Another Tesla crashes into guardrail in Montana

[UncleDave]

You may not be recalling correctly. 08-09 The NHSTA data I can find shows no sub 1.0 death per million miles.

They show 1.26 and 1.15 death per million miles - both worse than Tesla Number.

The lowest they show (that I can find is) 1.08 per 100M miles I can only find to 2014.

http://www-fars.nhtsa.dot.gov/Main/index.aspx

At 1.0 per 130M Tesla autopilot is markedly better with less fatalities and one third more miles.

Good commercial Shannow.



UD

 

[Garak]

Originally Posted By: mikered30
Tesla has said multiple times that auto steer is not for roads that lack a center divider unless the software update was completed. They will have to fix this by warning the driver or disabling that feature while there is no center divider.

So, basically, Tesla operators are alpha testers?

 

[OilyWaterMIXER]

Originally Posted By: Garak
Originally Posted By: mikered30
Tesla has said multiple times that auto steer is not for roads that lack a center divider unless the software update was completed. They will have to fix this by warning the driver or disabling that feature while there is no center divider.

So, basically, Tesla operators are alpha testers?


No, Tesla operators are Crash Test Dummies.

 

[HTSS_TR]

Originally Posted By: UncleDave
You may not be recalling correctly. 08-09 The NHSTA data I can find shows no sub 1.0 death per million miles.

They show 1.26 and 1.15 death per million miles - both worse than Tesla Number.

The lowest they show (that I can find is) 1.08 per 100M miles I can only find to 2014.

http://www-fars.nhtsa.dot.gov/Main/index.aspx

At 1.0 per 130M Tesla autopilot is markedly better with less fatalities and one third more miles.

Good commercial Shannow.

UD

I remembered I saw 0.95 death per 100M miles last decade. My memory may be deteriorating.

For Tesla's auto-steer to work it needs clear lane markings on both sides, also looks like it can't handle tight curve with higher speed.

If Tesla autopilot was on with this accident, Tesla is in bigger trouble then they thought, NHTSA will bug them until they fixed the deficiencies with current release autopilot.

 

[SatinSilver]

Originally Posted By: HTSS_TR
that why a driver got kill in Florida accident.


What nationality are you if you don't mind me asking? I detect an accent lately. Thanks.

 

[HTSS_TR]

Asian

 

[HTSS_TR]

Originally Posted By: Garak
Originally Posted By: mikered30
Tesla has said multiple times that auto steer is not for roads that lack a center divider unless the software update was completed. They will have to fix this by warning the driver or disabling that feature while there is no center divider.

So, basically, Tesla operators are alpha testers?

Many drivers used autopilot on 2 lane streets(one lane each way) without problem, as long as lane markings are clearly visible. The main missing capability is autopilot doesn't know what to do with stop sign and red light, and it doesn't have the ability to reduce speed for sharp turns.

 

[Garak]

Originally Posted By: HTSS_TR
Many drivers used autopilot on 2 lane streets(one lane each way) without problem, as long as lane markings are clearly visible. The main missing capability is autopilot doesn't know what to do with stop sign and red light, and it doesn't have the ability to reduce speed for sharp turns.

So it can't handle heavy traffic with a modern infrastructure and it can't handle no traffic in a geographic large, population light state without a brand new infrastructure, then? What's it going to do with anything less than perfect weather? Outside of California, we occasionally see street signs and reflective markings disappear thanks to snow covering, not to mention what happens to lane markings.

"Autopilot" relying on lane markings is an absolute dead end. Why they're even bothering forking in that direction, I'll never know.

 

[UncleDave]

Welcome to the deteriorating memory club at 49 Im feeling it as well.

Teslas number is lower than the lowest rate in 10 years according to the data.

"My personal opinion is death per 100 million miles must be less than the lowest rate of the last 10 years."

130M-1 is safer than 100M-1.08. The best year on record ever I can find.

By your own metric it passes. The numbers show its flat safer to accumulate miles with it engaged than to drive yourself.

Totally happy to change my mind if the data changes, but the numbers says it works pretty well.

UD

 

[Shannow]

Originally Posted By: UncleDave

By your own metric it passes. The numbers show its flat safer to accumulate miles with it engaged than to drive yourself.


You still haven't provided the death rates for the general public when they are limited to ONLY the same operational scenarios that Teslas autopilot operates in, highways with barriers and clear lane markings.

As per HTSS_TRs explaination of the limitations of operation in his last few posts.

Tesla's run rate at traffic lights ?
Tesla's run rate in the snow and ice ?
Tesla's run rate on rural 1-1/2 lane roads with wandering deer (we KNWO it can't see semis)?

Like I said, the data COMES from cherry picked operational situations, while the driving populace operates on all the OTHER bits as well.

 

[UncleDave]

Nope I can't provide you or anyone with a custom data request out of what we know today.

We know what the NHSTA posts because I posted that link.

We know what tesla tells us they logged 130M with it engaged. Well soon find out what else we do and dont know.

Do we agree on the 1.08M per million is the NHSTA's published rate?

What we will also learn is how many more or less accidents have happened and that will tell us a trove of more data on how good it is or isn't.

Like I said Im open to new data, until I do I have to look at the data we have.

The data we have is very compelling.



UD

 

[Shannow]

Here's another way of looking at the stats

http://www.iihs.org/iihs/topics/t/general-statistics/fatalityfacts/state-by-state-overview

Vast variance between the states...

 

[UncleDave]

Many states keep different records than the NHSTA, but thats the governing body that Tesla is being requested to report to, so thats most relevant to how this data is being looked at.

What interesting is that nearly every data point breakdown can be gleaned from the telemetry onboard the car instantly vs waiting for a police report, prob high 90% can now be automated.

If Teslas data quality meets or beats the NHSTA's meaning of the statistic then they will rule the relevant number in terms of deaths per million miles. NHSTA either has to agree with or show why teslas number isn't what they claim and why their conclusions vs the conclusions of a rocket corporation thats pretty darn good at carrying the 1 should hold water. Gonna be fun to watch.

Based on the early top level numbers this could be the equal of the safety test where the tesla scored more than any other car, or we could find through the data its years off.

Its also interesting to see who gets held to the same standards.

Lots of guys are real interested in what kind of safety a car they don't own has, and are very eager to restrict this and that all the while driving much much higher fatality per million mile vehicles themselves.

But then again thats all part of BITOG!

 

[Shannow]

I'm just not as enamoured at an apples/bananas comparison at "proving" something as others clearly are.