Safety Score

[EVNow]

And what is up with these midnight releases, meaning the first people to ever go and use FSD will have to do in the dark? Is that really actuarially sound?

My guess is the bandwidth costs are low around that time.

Also, from a project management perspective, you tell the team they have to release by midnight. So, guess what they do.

ps : Also with Tesla, the car is likely to be in the garage.

 

[ACAG]

When the car is in valet mode, is the safety score still being calculated?

 

[Dan D.]

There were crashes in the 6 billion miles of data, that’s how they came up with the safety score. People who have a lower safety score are more likely to crash.

The obvious refinement to the safety score will be to figure out what metrics correlate to FSD Beta crashes which is different than driving manually. Also they may want to somehow evaluate how well people using FSD Beta monitor the system to see who to kick out of the beta.

Agree to a point.

They pared down the parameters for this Safety Score (according to some who said acceleration at least used to be in the insurance score). From this point on there have to be further crashes to make refinements. But as you imply (IMHO), there will be crashes, so sure, they can do what they want with their predictor algorithm.

As far as on-FSD Beta safety score, we'll see if they create one. Let's hope there are no crashes on FSD, or at least so long as crashes include curb-strike and things that don't rise to the level of "accident" as Tesla describes an airbag/restraint deployment.

 

[MP3Mike]

When the car is in valet mode, is the safety score still being calculated?

Currently, yes.

 

[Zextraterrestrial]

Donate $100 to me (I have EAP already) and I'll sign up for the FSD sub and push the button and run any tests you want. I'm going racing this weekend, my score would be trash anyway. I can even do tests on private property.

The fact you can't see your "safety score" without having paid for FSD...

As of now TeslaFi reports that 9% of their 1,569 full self driving fleet have score of 100 (which is 141 vehicles). If we assume (per Elon's tweet) that there will be 1,200 with a score of 100 as of the 10.2 release then 1,200 divided by .09 would indicate that 13,333 vehicles are participating in the test.

This is for discussion...I'm not certain on this...any thoughts...is it logical?

View attachment 718785

I'm racing auto X this weekend too...currently at 87 because of 3 days with one collision warning each & total mileage somewhere ~150
I've tried hard to drive slow & careful. I'm near the bottom I'm tired of it. I'm about ready to drive like I normally do. I like to drive aggressively defensive. I stay away from other cars, avoid blind spots, pay attention to other drivers/cars intentions, watch my mirrors and use all of my cars capabilities where appropriate. When a car is not still in front of me I shouldn't be worried I'm going to hit it. I'm more worried I'll be rear ended avoiding the stupid false collision warnings.
Crappy to change your driving style and feel completely unsafe doing so & suck at it so bad

 

[Dan D.]

Why do they have to do it in the dark???? All the midnight (historically that means 0300 PDT ) release means is that the first opportunity to use Beta will be in the dark, at least in the west half of the country. No one will "have to" rush out and drive in the dark. I'm guessing that most will sleep until their normal wake up time. And even if some of us are excited and rush out and drive in the dark, so what? Certainly the early morning hours in the dark are the safest hours to drive. The biggest threat to a good driver is other drivers, and that threat is minimized at 0400. Even wildlife encounters are down in the early AM, the highest level of wildlife activity occurs in the evening twilight hours and slowly drops in the first few hours of dark.

Yeah, I'm not sure what the best time to release this would be. Certainly not during rush-hour. Probably at the lowest risk point, where few people are driving, so any problems caused affect the fewest people and can be identified and fixed.

Is Friday midnight (Saturday 12:01am) that time? Less people - check, less work traffic - check, Saturday morning isn't a typical workday - check, but at midnight Friday there could be a few more weekend-crazies on the road. Also because Tesla fans are so excited you get people who've stayed up late until it's installed and go for the 12:30 PDT to 3:30 EDT drive, probably a little sleepy/wired.

FSD seems to do quite well at night, and certainly does better with less pedestrians and other drivers.

Meh, midnight Friday PDT is probably a good time, all things considered. Perhaps Friday midnight EDT might be a bit nicer though.

 

[Nakk]

So you can't get the accelerometer to indicate -0.3 gs with regen alone.

Yes you can. I've done it several times. Like my previous post pointed out, you're mistaken in assuming the car will be in freefall. Once it reaches terminal velocity it is no longer in freefall and the force due to gravity is added to regen force and easily exceeds .3 Gs even on reasonably gentle slopes. The braking force is the sum of regen, rolling resistance and aerodynamic drag. A slope that has a terminal velocity of around 25 mph is sufficient to exceed .3 Gs with normal regen as shown by my testing.

