FSD Beta v11.x

TK211X · Dec 18, 2023

Welp, I burned my check tonight. Going to go wack it.

rlsd · Dec 18, 2023

Supcom said:
Teslafi reports software version data from the cars that are subscribed to their service. Since this includes only those who are interested in more technical details or their cars, it is likely not to be an accurate reflection of the Tesla population as a whole.

I guess Teslafi has more older tech geek owners than newer owners.

KArnold · Dec 18, 2023

swedge said:
But.... almost none of the updates are from 2023.27.x, the fed tester crowd. I

Very surprisingly I went from 2023.27.11 to 2023.44.30.1. So not entirely true but no idea on how many. Probably not many as you observe. So far nothing newer, neither would I expect it yet historically speaking.

Very strange as my typical deployment times are typically in the last 20% of the fleet. I've never been early. Always thought it was a Midwest thing. Maybe it truly is random once past the BFF's.

kabin · Dec 18, 2023

Still wrapping my head around the Forbes article showing Teslas have the highest crash rate of 30 auto brands. It's a curious stat normalizing to drivers versus miles. Apparently LendingTree looked at tens of millions of quotes over a 1 year period ending Nov 2023. In any event, accidents per 1000 drivers: Tesla 23.54, Ram 22.76, and Subaru 20.90.

The article ties it to the recent AP recall.

Tesla Has The Highest Accident Rate Of Any Auto Brand

jebinc · Dec 18, 2023

kabin said:
Still wrapping my head around the Forbes article showing Teslas have the highest crash rate of 30 auto brands. It's a curious stat normalizing to drivers versus miles. Apparently LendingTree looked at tens of millions of quotes over a 1 year period ending Nov 2023. In any event, accidents per 1000 drivers: Tesla 23.54, Ram 22.76, and Subaru 20.90.

The article ties it to the recent AP recall.

Tesla Has The Highest Accident Rate Of Any Auto Brand

Could be many things, such as, folks not used to the EV acceleration/torque/one pedal driving, crazed devotionalists testing Elon’s claims, or many other reasons. How do they stack up against other EVs?

archae86 · Dec 18, 2023

kabin said:
Still wrapping my head around the Forbes article showing Teslas have the highest crash rate of 30 auto brands. It's a curious stat normalizing to drivers versus miles. Apparently LendingTree looked at tens of millions of quotes over a 1 year period ending Nov 2023. In any event, accidents per 1000 drivers: Tesla 23.54, Ram 22.76, and Subaru 20.90.

The article ties it to the recent AP recall.

Tesla Has The Highest Accident Rate Of Any Auto Brand

A major differentiator in crash rates is the selection bias of who buys a particular car. How they drive, where they drive...
I remember years ago routinely perusing a set of model-specific crash statistics that was published annually. For several years there was a consistent pattern that cars that were sold in both 2-door and 4-door versions had drastically higher crash rates in the 2-door version (something like 2X!!!). As no one suggested there was a sufficient structural deficiency to cause this, selection bias by driver and use was likely the cause.

I'd revise the claim to say that "Tesla Drivers have the highest crash rate among the drivers of 30 auto brands". That is assuming they actually did the math correctly and are reporting fairly otherwise.

DrGriz · Dec 18, 2023

archae86 said:
A major differentiator in crash rates is the selection bias of who buys a particular car. How they drive, where they drive...
I remember years ago routinely perusing a set of model-specific crash statistics that was published annually. For several years there was a consistent pattern that cars that were sold in both 2-door and 4-door versions had drastically higher crash rates in the 2-door version (something like 2X!!!). As no one suggested there was a sufficient structural deficiency to cause this, selection bias by driver and use was likely the cause.

I'd revise the claim to say that "Tesla Drivers have the highest crash rate among the drivers of 30 auto brands". That is assuming they actually did the math correctly and are reporting fairly otherwise.

The actual Lending Tree article starts with a disclaimer that the content is based on the opinion of the author and hasn't been "reviewed."

That's not to say it isn't accurate, but it's hard to say how things like miles driven affect the results, for instance, just at first glance.

Mardak · Dec 18, 2023

swedge said:
So far only 9% of the .2 installs are from 2023.27x, but those represent ~22% of the TeslaFi fleet

Yeah, even worse with the large rollout of 2023.44.30.4 just now to over 2000 TeslaFi vehicles. 90% of Teslascope pending are on 2023.44.1 and not even 1% for 2023.27.x. Although .4 seems to newly add older vehicles with HW0/HW1, so if that's the focus, makes sense why it's not going to those with 11.4.7.

swedge · Dec 18, 2023

TK211X said:
Can someone please confirm that Teslafi sources data from opted in vehicles and that it only represents a disproportionate pool of users.

This would mean that TeslaFi could report 500 downloads when the backend number is 50,000 or conversely report 1500 downloads when the backend number is 5,000?

