Safety Score

[agordon117]

You know, I have to wonder if this was done in this specific way on purpose. I have a score of 100, so I'm not complaining (just sweating). I wonder if they figured that people who got the highest scores would be most likely to not live/work in highly crowded cities, which they already have plenty of data for. So they wanted to pick people who were capable of driving the smoothest in more rural areas, to bolster the data that they have outside of cities before the wider rollout.

I don't agree with the settings for the braking and turning and how that correlates to safety instead of just smooth driving, this was just my thought this morning when elon tweeted about delaying further rollout past the perfect scores.

 

[Ramphex]

You know, I have to wonder if this was done in this specific way on purpose. I have a score of 100, so I'm not complaining (just sweating). I wonder if they figured that people who got the highest scores would be most likely to not live/work in highly crowded cities, which they already have plenty of data for. So they wanted to pick people who were capable of driving the smoothest in more rural areas, to bolster the data that they have outside of cities before the wider rollout.

I don't agree with the settings for the braking and turning and how that correlates to safety instead of just smooth driving, this was just my thought this morning when elon tweeted about delaying further rollout past the perfect scores.

I drive a mixture and busier places it’s literally impossible to get 100 because of people always trying to get into the gap in between you and the vehicle ahead of you (trying to keep distance to avoid unsafe following). The threshold is so weird for that, I feel that as unless you’re drafting, in traffic commute you’re not “unsafe” following.

 

[gearchruncher]

I wonder if they figured that people who got the highest scores would be most likely to not live/work in highly crowded cities, which they already have plenty of data for. So they wanted to pick people who were capable of driving the smoothest in more rural areas, to bolster the data that they have outside of cities before the wider rollout.

Well @powertoold says it's obvious that the safety score measures both how you drive and where you drive, so yeah, they are only rolling out to people in very rural areas.

Because of course, where else is the best place to roll out CITY STREETS AUTOSTEER (coming this year since 2019!) except to people that never drive on city streets!?

Oh right, let's make the excuse that they have "plenty of data in cities" but somehow need more data in the easy places to drive.

It makes you wonder just how bad CSA is if they are so afraid of giving it to anyone that lives anywhere near a city.

For those of you that don't have percent 100 scores, it's time to just resign yourself and understand that you are unsafe drivers. The algorithm has spoken, and it cannot be wrong with 6B miles of training data. You should seriously consider if you should keep driving, given how you are endangering your families and society.

 

[powertoold]

Well @powertoold says it's obvious that the safety score measures both how you drive and where you drive, so yeah, they are only rolling out to people in very rural areas.

Not true, you could get 100 in urban areas with the right driving style.

Anyway, doesn't seem like you pressed the button, so it's difficult to gauge your experience with driving it.

 

[nvx1977]

Well @powertoold says it's obvious that the safety score measures both how you drive and where you drive, so yeah, they are only rolling out to people in very rural areas.

Because of course, where else is the best place to roll out CITY STREETS AUTOSTEER (coming this year since 2019!) except to people that never drive on city streets!?

Oh right, let's make the excuse that they have "plenty of data in cities" but somehow need more data in the easy places to drive.

It makes you wonder just how bad CSA is if they are so afraid of giving it to anyone that lives anywhere near a city.

For those of you that don't have percent 100 scores, it's time to just resign yourself and understand that you are unsafe drivers. The algorithm has spoken, and it cannot be wrong with 6B miles of training data. You should seriously consider if you should keep driving, given how you are endangering your families and society.

seems logical to me to start in areas with least chance of collision. least amount of PR sh*tshow. If it turns out that "rural" FSD testing is not making good progress, then it seem logical to unleash the next batch of lower scores. Rural > exurbs > suburbs > cities > downtown.

 

[gearchruncher]

Speeding isn’t that big a deal, IMO. Far more important is speed differential in relation to other traffic on the road. If you’re going with the flow of traffic, even if everyone is going 10-20 mph over the speed limit, you’re a far safer driver than one who is his own island chugging along at a fairly arbitrary speed limit forcing everyone else on the road to constantly change lanes to pass you, or if you’re a speedster, constantly weaving between lanes trying to go significantly faster than the flow of traffic.

