Safety Score

[S4WRXTTCS]

Totally agree. I've been driving super-carefully I admit (but on my usual mix of shopping and work trips) and managed to stay at 99%, but all the "dings" I got were justified by "good" (imho) driving such stopping for yellow lights downhill, and once for an ambulance that did a U turn in front of me.

If Tesla genuinely want decent testers they are better off disqualifying the 100% scores as gaming the system (yeah sure, it will outrage many people, but they have already done that anyway).

Tesla needed three different aspects in FSD Beta drivers

Diversity in location
People adept at interpreting when its screwing up, and reporting bugs/issues
Safe Drivers who take over quickly.

I think they could have taken a better approach by adding mechanism to report issues within the existing SW. Then they could have released a limited version of FSD Beta where it didn't do a lot of the dangerous stuff, but where it was all unified (same vision stack used on the freeway as the city).

This way TONS of people could use the reporting to report issues with routing, and other things that will clearly mess up FSD Beta before that element is even turned on. In addition to this it would gave given customers improvements in visualizations, and NoA.

Then they could have used those three aspects to pick people. The approach I would have taken would have been really geo-specific (I'm a strong advocate of good mapping). Like 1000 people in the Portland area or something like that. Where they gather a bunch of data, and they might even turn off the Beta for those people before moving to another area with another 1000 people. Maybe even have a Calendar where local groups could organize events so non-FSD beta people could get a glimpse.

It's an early beta so part of whole opt-in should be an acknowledgement that it might come and go. It should also come with a Quiz to demonstrate that the owner understands all the liability. Nio added this for their L2 system after a recent fatality.

For safety I wouldn't have reported what I used as the mechanism, but it would have been pretty liberal where the only goal was to weed out the terrible. I certainly wouldn't have released a Safety Score system that didn't have situational awareness. The Safety Score does have a long term purpose of insuring a vehicle capable of both L2 and L4 driving. I think that's a big component of the long term plan of Tesla insurance.

 

[mwasylyk]

Tesla needed three different aspects in FSD Beta drivers

Diversity in location
People adept at interpreting when its screwing up, and reporting bugs/issues
Safe Drivers who take over quickly.

I think they could have taken a better approach by adding mechanism to report issues within the existing SW. Then they could have released a limited version of FSD Beta where it didn't do a lot of the dangerous stuff, but where it was all unified (same vision stack used on the freeway as the city).

This way TONS of people could use the reporting to report issues with routing, and other things that will clearly mess up FSD Beta before that element is even turned on. In addition to this it would gave given customers improvements in visualizations, and NoA.

Then they could have used those three aspects to pick people. The approach I would have taken would have been really geo-specific (I'm a strong advocate of good mapping). Like 1000 people in the Portland area or something like that. Where they gather a bunch of data, and they might even turn off the Beta for those people before moving to another area with another 1000 people. Maybe even have a Calendar where local groups could organize events so non-FSD beta people could get a glimpse.

It's an early beta so part of whole opt-in should be an acknowledgement that it might come and go. It should also come with a Quiz to demonstrate that the owner understands all the liability. Nio added this for their L2 system after a recent fatality.

For safety I wouldn't have reported what I used as the mechanism, but it would have been pretty liberal where the only goal was to weed out the terrible. I certainly wouldn't have released a Safety Score system that didn't have situational awareness. The Safety Score does have a long term purpose of insuring a vehicle capable of both L2 and L4 driving. I think that's a big component of the long term plan of Tesla insurance.

Pain and simple, I really think it was just easier for them to just have a score that gave them a metric(super accurate or not) about how safe of a driving environment the car is in to pick the early stages of the rollout. I think the phrasing that it's a safety score and somehow supposed to be representative of your driving skills is what is irking some people. Because a lot of the score at the moment is really just being affected by where you are driving and how able(situational) you are to drive ULTRA slow and careful. Locations where it's easy enough to get 100 are probably locations as a generalization they would be more comfortable handing out the beta, just to reduce risk at first.

 

[ConSaab]

I got mine up from 99 to 100 on the 8th by driving an extra ~120 miles with a 100 trip score yes. That was on the 8th into the 9th early am, doubtful they locked the score from changing, but no one knows if they already picked the first people to roll out to. But I also doubt the selected people already

Thanks for confirming. Did yours update to commutative 100 right after you logged 100 perfect drive? I’ve had perfect 370 with 10 trips but it hasn’t budged….

 

[mwasylyk]

Thanks for confirming. Did yours update to commutative 100 right after you logged 100 perfect drive? I’ve had perfect 370 with 10 trips but it hasn’t budged….

Well I logged a 100 drive with too few miles, the overall score was still 99. Then after the next 100 score drive the overall score updated to 100 instantly yes. Maybe double check the math on how many miles you need to get to 99.5/maybe you're just a handful of miles away from having enough to push it over.