It gives an illusion that there’s “only” X amount of downloads going out when in reality it can be Z so you could be discouraged from using your once per 24hr check.

Though if you get it doesn’t it push to you regardless of time of day?

We can hope that there is a good spread of TeslaFi users that yield hopefully an equal % of the real Tesla server numbers whatever it may be like 1:150 or something.

Right, TeslaFi is a not very random subset of all Teslas. There are roughly 20,000 on Fi, well over 2 million world wide Tesla vehicles. While the sample size is statistically decent, it is probably biased for early adopter bit head types. (TMC has it's own demographics, including folks with time on their hands.)

Recent reporting suggest that the one check per 24 hours is moot. It appears that Tesla dumps a number of specific vehicles into the Pending pile, and when they connect to WiFi, download and install at their own pace they get included in the installed count*. Who gets into the Pending pool depends on hardware variants and FSD beta tester status. A few folks, like Chuck and Omar get early access. Vehicles are added to the pool periodically, often daily, and typically in increasing numbers in each batch. But besides the criteria, who gets into the pool appears random.

The current 2023.44.30 updates appear to be mostly, but not exclusively USA, as are TeslaFi. The recall speaks of roughly 2 million, so 1:100 is a decent SWAG.

*TeslaFi gets account access for it's subscribers. It inquires of Teslas server, same as the Tesla App. This is where the data comes from. By looking at the timestamps of installs, you can see roughly when the batches are added to the pending pile, because the number of installs per minute goes from fractional to multiple. It does take maybe an hour for TeslaFi to get the full count of Pending after a batch is added.

Anyway, don't sweat the 1 check per day. When your time comes, it'll appear in your app, and you can click to install when you are ready.

rlsd · Dec 18, 2023

DrGriz said:
The actual Lending Tree article starts with a disclaimer that the content is based on the opinion of the author and hasn't been "reviewed."

That's not to say it isn't accurate, but it's hard to say how things like miles driven affect the results, for instance, just at first glance.

Here is another study using more than 4 million car insurance applications:

https://insurifycdn.com/wp-content/uploads/2021/07/car-makes-with-most-accidents-2021-1.png

and

Car Models With the Most Accidents in 2023 - Insurify

Insurify identified 2023’s most accident-prone vehicles. Sports cars like the Audi S4 and Subaru WRX sit atop the list, but some unexpected models made the top 10.

insurify.com

and

These Are The 5 Most Accident-Prone Car Brands In America | Carscoops

Data on own-fault mishaps suggests giving RAM 3500 owners a seriously wide berth

www.carscoops.com

Which one should we believe?

swedge · Dec 18, 2023

rlsd said:
Here is another study using more than 4 million car insurance applications:

https://insurifycdn.com/wp-content/uploads/2021/07/car-makes-with-most-accidents-2021-1.png

and

Car Models With the Most Accidents in 2023 - Insurify

Insurify identified 2023’s most accident-prone vehicles. Sports cars like the Audi S4 and Subaru WRX sit atop the list, but some unexpected models made the top 10.

insurify.com

and

These Are The 5 Most Accident-Prone Car Brands In America | Carscoops

Data on own-fault mishaps suggests giving RAM 3500 owners a seriously wide berth

www.carscoops.com

Which one should we believe?

Here is what Todd Snyder has to say about this: (link)

64 percent of all the world's statistics are made up right there on the spot
82.4 percent of people believe 'em whether they're accurate statistics or not
I don't know what you believe but I do know there's no doubt
I need another double shot of something 90 proof
I got too much to think about

swedge · Dec 18, 2023

kabin said:
Still wrapping my head around the Forbes article showing Teslas have the highest crash rate of 30 auto brands. It's a curious stat normalizing to drivers versus miles. Apparently LendingTree looked at tens of millions of quotes over a 1 year period ending Nov 2023. In any event, accidents per 1000 drivers: Tesla 23.54, Ram 22.76, and Subaru 20.90.

The article ties it to the recent AP recall.

Tesla Has The Highest Accident Rate Of Any Auto Brand

What if Tesla drivers drive more miles per year than Ram drivers? I mean besides infinitely lower fuel cost. Oh, right, lower accidents per driver.

Then there is selection bias: which of Tesla drivers use Lending Tree?

swedge · Dec 18, 2023

Mardak said:
2023.44.30.4 just now to over 2000 TeslaFi vehicles.

That is interesting news! It happened while I was typing...

It looks like they think .4 is clean. They are pushing 44.30.x out very quickly. Maybe us lowly testers will be home for Christmas after all.

Of course the download is likely to become available to download as soon as we leave to drive to my sister's house. Not that I'm waiting up to hear reindeer hooves or anything. Last year's Santa Visualization will do for the drive, as long as it works with FSD...

aataskin · Dec 19, 2023

swedge said:
What if Tesla drivers drive more miles per year than Ram drivers? I mean besides infinitely lower fuel cost. Oh, right, lower accidents per driver.