I love that @FalconFour downvoted you for this, even though he said exactly the same thing yesterday. He literally argued that going under the speed limit was unsafe for the same reason, and that you had to go above the speed limit,:

There is one speed on the road: the speed limit. You drive THAT (plus a few because California literally won't allow people to just drive that). If you can't drive that speed, you put the hazards on and either stay in the right lane if it's literally the only way to get to your destination, or you get off the road and go to a smaller road appropriate to the vehicle or situation. You don't ride a bike on the freeway for this reason. This simple logic eliminates Carlin's quote from being a concern and transforms it literally into "anyone driving slower than you is unsafe, anyone driving faster than you is unsafe". The rule-maker is the speed limit (and all CA freeways are 65 at most - so it's really easy to know!). This is the foundation of the logic that AP ends up creating in operation - there's one speed that traffic ought to be going, and you can pass on the left but get out of the passing lane when it's clear and you're done.

 

[n.one.one]

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?

 

[gearchruncher]

Not true, you could get 100 in urban areas with the right driving style.

Pretty obvious to me, but it's not stated explicitly. Wouldn't that cause more drama if they mentioned that?

Insurance rates are based on driving locale, based on collision risk. Same thing with safety score.

OK, so a 100% score is "possible" in a urban area, just much, much harder. So they're giving it to amazing drivers (per their made up score) in urban areas, but just any random old person in a rural area? Sounds like they are not very confident that CITY STREETS autosteer is mind blowing.

Anyway, doesn't seem like you pressed the button, so it's difficult to gauge your experience with driving it.

Tesla's "safety score" is only available to those people that have paid for a FSD lottery ticket. I don't play lotteries, I make purchases. I'll buy FSD when it does something useful that has any relation to the cost.

You'd think if it really was a "safety score" and how at Tesla "safety is always the top priority" they would make this score available to everyone. But weirdly they only allow people that are entering the lottery to see it.

 

[Daniel in SD]

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 718782

Sounds logical. However I would guess that people submitting their scores to TeslaFi have a disproportionate number of 100's because they're more likely to be studying threads on TMC about how to optimize their scores. That would make the number of vehicles participating lower.

 

[nvx1977]

You'd think if it really was a "safety score" and how at Tesla "safety is always the top priority" they would make this score available to everyone. But weirdly they only allow people that are entering the lottery to see it.

I'm hoping they release an efficiency score for all some day (beyond just the Wh/mi readout). This score breaks things down so you see what driving characteristics make you more efficient. I loved the analog gauge in my old BMW M3 that showed me my instantaneous MPG. That made me watch my accel when I wanted to just go easy.

Feels like less pressure to be efficient than to be safe, especially since there's no potentially monetary penalties on the insurance safety side. Or they offer both, and you can toggle them off/on or opt out/in.

 

[gearchruncher]

If it turns out that "rural" FSD testing is not making good progress, then it seem logical to unleash the next batch of lower scores. Rural > exurbs > suburbs > cities > downtown.

Of course. You can see this is exactly how they did it when they released to Youtubers too. That's why we have no breathless videos of lombard street or downtown Seattle.

Oh wait. Are you telling me Tesla marketed CITY STREETS autosteer with a bunch of people driving it in cities, yet the actual product is really only ready for rural roads? The same company that calls an L2 system FULL SELF DRIVING? Well, I never....

 

[gearchruncher]

I'm hoping they release an efficiency score for all some day.

My screen already shows wh/mi. You want them to gameify a simple, quantitative measurement even more?

 

[tivoboy]

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

Welp guess I’m out for a while with my 99 score

Yes, sadly as I had indicated last week if one isn’t probably above 96+, its most likely going to be months before getting the FSD beta. Certainly we still have no idea how exactly they will calc this or roll it out, but its almost certainly NOT going to be 1000/day for weeks or months. time will tell, I’m holding at 96 and it ain’t going to get better and I’m not holding my breath.

