Tesla Vehicle Safety Report Shows Autopilot Crashes Less Than U.S. Average

Tesla’s Q4 Vehicle Safety Report shows vehicles operating on the company’s Autopilot system are involved in fewer accidents than average in the U.S.

Tesla registered one accident for every 2.91 million miles driven in which drivers had Autopilot engaged and one accident for every 1.58 million miles driven without Autopilot, according to the report. The report points out the the National Highway Traffic Safety Administration’s most recent data shows that in the U.S. there is an automobile crash every 436,000 miles.

Tesla reported an increase in accidents over the previous quarter, but did not provide any additional context. For the Q3 2018 results were: one accident for every 3.34 million miles driven with Autopilot and one accident for ever 1.92 million miles driven without. NHTSA’s average at the time was an automobile crash every 492,000 miles.

Tesla committed last year to providing accident statistics for vehicles on Autopilot in an effort to be transparent and build public confidence in autonomous vehicles.

The Vehicle Safety Report is available here.

 

[Barklikeadog]

NHTSA's average accident rate includes teenagers, elderly, drunks, etc (People who don't drive Tesla and really skew the numbers)
Accident rates for all vehicles that have safety features like 'blind spot detection' have fallen dramatically.
I like the additional safety that 'autopilot' systems create, but to compare 'tesla drivers' (AGI $300k age 35-60) to the general population
of drivers is wrong wrong wrong.
Nobody dies driving a Volvo XC90 for a reason.

 

[ronm2948]

I was also wondering about the types of roads being driven on. I would guess that accidents per mile are much higher on city streets vs on freeways where people generally use autopilot. If you looked at cars without autopilot driving on freeways, I wonder what the accidents per mile would be.

I love autopilot, and think there are a lot of great safety features in Teslas, but I don’t think that the numbers Tesla touts are really a fair apples to apples comparison.

 

[omgwtfbyobbq]

NHTSA's average accident rate includes teenagers, elderly, drunks, etc (People who don't drive Tesla and really skew the numbers)
Accident rates for all vehicles that have safety features like 'blind spot detection' have fallen dramatically.
I like the additional safety that 'autopilot' systems create, but to compare 'tesla drivers' (AGI $300k age 35-60) to the general population
of drivers is wrong wrong wrong.
Nobody dies driving a Volvo XC90 for a reason.

Tell that to this guy.

Man Arrested for Drunk Driving after Officers Found Him Asleep in Tesla Running in Autopilot Mode

 

[mongo]

I was also wondering about the types of roads being driven on. I would guess that accidents per mile are much higher on city streets vs on freeways where people generally use autopilot. If you looked at cars without autopilot driving on freeways, I wonder what the accidents per mile would be.

I love autopilot, and think there are a lot of great safety features in Teslas, but I don’t think that the numbers Tesla touts are really a fair apples to apples comparison.

The non-Autopilot Tesla accident numbers are better than non-Tesla rates also. It is not just a places-where-you-can-use-AP comparison.

 

[mswlogo]

NHTSA's average accident rate includes teenagers, elderly, drunks, etc (People who don't drive Tesla and really skew the numbers)
Accident rates for all vehicles that have safety features like 'blind spot detection' have fallen dramatically.
I like the additional safety that 'autopilot' systems create, but to compare 'tesla drivers' (AGI $300k age 35-60) to the general population
of drivers is wrong wrong wrong.
Nobody dies driving a Volvo XC90 for a reason.

True. But they also compared their own vehicles with AP on and off and it was better on.

But you have to be careful with that too. That we tend to only engage AP when traffic conditions are "Simple", good weather and clearly marked. So that could skew results as well.

And we also tend to not engage in off the highway, congested areas etc. where most accidents happen.

Only if AP is on 100% could you really make any kind of statistical conclusion.

 

[ChadS]

I completely agree that there are issues with the numbers being compared.

But frankly I don't know where to get the precise apples-to-apples numbers that we all would like to see (if anybody does know, it would be very helpful to see that mentioned here). And Tesla, of course, would be rightly criticized if they didn't release any numbers at all.

What they have released is definitely not perfect, but I'm glad they put them out. We can at least see that Tesla non-autopilot numbers are roughly 1/3 of general numbers, and then autopilot numbers are half of that. Indeed we should apply some discounts for it not being apples-to-apples, but I still think the numbers bode well. (Especially when you note that the numbers are for accidents the cars are involved in, rather than accidents where the car in question is at fault. That somewhat - though by no means wholly - reduces the effects of different driving styles, etc).

 

[Barklikeadog]

The non-Autopilot Tesla accident numbers are better than non-Tesla rates also. It is not just a places-where-you-can-use-AP comparison.

...because Tesla drivers are $300k agi households age 35-60. The safest drivers on the road. Autopilot or no autopilot, if the vast majority of your drivers come from the safest group
of drivers, of course you will get these results. Compare Tesla safety stats to similar drivers in other cars, like Volvo drivers, and suddenly there's no difference.

