Safety Score: parameter dependence

[Michael_Vinson]

So I need to work on my unsafe following parameter.

Am I correct in understanding that I will only be able to improve it by driving without AP and between 1-3 sec following distance?

Doesn't seem like maintaining a 3+ distance or AP helps this parameter in any way.

I agree, Che. Greater than 3 seconds, or less than 50mph, or driving in AP, or any combination thereof will have no effect on your UFT. But it's also the least impactful on your total score. You can get your UFT up to 22% and still get a perfect 100. Indeed, if all your other parameters are zero, and your UFT is 100%, your total score for that day will be 98! That's higher than if you have just 2% on Hard Braking and everything else is zero. So my advice is not to sweat the UFT. And forced AP disengagements are also easy to avoid (although more impactful on your score).

However, be warned that my total score is a wretched 88, so what do I know?

Michael

 

[AlanSubie4Life]

But the real bitch is that I was in manual (and subject to scoring) only for less than 1/4 mile. So evidently that 1/4 mile was used to assign 60% Unsafe Following to the whole 28.8 mile trip, of which 28.6 was entirely in NOA. That's not right.

Totally expected and correct. Someone flashing across your bow for a 1/2 second will take you to 60% in this case. The problem is that you used NOA. You have to bank good events before you transition to NOA.

Fortunately close following doesn’t matter much and it is also easy to get it down to 5% or whatever you want, from 60% (on the same day).

 

[ATPMSD]

This has been an entertaining thread! What we seem to miss is that this is a first release and flaws are to be expected - and there seem to be a lot of them. What bothers me is not that the Safety Score systems needs to be adjusted, but that Tesla is using a clearly flawed app to make judgments on who to release the beta too.

If Tesla wants to rollout the new FSD slowly (which really means they are breaking yet another delivery promise) then it should be rolled out based on when you purchased FSD i.e. those who have been waiting the longest should get it first.

 

[AlanSubie4Life]

then it should be rolled out based on when you purchased FSD i.e. those who have been waiting the longest should get it first.

Whether this is the right call depends on whether “wait time for FSD” is a better predictor of safety & driver attentiveness while monitoring the beta of FSD Beta, than the Safety Score.

 

[ATPMSD]

“wait time for FSD” is a better predictor

Obviously wait time predicts nothing about driver safety, but it is a matter of principle and showing respect for those who joined the program early. Which, by the way, is not me as I got my car just a year ago.

 

[AlanSubie4Life]

Obviously wait time predicts nothing about driver safety, but it is a matter of principle and showing respect for those who joined the program early.

What is being delivered is not what anyone who purchased FSD paid for, though. This is the beta of FSD Beta.

So it's not clear that delivering a product which is not ready to the early customers, who expect a completed product, who then overestimate its capabilities and have accidents, resulting in the completed product never being delivered, serves the best interests of those who joined the program early.

That being said, anyone with a vehicle with the hardware capable of doing so (big caveat), who purchased early, is able to join the beta of FSD Beta if they want to, as long as they wish to participate and decide to excel at the scoring. The current Safety Score program doesn't appear that it will exclude anyone who wants to join (with the prior caveats). We'll know tomorrow!

 

[EVNow]

If Tesla wants to rollout the new FSD slowly (which really means they are breaking yet another delivery promise) then it should be rolled out based on when you purchased FSD i.e. those who have been waiting the longest should get it first.

I'd have no problems if Tesla implements this way with a safety score cut off. Take everyone with 95%+ safety score and then deliver based on when they purchased. I think it would have been equally fair -if they had decided on that and communicated in the beginning.

 

[brianman]

Obviously wait time predicts nothing about driver safety, but it is a matter of principle and showing respect for those who joined the program early. Which, by the way, is not me as I got my car just a year ago.

Speaking as a previous owner of AP1 (traded in), previous owner of FSD with AP2.x hardware (traded in), and now current owner 2x FSD with AP3 hardware... I'm pretty sure there's no way they would be willing to properly "show respect" for how early I "invested" in AP and FSD with Tesla.

"But but you traded those in so those don't matter"
Perhaps, but Tesla (corp.) has always had the philosophy of "trade in because we don't really support upgrading your vehicle" (or upgrading your battery or ...).

 

[aronth5]

I'd have no problems if Tesla implements this way with a safety score cut off. Take everyone with 95%+ safety score and then deliver based on when they purchased. I think it would have been equally fair -f they had decided on that and communicated in the beginning.

Agree, and for those who where not able to transfer the FSD software from one Tesla to a new Tesla since they paid twice they should be at the top of the list.

 

[ATPMSD]

What is being delivered is not what anyone who purchased FSD paid for, though. This is the beta of FSD Beta.

Respectually disagree. What we have in the car is “Navigate on Autopilot (BETA)”, Telsa has yet to deliver the “production” version of this. Elon keeps putting it off an now we are talking about FSD Beta, so at this point this is exactly what everyone purchased. Also, keep in mind that “eventually” everyone will get the FSD Beta and, someday, the production version. So unless he is going to send out a bill we have in fact paid for this.

 

[AlanSubie4Life]

now we are talking about FSD Beta

No, we're talking about the beta of FSD Beta. That's the issue.

But anyway, I certainly understand people wanting what they paid for. I'm not arguing that people should not be getting that. It's just not clear when that product will exist. But there are a lot of discussions about that elsewhere. I guess we should stick to parameter dependence in this thread. Sorry for getting OT.

