[gluxon]

> For those driving without Autopilot, we registered one accident or crash-like event for every 1.92 million miles driven. By comparison, the National Highway Traffic Safety Administration’s (NHTSA) most recent data shows that in the United States, there is an automobile crash every 492,000 miles. While NHTSA’s data includes accidents that have occurred, our records include accidents as well as near misses (what we are calling crash-like events).

I'm not quite sure what to make of this data. How is it that Tesla cars (without autopilot) are in 4x less accidents than average?

[Shikadi]

Probably demographics. People who buy Teslas are probably less likely to be in an accident regardless of whether or not they're in a Tesla. Other things might matter too, like lane departure warning and blind spot detection, which would probably lead to less accidents on other newer cars with those features as well

[nolok]

Because the NHTSA data includes all car, including the very cheap cars with bare minimum safety/help systems, old restored cars, ...

For it to have any value, it should compare each Tesla cars to modern/similar age car of the same category and price range.

[AnthonyMouse]

> Because the NHTSA data includes all car, including the very cheap cars with bare minimum safety/help systems, old restored cars, ...

https://crashstats.nhtsa.dot.gov/Api/Public/ViewPublication/...

94% of crashes were caused by drivers, 2% by vehicle failures. This implies that the age or state of the vehicle has minimal influence (only the 2% that actually failed) outside of the propensity for older vehicles to have less safety features.

And comparing the new safety features that reduce or mitigate the driver error to those of the average vehicle is exactly the point -- the difference in outcomes is large. This actually matters even across price ranges because people may be willing to pay more for a car if it's more than four times less likely to kill them or their family.

It would be nice to have a comparison separated by class as well (obviously that data would have to come from the government rather than Tesla), but the comparison to the average still has value because it at least informs you whether a more expensive car is worth it.

[plasticchris]

Would switching lanes into another vehicle be an accident caused by a driver? Blind spot warnings. These safety features do reduce accidents caused by drivers.

[AnthonyMouse]

> Would switching lanes into another vehicle be an accident caused by a driver? Blind spot warnings. These safety features do reduce accidents caused by drivers.

That's the point.

The comparison to existing vehicles is valuable because the first requirement before someone is interested in a new car is that it be sufficiently better than their existing car. A difference of that magnitude is relevant.

[btilly]

That doesn't negate the point exactly because older vehicles have fewer safety features, and we know that those safety features prevent accidents. Take a look in https://en.wikipedia.org/wiki/Motor_vehicle_fatality_rate_in... and see that the fatality rate has dropped by a factor of 2 in the last 40 years. Features like ABS are a big chunk of that.

For example ABS systems reduce accident rates by

[Retric]

Drunk driving reduction the majority of that.

In 1982 the fatality rate per 100,000 people was: 18.969, in 2016 it was 11.59. By comparison Drunk driving fell from 9.1 per 100k to 3.3 per 100k in 2016.

As an approximation drunk driving is ~5.8 and everything else is ~1.58. (Clearly some of those non drunk driving related saved lives also saved drunk drivers, but it's still ~3x as important as everything else put together.)

https://www.responsibility.org/wp-content/uploads/2018/02/Al...

[x38iq84n]

AP racks up miles on highways and in favorable conditions, i.e. out of city traffic, parking, most pedestrians interaction and so on.

[jnevelson]

I'd be curious to see if the crash rate is similar to the crash rate of other cars in the same class (BMW, Mercedes, etc). Is it possible that people tend to drive more cautiously in a more expensive car?

[tomatocracy]

It's more like: people who can afford and buy expensive cars tend to be middle-aged, which means they are on the whole (i) more experienced drivers (without being at the stage of their lives where eyesight/hearing/reaction times might be failing without them realising) and (ii) less prone to risky behaviour of any sort vs (say) 16-25 year olds.

In addition, if you can afford an expensive car and choose a Tesla over (say) a sports car, I'd guess that probably means you aren't likely to be an aggressive driver which in turn means lower accident rates.

[ssheth]

Probably because when you're off autopilot, Tesla drivers (including me) are much more alert and aware of what's going on compared to a normal car where you have to alert/aware the whole time and in so doing, tend to "zone out" when doing monotonous driving. I tend to use Autopilot for the monotonous driving parts .. and then take over myself if there is something that needs more attention (construction, weather conditions, crazy drivers, etc.)

[gizmo385]

> I tend to use Autopilot for the monotonous driving parts .. and then take over myself if there is something that needs more attention (construction, weather conditions, crazy drivers, etc.)

To me, this seems to imply that you're not paying attention when Autopilot is engaged.