[_ea1k]

Good analysis. Just over a month ago, Electrek was posted here claiming that Teslas with humans were crashing 10x more than with humans alone.

That was based on a sample size of 9 crashes. In the month following that, they've added one more crash while also increasing the miles driven per month.

The headline could just as easily be about the dramatic decline in their crash rate! Or perhaps the data is just too small to analyze like this, and Electrek authors being their usual overly dramatic selves.

[epistasis]

That is an overly optimistic way to phrase an apparent decrease in crashes, when Tesla is not being upfront about data that at best looks like it's worse than human crash rates.

Unless one was a Tesla insider, or had a huge interest in Tesla over other people on the road, such spin would not be a normal thing to propose saying.

Media outlets, even ones devoted to EVs, should not adopt the very biased framing you propose.

[hedora]

I don't understand your claim.

Previous article: Tesla with human supervisor at wheel: 10x worse than human alone.

Current article: Tesla with remote supervisor: 3-9x worse than human alone.

Given the small sample sizes, this shows a clear trend: Tesla's autopilot stuff (or perhaps vehicle design) is causing a ton of accidents, regardless of whether it's being operated locally by customers or remotely by professionals.

I'd like to see similar studies broken down by vehicle manufacturer.

The ADAS in one of our cars is great, but occasionally beeps when it shouldn't.

The ADAS in our other car cannot be disabled and false positives every 10-20 miles. Every week or so it forces the vehicle out of lane (either left of double yellow line center, or into another car's lane).

If the data on crash rates for those two models were public, I guarantee the latter car would have been recalled by now.

[brookst]

I don’t think statistics work that way. A study of all Teslas and all humans in Austin for 5 months is valid because Electrek ran a ridiculous “study”, and this headline could “just as easily” have presented the flawed Elektrek stork as a legit baseline?

[fwip]

The 10x would be 9x if the methodology were the same. 9x->3x is going from reported accidents to inferred true accident rate, as the article points out.