a 5-star rating.
Yeah, if ratings are intended to communicate useful information to consumers, and to encourage automakers to compete on safety, to make vehicles better, then a rating system where nearly half of all new cars get the highest possible score is failing to to its job.
The NHTSA needs to either add some new categories -- six star, seven star, etc. -- or if they don't believe their system can usefully make those distinctions, why do they think that it can distinguish among the lower tiers?
I suspect the truth i
Re: (Score:2)
As I alluded to, this hasn't been Tesla's only struggle with telling the truth.
State auto tax (Score:1)
I rather cutely come up with the formula to tax automobile registered in the state of California. An annual state auto tax of $1/lb.
"But the NHTSA argues that this is statistical malpractice because it doesn't take into account vehicle weight. In a vehicle-to-vehicle crash, the occupant of the heavier vehicle is less likely to be injured."
This is part of the justification. Also, road repair, and simply public space utilization. A concerted effort to sell and own smaller and lighter cars helps everyone.
Tesla's Response (Score:5, Informative)