 

[FalconFour]

"Everyone with a perfect score will get it". "Rollouts will hold for several days to see how that goes. If that looks good, beta will gradually begin rolling out to 99 scores and below".

Elon, you absolute ... trollmaster.

https://twitter.com/x/status/1446107701839343620

I can't believe this madness. I mean, I'm (*checks app*)... whaaa? 99 total and 97 today with 1.3% hard braking as the ONLY factor? okay i'm gonna fix this during my lunch break today THEN I'LL BE 100.

This game is ... I dunno what to say about it. I wouldn't expect Elon to double-down on the ill-advised gamification angle, and especially wouldn't expect a WEEK gap between 100 and 99. That's frustrating.

I hope I can get to 100 in time, but having spent the entire past 2 weeks in 99, I can relate hard to the people this is hurting.

 

[Ramphex]

I stand corrected, I really thought there’d be more people with a score of 100

Maybe I will do a little more driving tonight to get to 97, but that's as high as I can go by the end of Friday need close to 400 more miles just to get to 99

I feel like the more I drive the worse it is because it opens up more opportunities for “hard braking” or “following too closely” those are the only sections I get strikes in no matter what I do driving manually, pretty much have to drive on AP all the time.

If AP got ranked for the type of braking it does, the AP would never qualify for FSD beta

 

[TIppy]

They aren't. The car is not brought to a stop in a fixed distance from a given speed. Not even a little bit close. Regen applies a braking force that decreases with lowered speed because the kinetic energy available to convert is decreasing. I drive up and down hills constantly where I live. This becomes obvious instantly. As I said before, regen won't even slow you down on many hills as the braking force available to regen is less than the forward force gravity is applying to the car.



I think I see where you're mistaken now. The accelerometer sees .1 G on a certain slope whether or not you brake, once you reach terminal velocity. Braking adds to that .1 G What you're assuming is that the car will be accelerating unimpeded if the brakes aren't applied, it won't be. There's plenty of drag available to keep that from happening. At some speed, the rolling resistance combined with aerodynamic drag will equal the accelerating force from gravity. Depending on slope this happens anywhere from a very slow speed all the way up to 40 or 50 mph for a medium gentle slope. (Medium gentle is relative to where you live of course.) Steeper slopes will result in a higher terminal velocity of course. It's been my experience that a slope that results in a terminal velocity as low as 20 mph combined with normal regen from above that speed will exceed .3 G and cause a hard braking penalty.

Back to EVNow's example. He was traveling at 35 mph. At that speed a model 3 has 3855 watts of drag. At that power and speed the deceleration is 0.014 gs.

Regen doesn't decrease with lowered speed. It's less at higher speeds because there isn't enough regen power to cause 0.2 gs of deceleration. Anything above 47 mph for a model 3 would be less than 0.2g at full regen power. Below that speed they begin to lower the power to keep regen at 0.2gs.

 

[THXULTRA]

has the avg temps in your area dropped these last few weeks? more rain? definitely did here. that could affect Wh/mi readings.

Ya but temps have dropped to where the car should be more efficient. Gone from the 80s to the upper 60's for highs. Always get the best range out of my old model S this time of the year. Going to be going back to my normal driving now since there is no way I'm getting the beta anytime soon with a 98 score. Not worth playing Elon's games for weeks before he trickles it down to people with my score.

 

[_jal_]

I wonder if I can put my car up on blocks, put a cinderblock in the seat, set the cruise to 65 and let it go. I wonder if it would run! Ferris Bueller style. A little over 50 hours I'd be at enough miles to counteract The Bad Day. I wonder what my Wh/mi would be

 

[johnmodely]

"Everyone with a perfect score will get it". "Rollouts will hold for several days to see how that goes. If that looks good, beta will gradually begin rolling out to 99 scores and below".

Elon, you absolute ... trollmaster.

https://twitter.com/x/status/1446107701839343620

I can't believe this madness. I mean, I'm (*checks app*)... whaaa? 99 total and 97 today with 1.3% hard braking as the ONLY factor? okay i'm gonna fix this during my lunch break today THEN I'LL BE 100.

This game is ... I dunno what to say about it. I wouldn't expect Elon to double-down on the ill-advised gamification angle, and especially wouldn't expect a WEEK gap between 100 and 99. That's frustrating.