 

[EVNow]

so at this point this is exactly what everyone purchased.

No. What a person purchased is whatever was on the website at the time the purchase was made.

 

[KArnold]

Obviously wait time predicts nothing about driver safety,

Sure it does. Chances are extremely high they have less at-fault accident-free miles in a Tesla than someone who bought a month ago. By definition, they are the safest among us in this car. Would you prefer a surgeon who has done 3 perfect surgeries, or one who has done a thousand? Is the 3-try guy less safe than the 1000-try guy? Maybe not, but I'd put my life in the hands of the latter every time as statistically, he has a better skillset and less likely to screw up.

 

[brianman]

Respectually disagree. What we have in the car is “Navigate on Autopilot (BETA)”, Telsa has yet to deliver the “production” version of this. Elon keeps putting it off an now we are talking about FSD Beta, so at this point this is exactly what everyone purchased. Also, keep in mind that “eventually” everyone will get the FSD Beta and, someday, the production version. So unless he is going to send out a bill we have in fact paid for this.

I think you're actually in agreement but missing each other on phrasing.

The "1000ish" people that get the download at midnight PST tonight are not "getting what they purchased / paid for". It's less than was paid for.

I think the argument you're trying to make is that "people that paid for FSD are overdue for getting 'whatever Tesla has' in the FSD space offered to them". Some will disagree, some won't. But either way it's a different point entirely from whether the limited 10.2 rollout is "what was paid for". It is most definitely not. And if Tesla tries to "claim delivery" of "paid for FSD", some legal fireworks will likely ensue.

 

[Sparrow]

Still trying to understand how this all works. If we look at Agressive Turning for example. From what I am hearing driving 10 miles with 1 perfect turn would be equal to driving 10 miles with 10 perfect turns. At the end of the day you essentially book 10 miles with a perfect score in both cases.

The only time you would want to have alot of turns is if you made 1 or more bad ones that same day. Then you can dilute that one bad turn with a bunch of good turns, but only on the same day as the bad turn.

I then assume the day is then weighted by the total number of miles driven. So a 100 mile perfect turning day is better than a 10 mile day. Surely, they do not weight those 2 days as the same do they?

Do I have any of this right?

 

[AlanSubie4Life]

Still trying to understand how this all works. If we look at Agressive Turning for example. From what I am hearing driving 10 miles with 1 perfect turn would be equal to driving 10 miles with 10 perfect turns. At the end of the day you essentially book 10 miles with a perfect score in both cases.

The only time you would want to have alot of turns is if you made 1 or more bad ones that same day. Then you can dilute that one bad turn with a bunch of good turns, but only on the same day as the bad turn.

I then assume the day is then weighted by the total number of miles driven. So a 100 mile perfect turning day is better than a 10 mile day. Surely, they do not weight those 2 days as the same do they?

Do I have any of this right?

That is all correct.

Booking more good turns and such on a perfect day may make your overall average on the main page look better (that is the only thing they would affect I think - though these aggregate numbers are not well understood - I’m making an assumption here), but those numbers on the main page are definitely not used for the calculation of your score. So those good turns do not matter unless you need to dilute a “same-day” bad turn.

 

[Sparrow]

Going the opposite direction, if somehow you were able to drive 100 miles in a day and you made only 1 turn that counted and it turned out to be aggressive, that would be a 100% Aggressive Turning for that day. If on another day you drove only 10 miles and still had only one turn that counted and it too was considered Aggressive, that would also be a 100% Aggressive turn day. If I am understanding the metrics, the 100 mile day with the 100% Aggressive Turning scores would be much worse for your cumulative Safety Score than the 10 miles day with 100% Aggressive Turning.

Does that reasoning hold too?

 

[AlanSubie4Life]

Going the opposite direction, if somehow you were able to drive 100 miles in a day and you made only 1 turn that counted and it turned out to be aggressive, that would be a 100% Aggressive Turning for that day. If on another day you drove only 10 miles and still had only one turn that counted and it too was considered Aggressive, that would also be a 100% Aggressive turn day. If I am understanding the metrics, the 100 mile day with the 100% Aggressive Turning scores would be much worse for your cumulative Safety Score than the 10 miles day with 100% Aggressive Turning.

Does that reasoning hold too?

Yes. Though aggressive turning is capped at 17.1% allegedly, so that’s what you’ll end up at. Which is catastrophic (94). The 100-mile day would be much more catastrophic though.

The idea in general would be to bank a lot of good turns in your 100-mile trip to make sure one bad one does not destroy your score.

This is more and more important to do for all the three % metrics, the longer the day. You really want to bank good events so you don’t have to spend dedicated time “repairing” your score in the evening (which is insane).

This is an argument for using AP judiciously in cases only where you need help avoiding violations, since you cannot bank any events on AP, good or bad. It potentially opens you up to catastrophe. Live by the sword, die by the sword, basically. AP could be great, but if you get unlucky off of AP, look out.

 

[ATPMSD]

So what the Safety Score is teaching us is how to game the system, which makes the score even less valid.

“How about a nice game of chess?” - War Games

 

[rickm8888]

So, wait. I leased my car (2 year lease) 9 months ago. If I never get FSD during the lifetime of my lease...I assume, Tesla being Tesla, there will be no refund for the FSD that I never get.