Safety Score

[EVNow]

Have all of the beta releases been on a Friday night? Off the top of my head, I can't recall.

Just the other week they delayed 10.0.1 by 2 days.

 

[MarcG]

I agree and have put a note on the sheet to that effect. In my case the Tesla app score changed to 100 when I reached 99.50 on the sheet, but others like you needed 99.51. I'm confident the formulas are correct but attribute this to some very small rounding error based on the individuals combination of bad events.

My experience shows that 99.5001 rounded up to 100.

 

[dcesq06]

Just the other week they delayed 10.0.1 by 2 days.

That's right! Thanks. Let's hope that is the case this time.

 

[jebinc]

I'm thinking good thoughts for those with scores that are within striking distance of 100 and hoping they can get there before the delayed release of 10.2 (and that Tesla then includes them). If you'd like to plan how many miles you need to drive consider using my spreadsheet (or other similar sheets) for your driving strategy. It helped me and I've gotten great feedback that it's been helpful to others. There are full instructions on the sheet as shown on the screenshot below the link. Good luck and good driving.

Tesla Safety Score - Tracking & Planning v10

Safety Score This is set to read only. To make your own sheet go to the file menu, 'Make a Copy' and open in Google Sheets (not in Excel). No share requests please. There are hidden columns with some of the working formulas so be careful what you change or move. <a href="https://teslamotorsclu...

docs.google.com

View attachment 719496

I think, by the time they release the software, whenever, that there will be more 100s then they need or want - resulting in another.....

 

[jebinc]

You haven’t wasted your money. You invested it. The price of FSD is going to go up.

In Elon we trust, right? … RIGHT!?

It's not an investment if you can't sell, trade or transfer it.....

 

[DaveG_NJ]

As an aside, I find it funny that the OG Beta Testers refer to us as the "button-pushers" on Twitter , kinda like the riff-raff from the cheap seats coming down and sitting in the box seats.

 

[jebinc]

By all of that being known up front.

Nobody's having a gun held to their head to buy it after all. If you failed to check into what you paid for before you paid for it, that's really on you.

Also- it's only been 10k for FSD for less than a year. It's been as cheap as 2k-3k for most of its existence.

@Knightshade Well, I feel this is only partially true. When I purchased FSD the first time, back in 2019, it was advertised on the ordering page as having certain features "by the end of the year;" features that never materialized - so we were mislead at the time of purchase, somewhat. For me, some of those items listed on the ordering page back then (2019) have still not materialized, like city driving. What did, is complete rubbish ("Smart Summon," which is really stupid drunk summon, in perpetual "Beta," but really pre-alpha, form). YOMV (Where "O" = Opinions)

 

[Daniel in SD]

Can you show your work here?

How do you know more than Tesla about the amount of data they need to make this judgement?

Remember this isn't them needing to decide if the software is safe- it's the judge if the DRIVERS are.

You don't need millions of miles of edge cases to check things like "Is driver inattention significantly higher among new testers than old testers"

They're trying to measure how safe the drivers are when performing a task (supervising FSD Beta software) that they've never done before and for which there is very little past experience to make predictions.
I just don't think you can measure safety without measuring collision rate. Maybe you could extrapolate from metrics like driver attention but that's only after you have a model that translates that to collision rate for which you need collision rate data. Another factor is that collisions vary wildly in severity so I would be skeptical about making predictions about fatal collisions by measuring curb strikes.

 

[JosiahLuscher]

What calculator?

Ski

Safety Score Calculator

safety.sfnet.app

 

[Knightshade]

They're trying to measure how safe the drivers are when performing a task (supervising FSD Beta software) that they've never done before and for which there is very little past experience to make predictions.

They have an entire year of original FSD testers (roughly 2k of them, so double the amount of the added group)

So they have a baseline already for all these things.

I just don't think you can measure safety without measuring collision rate.

...what?

AFAIK the collision rate from a year of FSDBeta among 2000 people is... 0.

Maybe you could extrapolate from metrics like driver attention but that's only after you have a model that translates that to collision rate for which you need collision rate data. Another factor is that collisions vary wildly in severity so I would be skeptical about making predictions about fatal collisions by measuring curb strikes.

Again this isn't making any sense.

They're not looking for how safely the FSDBeta drives in a couple of days. The current collision rate is ZERO. The idea they can't know ANYTHING ABOUT ANYTHING without collisions doesn't seem to be based on... anything.

They're looking for how well the human monitors the system.

There's multiple ways to do that WITHOUT collisions happening.

Driver attention is one I mentioned.

Rate of disengagements and/or interventions (since they're not quite the same thing) are others.


Compare those to the year of data they already have from testers.

If the new folks seem to be in line, great- all good- keep expanding.

If the new folks seem significantly worse in some way, pause adding people to address whatever differences you're finding.

@Knightshade Well, I feel this is only partially true. When I purchased FSD the first time, back in 2019, it was advertised on the ordering page as having certain features "by the end of the year;" features that never materialized - so we were mislead at the time of purchase, somewhat. For me, some of those items listed on the ordering page back then (2019) have still not materialized, like city driving. What did, is complete rubbish ("Smart Summon," which is really stupid drunk summon, in perpetual "Beta," but really pre-alpha, form). YOMV (Where "O" = Opinions)

So- 2 things.

1) Your comment regarding city streets "by end of the year" is absolutely fair and accurate.

