Why Full Self Driving is needed

You should be able to differentiate between your "signature" and your profile.

Your data in https://priuschat.com/account/personal-details is incorrect. for "Vehicle: Required" you list a 2017 Prius Prime. One would assume that if the vehicle model is required, you have an obligation to keep it updated. The board allows you to list your Tesla there. Don't forget to click "save" when you are done.

 

That's a bit of a stretch, to say that it saved your life. After all, there should have been rumble strips alongside the road, and your wife should have relieved you of your task if you were having problems. I don't understand why you kept driving after the first time you nodded off. That's just bad judgment.

I guess I can say that my car has saved my life a few times. Not long ago I made a left turn onto a boulevard without noticing that another car had sped up to catch the light such that he was on my left in my blind spot as we completed the turn. Fortunately, the car sounded the alarm as I started to change lanes. Otherwise I would have ended up bouncing off of him and might have ended up in oncoming traffic.

Then there are all the times that my cross traffic monitoring radar tells me when a car is approaching from either side as I back out of a parking lot. That's been a real life saver in areas where SUVs are the dominant vehicle. Yes, there is nothing as precious as good, reliable hardware and software.

 

LOL. You can't make 3 clicks of the mouse, yet you can spend all day typing posts on various forums? Give me a break!

I'll tell you what. Give me your password and I'll change it for you.

Or you can just click on https://priuschat.com/account/personal-details and then scroll down till you find the words "Vehicle: Required". There's a drop down box there where you can select your car, just like you did when you signed up on the board. Then scroll down and click on "Save Changes" near the bottom. It's so simple that you should be able to di it in your sleep.

Sincerely,

 

It wasn't Tesla that labeled it as "proof", it was you back at post #30.

It doesn't have to be either proof or a lie, with no possible states in between. It can be incomplete data that is suggestive, but not solid enough to be definitive, therefore really meaning an inconclusive outcome. At least for this moment.

I believe we all know that AP is used more on certain roads than others. In fact it can't be used on certain others.

We also know that certain roads have far higher fatality rates than others, per vehicle mile traveled. And that AP's experience database is not an unbiased sample of America's driving miles. It can't be, because many vehicle miles are on roads that AP simple doesn't handle.

It would seem intuitively reasonable to believe that AP miles are heavily over-weighted with Interstates and other freeways, and under-weighted on rural non-freeways. If this reflects reality, then AP miles would inherently be biased to low-fatality-rate roads where even Teslas without AP engaged also achieve lower fatality rates. Widening the bias even further, these sometimes-AP users would be biasing up the total high-fatality non-AP miles when they have it turned off for non-AP roads, compared to Telsa drivers who always keep AP off even on good freeways. Without accounting and adjusting for this selection bias, we simply can't have definitive proof that AP is safer than not-AP.

I haven't yet found road type fatality rates for the whole U.S. As a proxy until we find better, here is some recent data from one of my neighboring states:




The figures you quote in post #30 show AP being (4.31/1.59) = "2.7 times safer" than non-AP Tesla use. However, this Oregon data shows that Interstates are (1.19/0.33) = "3.6 times safer" than the highway system as a whole, or ("2.12/0.33) = "6.4 times safer" than non-freeways. And even "8 times safer" than minor arterials, or "12 times safer" than rural major collectors. Rolling in the "Other Fwys/Expressways" barely budges those ratios. All that leaves enough theoretical room for AP's "safer" than non-AP to be entirely the result of selection or weighting bias.

I just gave you some of the evidence, facts, and data that you said I didn't have. Now drop the childish insults, and bring back some more data to better illuminate this selection bias and your so-far inconclusive "proof".

 

Bob: Note your home page is a "not found".

I did manage to change my profile but the best I could find at the time was a "Other hybrid" for a Rav4 hybrid.

If you click on a poster's name and go to their information you may find some notes about what cars they own. Bob has that.

 

Nothing in this portion addresses the data selection bias problem I described.

Where do you get this NHTSA crash rate of 0.498?

I hope you didn't miss noticing that the Oregon and Tesla figures are inverted. Reciprocals. Oregon lists crashes per (million) miles, while Tesla lists miles per crash. I.e. Tesla's report of "NHTSA’s most recent data shows that in the United States there is an automobile crash every 498,000 miles" translates to (1 crash / 498,000 miles) = 2.01 crashes / million miles. On the contrary, Oregon's crash rate isn't twice this NHTSA rate, it is really just half of this NHTSA rate.

Put another way, where NHTSA reports a crash every 498,000 miles, Oregon reports a crash just every 1,060,000 miles.

I ignored that line because the sample size, a single fatality, is too small to be statistically significant. But its 3.17 crashes per million miles translates to Tesla's basis as a mere 315,000 miles per crash. It isn't the closest to Tesla's report, it is really the farthest away, far worse.

OTOH, Oregon's Interstate rate of 0.47 crashes per million miles translates to Tesla's basis as 2,130,000 miles per crash, not quite as good as Tesla's AP, but better than any Tesla non-AP reports.

 

There is another selection bias with the Tesla numbers. Most of the Tesla cars are newish. Since many newer cars might have better safety systems and maybe more careful drivers it may be that they all have lower crash rates. You would need to obtain the raw data from the wider market data and select out the data for each model year in order to do a realistic comparison. New cars are probably driven more miles than older cars so it is unclear how significant this effect is.

Mike

 

Wouldn't it be funny to find out that Bob is actually Elon Musk.

 

Just looking at the potential problems with those statistics. What factors for age of drivers are appropriate? What factors for vehicle condition are appropriate? What factors for time of day/night the vehicles are driven? And using only freeway miles and comparing them to typical mixed route driving? How about comparing to just non-Tesla cars with warning systems (ones recent enough to have em)?

I'm thinking the Tesla AP might be favored by some of those adjustments and not by others. But without those aren't we comparing oranges and tangerines?

 

That's a neat trick; using 2021 data for Tesla and 2019 data for all others and then pretending that it's a valid comparison. That's two different areas with different climates (US as a whole VS Pacific Northwest) and different standards. Since the majority of the Tesla sales are in California (which had a mild, dry climate in 2021) you are not comparing similar data sets.

I spent a good deal of 2019 in Oregon, and we had some really bad weather that year. I can attest to the fact that there were almost no Teslas in Oregon outside the Portland metro area.