4 Crashes in a drive of 250k miles. How good/bad is this with human driving? I couldn't compare this with human driving. Though I found a report on fatal crashes here - https://www.iihs.org/research-areas/fatality-statistics/deta...
It's not just about fault.
Waymo has a well deserved reputation for vastly reducing the frequency and severity of accidents where it is at fault.
But if you look at the data for all crashes, regardless of fault, it's clear that Waymo also reduces the frequency and severity of crashes where other drivers are at fault.
Waymo's "we got hit by someone else" crashes are substantially lower per mil, probably on the order of 50% to 70% reduction, not just the crashes it causes.
A month or two back, I was driving down a steep one-way, three-lane street in SF, late on a rainy night. I saw a Waymo stopped at the left curb and I moved to the from the left lane to the center lane in case it started to pull out into the left lane. There were no cars in front of it or behind it, so I was shocked to see it quickly leave the curb at about a 45 degree angle, as if it were pulling out of a tight spot with a car parked just in front of it, but much faster. If I saw a human driver doing that, it would almost certainly mean they are trying to get all the way across the street immediately. If it was doing that, there is no way I could stop in time on the wet downhill. I tried, but that just made steering difficult as my anti-lock brakes struggled to find any traction at all. Then, just as quickly, it straightened out in the left lane. I'm glad I was the only other car around.
One element of defensive driving is thinking about how to avoid surprising other drivers. When will self-driving cars' defensive driving rise to that level? Waymo certainly wasn't there in that situation on that night.
If we use insurance claim as the definition then: - The average driver files an insurance claim for a car crash about once every 17.9 years [1] - The average driver drives 13,476 miles per year [2] - This means one insurance claim per 241,220 miles driven by a human driver.
1. https://www.gtslawfirm.com/what-are-the-chances-of-getting-i... 2. https://www.fhwa.dot.gov/ohim/onh00/bar8.htm
The waymo/robotaxi driving is basically entirely city driving, so I think I think it's reasonable to say human accidents for that type of driving are higher, possibly nearly twice as high as the estimate you got.
It's reasonable to assume that a seasoned taxi driver will be _better_ than the average human driver. Many serious accidents happen because the driver was inebriated or drowsy/falling asleep, which shouldn't be the case with taxi drivers.
For a proper comparison, we should look at the crash record for _taxi drivers_, not the general population.
Furthermore, if I'm getting an Uber/Lyft I'm likely to choose drivers with a 5-star rating or close to it. That basically filters out all poor drivers. Which means that not only should we compare robo-taxis with taxi drivers, but with the _top rated taxi drivers_.
i would calculate it that way to compare it to the last stat i had on waymo which was +/- an accident every 98,000 in an unattended vehicle.
comparing that to a human driver using stats from [0] the “AAA Foundation for Traffic Safety”
page down for maintenance, but summarized at [1]:
ages 16-17: crash every 69,000mi
ages 18-19: crash every 136,986mi
ages 20-29: crash every 182,149mi
ages 30-79: crash < every 303,030mi
ages 80: crash every 231,481mi
[0]: https://aaafoundation.org/rates-motor-vehicle-crashes-injuri...
[1]: https://www.friedmansimon.com/faqs/how-common-are-car-accide...
with that said: unattended vehicles are required to report every single incident no matter how minor or who is at fault. adults might just shrug off a minor contact and get on with their day without reporting or claiming it.
readily located a variety of news outlets investigating the same question and summarizing that in waymos case human drivers were mostly at fault… even if you ignore stuff that smacks of PR/industry or from waymo itself…
piling on to the flaws in comparing these: waymos operate in major metros that just have higher crashes and claims to begin with. im spitballing with nationwide averages but i think a serious inquiry/research would have to drill into the crash rates for the cities theyre operating in and where each incident occurred.
eg, i live in los angeles and have over 2,000,000 incident free miles out here but ive been hit no less than four times in austin TX and doubt i even cracked 20,000mi there.
I’d be interested in how waymo compares against LA drivers operating in LA…… against Austin drivers operating in Austin. More so than how they compare to all drivers on average nationwide. Without assessing that… say if waymo has a comparable incident rate as Tesla in Austin, youd be overlooking that by comparing a company operating in multiple markets- to one thats only in Austin.
https://www.azfamily.com/2025/05/01/what-happens-when-waymo-...
https://arstechnica.com/cars/2025/03/after-50-million-miles-...
Most Austin human drivers don't crash 4 times in 250k miles.