* Also I misspoke I did this on the 7th not the 8th. Not that it should matter *

 

[EVNow]

There are definitely tricky situations (e.g. my video posted here earlier). But it is still not that hard and for the most part these issues can be resolved by driving at the speed limit. I live on a hill at 870 feet and elevation goes to 1000 feet before plunging to 300-500 feet over about one mile.

Yes - similar situation here.

Plus people who live on hills may be higher risk.

Sure - all other things being equal.

I think having to stop quickly because of yellow light etc could still be considered for insurance rating. Afterall from insurance rating perspective there won't be too much of a difference between 95 and 100. Its just for this fsd beta rating "game" that becomes impractical.

If I were to do this all over again - I'd use AP everywhere. I think that takes care of yellow lights and hills.

 

[mwasylyk]

Yes - similar situation here.


Sure - all other things being equal.

I think having to stop quickly because of yellow light etc could still be considered for insurance rating. Afterall from insurance rating perspective there won't be too much of a difference between 95 and 100. Its just for this fsd beta rating "game" that becomes impractical.

If I were to do this all over again - I'd use AP everywhere. I think that takes care of yellow lights and hills.

I never tried it but I know Progressive has that ODB device that monitors your driving statistics to either give you a discount/increase your insurance rate. I'm willing to bet if the raw scores from that monitoring tool were made available, people would be just as outraged over how it was scoring them lol

 

[mwasylyk]

If I were to do this all over again - I'd use AP everywhere. I think that takes care of yellow lights and hills.

Exactly, as soon as I found out that AP miles count towards your overall daily millage that was pretty much all I used. Miles were what I needed to push my score to 100 and I didn't need to improve my current days score so AP was a perfect fit. I didn't read their page on the score well enough the first time and assumed all AP miles were ignored

 

[EVNow]

I never tried it but I know Progressive has that ODB device that monitors your driving statistics to either give you a discount/increase your insurance rate. I'm willing to bet if the raw scores from that monitoring tool were made available, people would be just as outraged over how it was scoring them lol

One of the insurance company collection of stats posted here showed how they collected braking and other stats in a number of tiers (some 10 steps) - that kind of nuanced collection if used properly will give a better rating overall. I'm sure as Tesla rating improved they will figure these things out.

BTW, if they collected $200 a month from some one with 100 rating vs $500 from some one with 99 rating - it would cause outrage. Not if the 99 score person need to pay $10 more per month.

 

[EVNow]

Exactly, as soon as I found out that AP miles count towards your overall daily millage that was pretty much all I used. Miles were what I needed to push my score to 100 and I didn't need to improve my current days score so AP was a perfect fit. I didn't read their page on the score well enough the first time and assumed all AP miles were ignored

The issue was - we weren't sure if infractions under AP in the city counted or not. There were some false alarms in the beginning that created uncertainty (because of the terminology - when used "appropriately".

 

[ConSaab]

Well I logged a 100 drive with too few miles, the overall score was still 99. Then after the next 100 score drive the overall score updated to 100 instantly yes. Maybe double check the math on how many miles you need to get to 99.5/maybe you're just a handful of miles away from having enough to push it over.

* Also I misspoke I did this on the 7th not the 8th. Not that it should matter *

So I share my car with my wife. My profile has perfect drives however the wife got dinged for hard braking couple of time which knocked my score to 99… I’ve been driving in my profile to get it back to 100, you think the app won’t round me to 100 since the wife profile hasn’t driven to 100?

 

[mwasylyk]

One of the insurance company collection of stats posted here showed how they collected braking and other stats in a number of tiers (some 10 steps) - that kind of nuanced collection if used properly will give a better rating overall. I'm sure as Tesla rating improved they will figure these things out.

BTW, if they collected $200 a month from some one with 100 rating vs $500 from some one with 99 rating - it would cause outrage. Not if the 99 score person need to pay $10 more per month.

Of course its not a direct comparison and I didn't mean to say the insurance companies were as exclusive as Tesla is right now FSD, I was just poking fun at the idea. Interesting to know that info is around, I'll have to go find it to see what it looks like

 

[gearchruncher]

BTW, if they collected $200 a month from some one with 100 rating vs $500 from some one with 99 rating - it would cause outrage. Not if the 99 score person need to pay $10 more per month.

I love the idea that you would need to pay more for software that is supposed to drive your car for you at a level substantially safer than you, because when you do drive the car, you're not "safe" enough.

This tells you all you need to know about where FSD is right now. It's entertainment that can only be handled by the very safest drivers. It's not autonomy in any way.

Technically, it should be the other way around. The better a driver you are, the harder it is for FSD to be better than you, so you should have to pay more.

 

[mwasylyk]

So I share my car with my wife. My profile has perfect drives however the wife got dinged for hard braking couple of time which knocked my score to 99… I’ve been driving in my profile to get it back to 100, you think the app won’t round me to 100 since the wife profile hasn’t driven to 100?