I hope I can get to 100 in time, but having spent the entire past 2 weeks in 99, I can relate hard to the people this is hurting.

You won't get it even if you get to 100. When he says perfect score he means 0 dings across all the metrics. No way would the group be that small if it included people with 99.8% scores with one or two dings here and there.

Just hope people with almost no miles are not included in this group...would not be fair if all these people with perfect scores have 10 miles since scoring came out.

 

[FalconFour]

You won't get it even if you get to 100. When he says perfect score he means 0 dings across all the metrics. No way would the group be that small if it included people with 99.8% scores with one or two dings here and there.

Just hope people with almost no miles are not included in this group...would not be fair if all these people with perfect scores have 10 miles since scoring came out.

not so fast there.

https://twitter.com/x/status/1446104772092772352

specifically says "100/100", which leads to mean perfect.

It seems very reasonable to me that there are ~1000 people with perfect scores, given what we know from TeslaFi and Teslascope metrics.

 

[gearchruncher]

I can't believe this madness. I mean, I'm (*checks app*)... whaaa? 99 total and 97 today with 1.3% hard braking as the ONLY factor? okay i'm gonna fix this during my lunch break today THEN I'LL BE 100.

This game is ... I dunno what to say about it. I wouldn't expect Elon to double-down on the ill-advised gamification angle, and especially wouldn't expect a WEEK gap between 100 and 99. That's frustrating.

I hope I can get to 100 in time, but having spent the entire past 2 weeks in 99, I can relate hard to the people this is hurting.

This coming from the person that is super pissed that CA can't use scores like this to set insurance rates and is sure others would be the ones charged more?

You see why it's a good thing CA doesn't allow arbitrary, unreviewed algorithms to set insurance rates?

I'm calling it, it will be way more than a week before someone with a 99 gets it. It will be a week in Elon time.

 

[FalconFour]

oh but I 110% freaking completely wholeheartedly agree that the people that gamed the system by driving 1 mile for a perfect 100, parked the car, then drove their gasser for a couple weeks... that they can straight burn in hell. Can't believe Elon would reward them by promising everyone with "100" would get it, regardless of mileage.

 

[gearchruncher]

Simple stuff - like take into account road gradient when looking for hard braking. Don't count FCW in the garage.

We're saying the same thing.

If you can detect road gradients and garages FROM THE DATA YOU HAVE, then you can use it. That's "algorithm improvement."

If you use data not in the data set to identify road gradient or being in the garage, and remove those events from the data set, then you are using external influence, and now any algorithm run on that data isn't representative.

This seems super obvious to me: If you know you have false FCW's in garages, so you go remove all garages from the data, train on that data, then that algorithm is only a valid estimate if fed data with garages removed. If you deploy it against the original data collection method, it's going to call people with garages much more dangerous than they really are, because it has no idea the majority of FCW's are false positives because they weren't in the training set.

Given Tesla is trying to run on the data set in real time, and claim this represents your actual safety, the algorithm needs to be trained with the false FCW's included, because they will be there in the actual used data. If there are so many false FCW's that it means it's no longer statistically correlated, well, then, yep, you can't use that. You can't say "this works great if I just remove these false positives by hand..."

 

[gearchruncher]

My guess is the bandwidth costs are low around that time.

ps : Also with Tesla, the car is likely to be in the garage.

Tesla is worried about BW costs for 1000 cars, yet when they release normal software to 10's of thousands of cars they just roll out in the middle of the day?
If this was the reason, then all updates would roll out at night. Yet normal updates are just all over the place and go to way more people.

 

[gearchruncher]

Can't believe Elon would reward them by promising everyone with "100" would get it, regardless of mileage.

...But the algorithm is an accurate measure of your safety as a driver. Why would they not use the algorithm score? It's a super effective way to hold people accountable for their driving from what I hear.

I mean, the one thing CA does allow is mileage, and it is super clear that the less you drive the lower risk you are.

 

[momo3605]

You won't get it even if you get to 100. When he says perfect score he means 0 dings across all the metrics. No way would the group be that small if it included people with 99.8% scores with one or two dings here and there.

Just hope people with almost no miles are not included in this group...would not be fair if all these people with perfect scores have 10 miles since scoring came out.

He said the number of 100s will increase by Friday. That sounds like people can improve their scores which would then round to 100. How else can the number of 100s increase otherwise