And in the various threads where folks have blustered about lawsuits and such, the one valid basis for one anybody has ever cited and with which I've agreed.


2) That said- all the things the poster I was replying to was complaining about were- it was non-transferable (which has been true since day 1 and Tesla has always been clear about), that it "might never come" which is true for anything not yet finished, and no definitive deadline (ironically, city streets DID have one in 2019- it just missed it).

It's not an investment if you can't sell, trade or transfer it.....

It's attached to the car.

Which you can sell, trade, or transfer.

 

[qdeathstar]

It's not an investment if you can't sell, trade or transfer it.....

Yes it is. You can’t sell trade or transfer your salvation but choosing to not spend the evening with your mistress is still an investment in it.

 

[insaneoctane]

My family is going to be SO happy with me.
I've already banned my wife from driving the Tesla for 2 weeks now.
Yesterday when it was announced that 100 scores were getting beta, I parked the car and nobody drove it since it was at 100 and the last thing I needed was to lower my score especially since the 1,000 per day rollout was delayed if you weren't at 100....Now that the goal posts have moved again, I'm going to be Mr. Popular around here! I think I'll keep it garaged another couple of days. I do have 567 cumulative miles, so it's not "gamed", I'm just trying to get this car back into the family rotation (with beta for me) as quickly as possible. Grrrrrr....

 

[Ramphex]

Can you show your work here?

How do you know more than Tesla about the amount of data they need to make this judgement?

Remember this isn't them needing to decide if the software is safe- it's the judge if the DRIVERS are.

You don't need millions of miles of edge cases to check things like "Is driver inattention significantly higher among new testers than old testers"







By all of that being known up front.

Nobody's having a gun held to their head to buy it after all. If you failed to check into what you paid for before you paid for it, that's really on you.

Also- it's only been 10k for FSD for less than a year. It's been as cheap as 2k-3k for most of its existence.

Known up front sure, the difference between me wasting my money and him taking my money is he knows the timeline, status, roadmap, etc, while we’re just hoping for the best

 

[EVNow]

It's not an investment if you can't sell, trade or transfer it.....

Still better than all the people who invested in their education but ended up as baristas.

 

[ChimpledPot]

on Teslafi, there’s some nutters who have a 100 score with over 2,000 miles since pressing the button. I’m 46th in terms of mikes accumulated myself and that’s a quarter of the miles of the guys at the top.

 

[brianman]

My family is going to be SO happy with me.
I've already banned my wife from driving the Tesla for 2 weeks now.
Yesterday when it was announced that 100 scores were getting beta, I parked the car and nobody drove it since it was at 100 and the last thing I needed was to lower my score especially since the 1,000 per day rollout was delayed if you weren't at 100....Now that the goal posts have moved again, I'm going to be Mr. Popular around here! I think I'll keep it garaged another couple of days. I do have 567 cumulative miles, so it's not "gamed", I'm just trying to get this car back into the family rotation (with beta for me) as quickly as possible. Grrrrrr....

After the assertive statement at the shareholder's meeting and then the follow-up, I kind of hope the power and internet goes out at the house Elon is staying at until he stops moving the goal posts. Having a delay in the software is fine (IMO) but playing with people on the cut-off for qualification is unnecessary and cruel.

 

[gearchruncher]

So, Tesla expects drivers to be perfect in their safety scores for 7 days.
However, they are testing the software build up until 4AM on a Saturday morning, and only learn then that it's not ready.
But they believe they can make fixes in 48 hours over a weekend, and release this very critical piece of software that if it fails will cause massive PR issues for the company, while asking that their own customers be much more stable than they expect of themsleves.

Tesla went 7 days saying it would be last Friday. Then it was this Friday. Why would anyone believe there is any reasonable chance this ships anytime soon?

 

[queue]

After the assertive statement at the shareholder's meeting and then the follow-up, I kind of hope the power and internet goes out at the house Elon is staying at until he stops moving the goal posts. Having a delay in the software is fine (IMO) but playing with people on the cut-off for qualification is unnecessary and cruel.

yeah, this is messed up, literally who knows how many thousands of people dropped what they had been doing 2 weeks ago to adapt to this mess called a "safety score" and to have it snatched from under your nose at literally the last second. doesn't play well.

 

[Daniel in SD]

AFAIK the collision rate from a year of FSDBeta among 2000 people is... 0.

Over how many miles?

Driver attention is one I mentioned.

Rate of disengagements and/or interventions (since they're not quite the same thing) are others.


Compare those to the year of data they already have from testers.

If the new folks seem to be in line, great- all good- keep expanding.

If the new folks seem significantly worse in some way, pause adding people to address whatever differences you're finding.

What is a worse rate of intervention? I've been arguing that it should be much higher given what we've seen in the beta videos.
They took a year to go from less than 100 public tester to go to over 1000. Why? FSD Beta doesn't really look any safer to me than it did a year ago (not saying it's not better, just that you still have to be hyper vigilant to use it safely). We have so little visibility into their decision making process so I'm going to assume that the rollout will continue to be slow. We'll see what happens!

 

[gearchruncher]

LOL at all the people not using their $50K asset trying to keep a score.
LOL at all the people driving extra miles in their environmentally sensitive car trying to up an arbitrary score.
LOL at all the people realizing that any drive "risks" their score, so maybe this isn't a great measure of you as a driver, and is really just a lottery.

Here's to Monday. Just wonder which Monday.....