Then there is selection bias: which of Tesla drivers use Lending Tree?

Or simply that Tesla's automatic collision prevention systems (those relying on vision only) are not adequate enough - especially once you control for the fact that it's a much more fun car to drive than a Ram. Honestly I wouldn't read much into this. You can't compare accident rate between a sports car and a truck.

Frankb70 · Dec 19, 2023

rlsd said:
Poor people who already received 2023.44.30.1. Now you have to install 2023.4.4.30.2 again unless you manage to block it and wait for 2023.44.30.3 tomorrow.

I received the 2023.44.30.2 software upgrade, I was longing for the updated for an improved park assist on vehicles fitted without USS sensors which up to now was tonally useless and could not be relied upon. Well sadly the new improved holiday (Christmas's) software upgrade can only be described as as a poor cracker joke). Whilst I was initially impressed by the surrounding detail on the updated park assist, the bubble quickly burst as it as useful as the previous version, totally unreliable. Drove out of my garage and it did not pick up the council refuge bin which is 1200 x 400 x 400 cm even when I made contact, so Tesla please bring back the USS sensors and would recommend owners that have USS sensors on their vehicles if Vision is released for vehicles with USS do not install. I strongly believe Tesla Vision needs to visit Speck Savers for a vision eye test.
I attach photos.

Frankb70 · Dec 19, 2023

jebinc said:
And yet another Tesla fail. I’ll keep my USS, thank you very much.

jebinc, I agree, Camera Vision is useless, I was hoping the 2023.44.30.2 would deliver more accuracy but its got worst could not see a 1200 x 400 x 400 refuge bin so no hope in saving your rims. I understand that its going to be rolled out on vehicles with USS sensors, hope Tesla dont remove the USS.

aronth5 · Dec 19, 2023

WilliamG said:
I mean, you took the words right out my mouth. Literally.

Pdubs said:
This is the first time I’ve seen this word used correctly in Literally 30 years.

Just wondering if "Literally" is always supposed to be capitalized?

Just want to be sure I use it correctly too.

kabin · Dec 19, 2023

archae86 said:
A major differentiator in crash rates is the selection bias of who buys a particular car. How they drive, where they drive...
I remember years ago routinely perusing a set of model-specific crash statistics that was published annually. For several years there was a consistent pattern that cars that were sold in both 2-door and 4-door versions had drastically higher crash rates in the 2-door version (something like 2X!!!). As no one suggested there was a sufficient structural deficiency to cause this, selection bias by driver and use was likely the cause.

I'd revise the claim to say that "Tesla Drivers have the highest crash rate among the drivers of 30 auto brands". That is assuming they actually did the math correctly and are reporting fairly otherwise.

Lots of considerations for sure. As they say, there's lies, damn lies, and statistics.

There's still a chance it's not nefarious. I'm thinking the insurance policies LendingTree had access to included used vehicles with mileage not attributable to the policy holder so they opted to normalize for the driver. One positive with the survey is it's likely the most up to date (Nov 14, 2023) which captures Tesla's most recent production ramp up.

Other surveys look at accident per manufacturer which is biased to the number of vehicles on the road. Subaru is a special case with lots of sales in snow country and the WRX being a long standing sporty vehicle for younger drivers (aka got rice).

jebinc · Dec 19, 2023

Frankb70 said:
jebinc, I agree, Camera Vision is useless, I was hoping the 2023.44.30.2 would deliver more accuracy but its got worst could not see a 1200 x 400 x 400 refuge bin so no hope in saving your rims. I understand that its going to be rolled out on vehicles with USS sensors, hope Tesla dont remove the USS.

Sorry to see/hear about your disappointment. I used to be very up beat about all things Telsa, but that eroded over time from a never-ending string of disappoints and cheap parlor tricks, touted as “mind blowing,” that just are not reliable or practically usable. And, to add insult to injury, Telsa leaves them in that state for years. For example; smart summon, FSD, auto wipers, ARNC, etc. The list of disappointments is long.

powertoold · Dec 19, 2023

It's really a matter of perspective

I was using 11.7.4.3 in the rain at night yesterday, and it blows my mind that it works as well with low visibility (highway recently repaved and reflective in the night rain) as it does on a clear day

Tbh, fsd still blows my mind routinely, as a software and tech guy

Yes, it may seem like L4 is far away, but for me, it's closer than we think

FSD Beta v11.x

Member

Active Member

Active Member

Active Member

Well-Known Member

Member

Active Member

Active Member

Member

Active Member

Member

Member

Member

Member

Member

Attachments

Member

Attachments

Long Time Follower

Active Member

Well-Known Member

Active Member

Similar threads