When they announce 'no autopilot accidents' I think... well there's a human backup safety system. How safe would autopilot be if the human didn't take over? Autopilot adds valuable safety
features to the car, but these misleading stats drive me nuts.

 

[Barklikeadog]

I completely agree that there are issues with the numbers being compared.

But frankly I don't know where to get the precise apples-to-apples numbers that we all would like to see (if anybody does know, it would be very helpful to see that mentioned here). And Tesla, of course, would be rightly criticized if they didn't release any numbers at all.

What they have released is definitely not perfect, but I'm glad they put them out. We can at least see that Tesla non-autopilot numbers are roughly 1/3 of general numbers, and then autopilot numbers are half of that. Indeed we should apply some discounts for it not being apples-to-apples, but I still think the numbers bode well. (Especially when you note that the numbers are for accidents the cars are involved in, rather than accidents where the car in question is at fault. That somewhat - though by no means wholly - reduces the effects of different driving styles, etc).

It is easy to compare apples to apples. you can find accident/deadly accident rates broken down by vehicle. I always use the Volvo xc90 as an example. year after year, no deaths in that vehicle.
and all of those years, no autopilot system, yet no deaths. these nonsense stats say more about the driver than the car

 

[Az_Rael]

More accidents this quarter over last. Who knows if that is because of the season or what. Will be interesting to see the long term trends as more and more Teslas get on the road.

 

[StealthP3D]

Nobody dies driving a Volvo XC90 for a reason.

Even XC90 owners have teenagers, drink alcohol and fall asleep at the wheel. The reason they have a very low death rate is that they are very safe cars. My 72-year-old brother's mother-in-law attributes the fact that she is still alive to the strength of the XC90 safety cage. She was tired and left the roadway going freeway speeds and hit an embankment. They had to cut her out with jaws of life. She was bruised with broken ribs and pelvis and shaken up but she lived to talk about it. Her husband, a retired orthopedic surgeon, also says she would likely not have survived in a lesser car.

This isn't magic, you can measure crash safety, and they do, with crash test dummies. The Model 3 performed the best of any car ever tested. So don't try to explain it away by safe drivers. Look at this horrific, high-speed drunk driving accident in a Model 3:

Deputies: Drunk driver crashes Tesla into pond in Canby, airlifted to hospital

This guy would be dead in a lesser car. And, no, Tesla drivers are not all rich, high income, smart, responsible people. The performance of the Model 3 P3D and the P100D has attracted plenty of thrill seekers.

 

[StealthP3D]

It is easy to compare apples to apples. you can find accident/deadly accident rates broken down by vehicle. I always use the Volvo xc90 as an example. year after year, no deaths in that vehicle.
and all of those years, no autopilot system, yet no deaths. these nonsense stats say more about the driver than the car

Tesla's statistics did not measure deaths, they measured the accident rate.

Tesla short-sellers have been telling us for years how the AutoPilot feature is extremely unsafe, some even say criminally so. The numbers of these Autopilots in real-world use tell the opposite story. You are less likely to be in an accident when AutoPilot is engaged. Much less. That's a fact.

Your smokescreen argument that it's all demographics falls apart when you consider that Tesla's with AP engaged are more safe than the same demographic of Tesla owners with AP turned off. Almost twice as safe. And Tesla's with the AP turned on are over 6 times as safe as the average car on the road.

Also, Volvo crash safety isn't what it once was. See this article:
Volvo Models Dropping From IIHS Lists of Safest Vehicles | Torque News

 

[mongo]

The wording this quarter didn't include "or crash-like events". However, based on the numbers it seems like it still does. (Or drivers got a lot worse)

 

[deonb]

Also to consider is that you’re not just comparing AutoPilot to Human drivers. You’re comparing 1 driver to 2 drivers.

Put two active human drivers in a car and the accident rate would be a fraction of what it is today.

 

[Barklikeadog]

Tesla's statistics did not measure deaths, they measured the accident rate.

Tesla short-sellers have been telling us for years how the AutoPilot feature is extremely unsafe, some even say criminally so. The numbers of these Autopilots in real-world use tell the opposite story. You are less likely to be in an accident when AutoPilot is engaged. Much less. That's a fact.

Your smokescreen argument that it's all demographics falls apart when you consider that Tesla's with AP engaged are more safe than the same demographic of Tesla owners with AP turned off. Almost twice as safe. And Tesla's with the AP turned on are over 6 times as safe as the average car on the road.

Also, Volvo crash safety isn't what it once was. See this article:
Volvo Models Dropping From IIHS Lists of Safest Vehicles | Torque News

I'm not a short seller saying that it is unsafe, I said autopilot adds a valuable safety feature. What we don't have is the information of autopilot with human backup, vs. other driving that is
a human with no backup. Sure it's safer. Just like other vehicles on the road that have added safety features like blind spot detection. Those features are causing drops in series accident
rates as well.

The "average car on the road" is 12 years old and driven by a less safe demographic. If teenagers could buy a Tesla off a used lot for $2000, the miles per accident rate would plummet.