Tesla, TSLA & the Investment World: the Perpetual Investors' Roundtable

[Knightshade]

For clarity's sake, when discussing the NHTSA data, ADS is level 3 and up, ADAS is level 2.

My post made no mention of ADS, and there are 0 ADS systems in consumer vehicles on US roads today- so not sure what you're pointing out the distinction for here?

Sort of disingenuous to conflate geofencing with accident rates when it's also cherry picking the easiest/ safest regions.

On the contrary- that's entirely the point.

The lower rates from legacy are almost certainly in part, possibly in large part, due to exactly that restriction.

A restriction NTSB has been screaming for Tesla to implement for over 5 years now-- NHTSA was resistant to compelling that- but is now collecting data that might serve as an excuse to do so.

"Your manual explicitly says this system is only intended for highways-- your data shows it has a higher accident rate elsewhere- why do you not geofence it to highways like everyone else with lower rates?"

 

[ZachF]

Here’s a 5 year rolling average motor Vehicle production graph for European countries:

(1954-2021)

 

[Prunesquallor]

Come on! Get with the Program.
Anything less than 0.75% tanks the market.
Anything more than 0.75% tanks the market.
In fact, exactly a 0.75% increase tanks the market as well.
The media will parade out their “experts” minutes after the rate announcement to tell us how this is all horrible.

Reminds me of the Second Law of Thermodynamics:
1) You can’t win
2) You can’t break even
3) You can’t quit the game

 

[jhm]

I have us crossing $420 somewhere between Labor Day and All Hallows Eve.

This is after the 3-to-1 stock split, right?

 

[Discoducky]

I should always explain acronyms. Engineering Change Order (ECO). This is how you change something on the production line. ECOs are usually done by legacy ICE in big batches usually occurring quarterly at a maximum or yearly at a minimum. Tesla does several per day; Elon has mentioned this publicly.

Spoiler: Engineering stuff

The cabling looks a bit crazy in Munro's Rivian teardown. I'd bet that there are significantly more cables than even the original Model S. In terms of overall distance of cabling, routing points and terminations/relays. These add cost, complexity and manufacture time. And overtime, additional points of failure lead to a lower quality product. They are indicative of a set of engineering and design challenges that could have had better outcomes. And it will get better for them, they will make big changes, they have to and very quickly. To change such a massive amount of issues with cabling means several teams need to work together better. You need design goals like reduce overall cabling by X%, reduce cooling/heating loops to a single loop or reduce weight by Y%. Major changes like these need to be done in layers or you risk massive problems with endless diagnostic tracing (I wonder how long it will take for Ford to trace how the Mach e contactors are failing and I wonder if the Lightning uses the same parts/design? I wonder how well these parts communicate and at what depth verbosity? How easy is it to see the whole data picture and then pinpoint an area to dive deep into to find root cause? What kinds of diagnostic tools have they developed?)

If this had come out with the Leaf in 2010 then great, but the world this truck enters now has higher EV standards (e.g. Octovalve, Superbottle, same/shared infrastructure/design between major parts)

Funny, Munro just posted a wiring video which covers some of what I mentioned and more not good stuff (very tight bends, looping, extra long unnecessary runs, orphaned cables with connectors, extra clips...etc). At the end it is mentioned that they will continue to deep dive into wiring so more fun stuff left to uncover.

What was covered in the video was that the R1T has dual 12V batteries and possibly redundancy at the 12V level. This could explain much of the wiring spaghetti. Huge overweight issues however abound!

If anyone wants to talk about this more, page me in the engineering thread...

 

[TheTalkingMule]

This is after the 3-to-1 stock split, right?

Correct.

 

[Huskyf]

Interesting quote in that article, very negative regarding Tesla.

View attachment 816863

As always, follow the money.

View attachment 816858

View attachment 816859

No radar not a problem In full self driving with fsd beta they are the best only one crash if you compare to Waymo 62 or others

 

[mongo]

From the report, it's citing:



In other news, in the last year, more iPhones have been reported to have had to be rebooted than Motorola Bricks.

Report seems more rational than yesterday's news blurb:

Summary Incident Report Data Are Not Normalized Reporting entities are not required to submit information regarding the number of vehicles they have manufactured, the number of vehicles they are operating, or the distances traveled by those vehicles. Data required to contextualize the incident rates are limited. Data regarding the number of crashes reported for any given manufacturer or operator, therefore, have not been normalized or adjusted by any measure of exposure, including the operational driving domain or vehicle miles traveled. For example, a reporting entity could report an absolute number of crashes that is higher than another reporting entity but operate a higher number of vehicles for many more miles.

Per the ADAS Level 2 data, Tesla reported 273 crashes, that's better than the national average by quite a bit seeing as the fleet size is more than 16k cars.

My post made no mention of ADS, and there are 0 ADS systems in consumer vehicles on US roads today- so not sure what you're pointing out the distinction for here?

Waymo is ADS and geofenced. Other systems are Level 3 and geofenced. So your calling out of ADAS manufacture's using geofenced systems was murky to me. (Actually, I'm having trouble thinking of a non-Tesla L2 geofenced system)

On the contrary- that's entirely the point.