 

[nvx1977]

My screen already shows wh/mi. You want them to gameify a simple, quantitative measurement even more?

yep, you beat me to my edit. The overall Wh/mi is great, but give you no feedback on how you earned that score.

 

[agordon117]

Well @powertoold says it's obvious that the safety score measures both how you drive and where you drive, so yeah, they are only rolling out to people in very rural areas.

Because of course, where else is the best place to roll out CITY STREETS AUTOSTEER (coming this year since 2019!) except to people that never drive on city streets!?

Oh right, let's make the excuse that they have "plenty of data in cities" but somehow need more data in the easy places to drive.

It makes you wonder just how bad CSA is if they are so afraid of giving it to anyone that lives anywhere near a city.

For those of you that don't have percent 100 scores, it's time to just resign yourself and understand that you are unsafe drivers. The algorithm has spoken, and it cannot be wrong with 6B miles of training data. You should seriously consider if you should keep driving, given how you are endangering your families and society.

I'm not making excuses for them about the data in cities, I'm just looking at it from one hypothetical perspective. With my assumption, CSA isn't necessarily that bad, but they don't want to start rolling this out more widely until they have more data in rural areas to better cover their bases.

 

[mongo]

The accelerometer only registers the -0.1g of the downslope when you are stopped because the brakes are supplying the required force. If you don't apply brakes the accelerometer sees 0 gs. Without brakes the only restraining force is regen which is providing a constant -0.2 gs in the direction of travel. Gravity is accelerating the car 0.1 g down the hill for a net of -0.1 gs. The car is decelerating at 0.1 g. The accelerometer still reads -0.2 gs though, because it zero for freefall minus the constant 0.2 from regen.

Regen isn't a constant kw from the motor. It depends on the speed. Power = f x v = m x a x v, so a = Power / (m x v). Therefore for constant deceleration, the power must decrease as the speed decreases. You can see this happen on the car's power meter as you slow down.

Yah, I was simplifying to a specific speed/force point (and region above max regen).
You bring up a great point that regen can't produce more force (acceleration) than it can produce. A car stopped on a hill has all its braking force available, whereas once in gear it is limited to the motor ouput. (Automatic pointed uphill for comparison, take your foot off the brake and it rolls backwards, but on a flat surface it would creep forwards). So you can have slope that regen can't hold the car in place on.

If regen is open loop controlled by a speed to kW lookup table to provide an approximately constant deceleration force, then regen can only trip the braking limit if it can achieve 0.3G on its own, which it would then also do on flat ground (unless the open loop includes vehicle angle).

If it is close loop controlled to the vehicle acceleration (up to the power limit), then 0.2G of net deceleration requires 0.3G of regen induced force if on a 0.1G slope and it could trip the braking limit. Due to the power limit, a quickly moving car could have less regen than needed to decelerate on the slope, but would slow at a lower speed.

Alternatively, regen could have both a power and speed upper bound (don't slow too quickly). That would also allow more regen force on a downhill than on a flat (clipped due to speed change)

 

[EVNow]

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

So, everyone with 100/100 score gets it on Friday/Saturday. I hope thats the rounded 100 and not actually "perfect" 100.

 

[EVNow]

they are only rolling out to people in very rural areas.

I expect most of the 100 to be in the suburbs. Afterall most Tesla buyers are in suburbs.

 

[gearchruncher]

The overall Wh/mi is great, but give you no feedback on how you earned that score.

And just like the "safety score" your wh/mi is more based on where you drive than how you drive.
But this is the company that refuses to put speed or acceleration in their safety score. What will the efficiency score look like without those?

 

[EricUSC]

I have no planned drives today (Thursday 10/7) but have enough planned drive for tomorrow (Friday 10/8) to bring my score to a 99.5 (rounded to 100 in app).

Which is Elon's definition of Friday midnight: Friday 10/7 at 12:00 AM or Saturday 10/8 at 12:00 AM? Also, wondering if I need to make the cutoff for this round or if I'll get FSD Beta once my score hits 100?

Trying to decide if I need a freeway cruise late tonight.