AFAIK the profile has nothing to do with it besides showing who drove the miles to you. Is your current daily score a 99 or 100? Because you would need a 100 for the current day you are driving in order to increase your overall score with the mile weighted day score. I know everyone uses this and it worked for me to calculate miles needed to reach 100. Safety Score Calculator

Anyone correct me if I am wrong.

Also if your score for the day is below 100 I think you could use the same formula to calculate how many miles you'd need to drive today to increase it to 100. I am just not sure if the daily score is each trips score(I think you'd have to manually put in each trip into the score simulator to get the score for each trip and write them down) average, or if each trip score is also weighted by miles when calculating the daily score. I never had to do it so don't take my word for it, and depending on how big of a hit it was today it might be way to many miles to try and bring up todays score rather than starting fresh with a 100 tomorrow

Edit: Per AlanSubie4Life

It’s neither of these things. It’s the measurement of each parameter over the course of the day without regard for the mileage or the % measurement of each individual trip. (FCW and AP disengage are more obviously calculated, not talking about them.). So you can have a two-hour drive with 60% unsafe following, followed by a carefully designed drive of 15 minutes with unsafe following of 0%, and end up with a daily unsafe follow of 0.5%.

It is all event duration based at the daily level.

 

[mwasylyk]

I love the idea that you would need to pay more for software that is supposed to drive your car for you at a level substantially safer than you, because when you do drive the car, you're not "safe" enough.

This tells you all you need to know about where FSD is right now. It's entertainment that can only be handled by the very safest drivers. It's not autonomy in any way.

Technically, it should be the other way around. The better a driver you are, the harder it is for FSD to be better than you, so you should have to pay more.

No one was even saying that. He wasn't even implying that, he was just using it as a comparative example to an insurance premium. What are you on about lol?

 

[dckiwi]

This tells you all you need to know about where FSD is right now.

I think most of the folks clamoring to get FSD Beta know this. If they’re like me, they’re not signing up so they son’t have to drive (that will come much later).

They’re signing up to experience the progress as it’s made, and be part of something big.

Plenty of people don’t feel that way, including many Tesla owners I know. They want it to ‘do what it says on the label’. And I totally get that! We’re the oddballs here….

 

[EVNow]

I love the idea that you would need to pay more for software that is supposed to drive your car for you at a level substantially safer than you, because when you do drive the car, you're not "safe" enough.

You totally misunderstood the comment. We were discussing insurance based on safety scores (that other insurers are using).

 

[AlanSubie4Life]

I am just not sure if the daily score is each trips score(I think you'd have to manually put in each trip into the score simulator to get the score for each trip and write them down) average, or if each trip score is also weighted by miles when calculating the daily score.

It’s neither of these things. It’s the measurement of each parameter over the course of the day without regard for the mileage or the % measurement of each individual trip. (FCW and AP disengage are more obviously calculated; not talking about them.) So you can very easily have a two-hour drive with 60% unsafe following, followed by a carefully designed drive of 15 minutes with unsafe following of 0%, and end up with a daily unsafe follow of 0.5%.

It is all event duration based at the daily level.

 

[mwasylyk]

It’s neither of these things. It’s the measurement of each parameter over the course of the day without regard for the mileage or the % measurement of each individual trip. (FCW and AP disengage are more obviously calculated, not talking about them.). So you can have a two-hour drive with 60% unsafe following, followed by a carefully designed drive of 15 minutes with unsafe following of 0%, and end up with a daily unsafe follow of 0.5%.

It is all event duration based at the daily level.

Good to know

 

[brianman]

A poster doesn’t bow down and worship Elon, or praise Tesla for demanding they to do unnecessary work to get the software they paid for after years = you’re a terrible human being. A total scum.

The cult is alive and well!

I think you misunderstand either the post I replied to, my reply, or both. Because your reply has nothing to do with what I said.

 

[ConSaab]

AFAIK the profile has nothing to do with it besides showing who drove the miles to you. Is your current daily score a 99 or 100? Because you would need a 100 for the current day you are driving in order to increase your overall score with the mile weighted day score. I know everyone uses this and it worked for me to calculate miles needed to reach 100. Safety Score Calculator

Anyone correct me if I am wrong.

Also if your score for the day is below 100 I think you could use the same formula to calculate how many miles you'd need to drive today to increase it to 100. I am just not sure if the daily score is each trips score(I think you'd have to manually put in each trip into the score simulator to get the score for each trip and write them down) average, or if each trip score is also weighted by miles when calculating the daily score. I never had to do it so don't take my word for it, and depending on how big of a hit it was today it might be way to many miles to try and bring up todays score rather than starting fresh with a 100 tomorrow

Edit: Per AlanSubie4Life

Thanks for the response. Correct, my daily scores have been 100…