The lower rates from legacy are almost certainly in part, possibly in large part, due to exactly that restriction.

A restriction NTSB has been screaming for Tesla to implement for over 5 years now-- NHTSA was resistant to compelling that- but is now collecting data that might serve as an excuse to do so.

"Your manual explicitly says this system is only intended for highways-- your data shows it has a higher accident rate elsewhere- why do you not geofence it to highways like everyone else with lower rates?"

Yah, lower rates on it's own doesn't mean a better system. I've missed less three pointers than Michael Jordan, am I a better basketball player?
A system that can only operate on straight, limited access roads at fixed speeds should have lower crashes. Allowing operation in higher risk areas will result in more crashes regardless of if those areas are expressly allowed in a geofence.

Preventing certain operational modes in unintended areas is a separate issue that has existed since speed limits and classic cruise control; but I see that as separate from system effectiveness.

 

[Huskyf]

I'm guessing the NHTSA report is out?

U.S. safety agency says Tesla accounts for most driver-assist crashes, but warns data lacks context

Federal safety officials stressed that the data is not meant to portray whether one advanced driver-assist system is safer than another.

www.cnbc.com

Surprised the CNBC article is so tempered.

And is not bad for Tesla in full self driving with fsd beta they are the best only one crash if you compare to Waymo 62 or others

 

[Singuy]

It could ALSO be the fact they only allow the system to work on a really limited set of ultra-safe, ultra-mapped, roads- where ALL accidents happen less often.

Supercruise can't work on unmapped roads- period. So there's no option to "allow" it to in hopes it's still safer than just a human.






Which again is data legacy has no way to collect.

And I don't believe NHTSA can force them to do so.... certainly not on existing vehicles- and I don't think they could compel real-time OTA data collection either if the company lacks the back-end to do so. Remember in this case they compelled car makers to provide all available data-- not "go invent systems to collect and report data you don't have access to do today"

As OEMs move toward some version of always-connected OTA capabilities that will change- and that data will start to get fed in because it'll be possible to collect then-- but you can't do much here on existing fleet.

Either way we don't know what is going on. And that's exactly the point of nthsa right now looking into this. And until other car companies can prove their adas system is safe through data, they can mandate a recall to have them disabled until they get the data. You don't allow cars to be on the road without a crash test to prove the car is safe right? So why should adas system be allowed without any proof that its at least average in safety? And if other car companies can't prove this then it's a sad day for them....

 

[Knightshade]

Waymo is ADS and geofenced.

And has nothing to do with the discussion since those aren't consumer vehicles one can purchase.

So your calling out of ADAS manufacture's using geofenced systems was murky to me.

Still not sure why.

(Actually, I'm having trouble thinking of a non-Tesla L2 geofenced system)

...most of them?

Supercruise was cited repeatedly as one example- it only operates on specific roads pre-mapped with LIDAR. Same with Fords and others system that are highway only systems- if you're not on the right type of road the system simply won't activate.


As I mention- NTSB has been insisting Tesla impose such a restriction for a long time- and especially every time there's an accident outside of the stated ODD of the system- but Tesla has shown no interest in doing so.... (it obviously benefits them greatly to collect data from all road types, even those AP isn't intended for use on)

Preventing certain operational modes in unintended areas is a separate issue that has existed since speed limits and classic cruise control; but I see that as separate from system effectiveness.

Seems pretty directly related though. Limiting operation to only those roads you know your system is most effectively obviously results in an increased score on effectiveness. Especially compared to another system that does not limit that way.

 

[Knightshade]

Either way we don't know what is going on. And that's exactly the point of nthsa right now looking into this. And until other car companies can prove their adas system is safe through data, they can mandate a recall to have them disabled until they get the data. You don't allow cars to be on the road without a crash test to prove the car is safe right?

That's not how any of that works.

You need data showing it's DANGEROUS for there to be a recall.

Their current data suggests they're instead very safe.


It's possible they're so safe because they're so restrictive on where you can use them of course- and we lack the data to be sure of THAT. But there's nothing in there that would remotely support a "recall" of such a system.

 

[bkp_duke]

Or... they insist Tesla geo-fence their features to lock them out in areas that Tesla themselves states in the manual you're not intended to be using them.

And point at the fact every other ADAS maker does that and has much lower accident rates.


NTSB has been banging that exact drum since at least 2016.


(this issue goes away whenever there's wide release of the One Stack To Rule Them All I suppose, but even then only for folks with FSD)

Actually, according to the data released by NHTSA today, Tesla probably has the lowest accident rate. They have something like 20X the number of vehicles on the road as the next competitor. Then you factor in miles driven with ADAS engaged, and their accident rate is a fraction of nearly everyone else.

Again, I'm incredibly surprised this was painted by CNBC in a much worse light.

 

[Knightshade]

Actually, according to the data released by NHTSA today, Tesla probably has the lowest accident rate. They have something like 20X the number of vehicles on the road as the next competitor

Actually that's not at all how they measured it.

their data is accidents per 1000 cars with an ADAS system.

So the # of cars is already normalized out.


That's one of the major objections we've been discussing- that the measurement standard is fairly arbitrary and not super useful and makes Tesla look significantly worse than it likely is because the other car makers heavily restrict where you can use their systems and Tesla does not.

We'd need data on things like:

Accidents per X miles
% of miles drive on each road type with ADAS on/off
And the corresponding accident rates to above

We have none of that though and neither does NHTSA. (and outside of Tesla, most of that data doesn't even exist since legacy has no routine collection or reporting of it- and even Tesla does not publicly report the last 2 out of those 3 in any detail-- which has been a common criticism of Teslas quarterly safety stats- they don't tell you how much of the miles are highway vs city in any of the categories they report or what % is driven one way or another)

 

[Singuy]

That's not how any of that works.

You need data showing it's DANGEROUS for there to be a recall.

Their current data suggests they're instead very safe.


It's possible they're so safe because they're so restrictive on where you can use them of course- and we lack the data to be sure of THAT. But there's nothing in there that would remotely support a "recall" of such a system.

Their report literally said no conclusions can be made because they don't have the proper data. And this is the underlying problem specified in the report. So in order to conclude anything, they will need proper data. So either they drop investigation or dig deeper which then mandates proper data. This is exactly the trap card Tesla unleashed for these type of investigations.

 

[Cobbler]

Hmmm.....
Which companies have high debts caused by the financing loan department?

 

[Singuy]

2 recalls and auto loan defaulting. Stocks is like "cool story bro" and continues to shoot up.

View attachment 816907

Hmmm.....
Which companies have high debts caused by the financing loan department?

 

[Knightshade]

Their report literally said no conclusions can be made because they don't have the proper data. And this is the underlying problem specified in the report. So in order to conclude anything, they will need proper data.

Right, which is why in what you quote, I said "we lack the data to be sure"

And why your suggestion this could lead to recalls of legacy systems made no sense- since even NHTSA said there wasn't enough data to draw any useful conclusions.

So either they drop investigation or dig deeper which then mandates proper data. This is exactly the trap card Tesla unleashed for these type of investigations.

Naah. It's not that they didn't ask for the right data. it's that the data does not exist

You can't mandate providing non-existent data.


You COULD mandate "All cars made after X date must collect Y data"

that's how we got mandated EDRs in cars in the first place.

But such mandates take years to create, develop, test, and then years more to actually be required in new vehicles- and never in old ones.

NHTSA began study of EDRs in 1998, but did not issue POSSIBLE mandated rules till 2004-2006 era, and did not get around to requiring them in new vehicles until 2012 model year.

14 years from start to actual new cars required to support it.

Heck it took longer than that to mandate ABS, a system that clearly made cars safer from the get-go.


No current legacy vehicle on the road- nor any current ADAS system from them on the road- will still be shipping by the time they'd be REQUIRED to collect and provide the type of data we're talking about.


It's certainly possible this is NHTSA starting to move that direction.... but this isn't any sort of "tesla trap" because the date by which legacy cars would need to actually do anything is beyond the 2035 date a lot of folks here think those companies won't even be in business anymore.

Industry-wide vehicle safety mandates move at glacial pace.

 

[mongo]

Actually that's not at all how they measured it.

their data is accidents per 1000 cars with an ADAS system.

So the # of cars is already normalized out.

Where does NHTSA call that out?

I can't find "1,000" anywhere on NHTSA.
From the report I pulled:

Summary Incident Report Data Are Not Normalized Reporting entities are not required to submit information regarding the number of vehicles they have manufactured, the number of vehicles they are operating, or the distances traveled by those vehicles. Data required to contextualize the incident rates are limited. Data regarding the number of crashes reported for any given manufacturer or operator, therefore, have not been normalized or adjusted by any measure of exposure, including the operational driving domain or vehicle miles traveled. For example, a reporting entity could report an absolute number of crashes that is higher than another reporting entity but operate a higher number of vehicles for many more miles.

 

[bkp_duke]

Actually that's not at all how they measured it.

their data is accidents per 1000 cars with an ADAS system.

So the # of cars is already normalized out.


That's one of the major objections we've been discussing- that the measurement standard is fairly arbitrary and not super useful and makes Tesla look significantly worse than it likely is because the other car makers heavily restrict where you can use their systems and Tesla does not.

We'd need data on things like:

Accidents per X miles
% of miles drive on each road type with ADAS on/off
And the corresponding accident rates to above

We have none of that though and neither does NHTSA. (and outside of Tesla, most of that data doesn't even exist since legacy has no routine collection or reporting of it- and even Tesla does not publicly report the last 2 out of those 3 in any detail-- which has been a common criticism of Teslas quarterly safety stats- they don't tell you how much of the miles are highway vs city in any of the categories they report or what % is driven one way or another)

I know HOW they measured it. My point was that any REASONABLE measurement should involve accidents per mile driven, and that makes Tesla's far safer than any other auto. Unfortunately, Tesla is probably the only company that has that data.

I don't need you to walk me through the simple math, it was an "argh this is stupidity on the part of NHTSA" post, that's all it was.