Safety Score

[S4WRXTTCS]

Just got done driving around and around and around and around in the roundabout near my house at 13 mph for 6 minutes. Safety score 100 today…if I can put 527 more miles on the car before Friday night at a 100 score, I should be able to bring my average score to 100. Been sitting at 99…

Wife’s car just needs like 120 more miles at 100 to get it to 100, should be able to do that tomorrow after work.

Is there anywhere we can rent a roundabout for about 4 days?

 

[nvx1977]

Is there anywhere we can rent a roundabout for about 4 days?

big empty parking lot works too. if anyone asks, you're doing slow donuts.

 

[powertoold]

Certain driving styles are conducive to better Safety Scores, as designed. I've had no problem maintaining my 100 score, although I must resist any road rage tendencies.

 

[AlanSubie4Life]

I’m surprised I got 100 today. Per my calcs, the 0.3% braking and 0.6% following numbers should have dinged me down to 99.4232 which typically rounds down to 99, not up to 100.

@AlanSubie4Life can you please check my math?

View attachment 717972

As @novox77 said, you could have 0.26% hard braking and 0.64% unsafe following and still end up on the right side of 99.5%. That would look like 0.3% and 0.6%. It does appear you got lucky - 0.27% hard braking puts you over the edge to a 99 (99.49). So it's possible to have 0.3% hard braking end up with a score of 99, or 100, depending on the next digit of that "0.3%."

That's my guess for what is happening here, anyway. Tesla likely isn't rounding the inputs to the formula very much (would lose some precision), just the output of the formula.

You can see this in any spreadsheet by just filling in the formula and playing with the inputs (while not rounding the output to more than two decimal places).

 

[dcesq06]

Not really. It might be time to expand access. You might want to double/quadruple usage.

How many people are in it now? 1k, maybe 2k? I can completely see expanding access, but it's the going 10x the current number in the beta group in a couple of weeks that I find odd. And why ramp up by so many testers, so quickly when there is still a lot of work left?

 

[BitJam]

And why ramp up by so many testers, so quickly when there is still a lot of work left?

The work needed is training the neural nets which requires vast amounts of data gathered from a large number of beta testers. Neural nets get better and better as you feed them more data. But it takes a lot of data. Hopefully FSD will improve quickly making it safer to increase the pool of beta testers which will help FSD improve more, etc.

 

[boonedocks]

Thank you for the worksheet posted above. Don't think I'll make 99.5+ (100) by Friday. The walk up is just too slow with the number of miles I put on my car

 

[techlogik]

This whole safety system is such a sham and destined to fail too make the daily commute to/from work where I live.

Waiting for Elon to post his Safety Score.....that should be a real laugh.

I'll just wait 2 more months anyway when City Streets driving is available to everybody like my App and their website says....not point in this stupid game anymore.

With that said, if Elon/Tesla really want to make their system learn real world driving and teach it, their neural network needs better real world environments like large multi-million people areas where driving is a necessity like Tampa/ATL. We have no city mass transit, everybody has to drive, roads are so under built and planning commissions (government for you) are so far behind where they should have been literally 20yrs ago in this area that it is just a recipe for utter garbage scores.

Feel free to come down during morning/afternoon rush hour and let us know how long your 99-100 score lasts. I give it 5 minutes...rofl.

 

[ahoen117]

Y'all probably don't care, but I'm excited this morning! Finally averaged back up to 100! (technically 99.52 over a total of 538 miles)

Also, I'd like to say thank you to everyone who posted tips in this thread! Over the weekend I figured out that I was actually decelerating too slowly in a lot of cases which, as I've seen a few people say here, wasn't counting towards my braking time between .1g and .3g. Because of this I swapped from chill to sport and have found it much easier to get good braking scores. Not having a delay of regen coming on makes it much easier to slow down smoothly without going so slow you don't actually reach .1g of braking. So actually the past few days, I've driven what I'd consider pretty normally, just looking way ahead and trying to avoid "bad" situations before they happen, and it's been tons easier to get consistent 100's.

 

[n.one.one]

Thank you for the worksheet posted above. Don't think I'll make 99.5+ (100) by Friday. The walk up is just too slow with the number of miles I put on my car


View attachment 718027

Thanks @boonedocks . For those who may have looked at this last night please note that I've made some changes to the formatting, instructions, notes, etc. If you had any challenges please take another look via this link.

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

 

[boonedocks]

This whole safety system is such a sham and destined to fail too make the daily commute to/from work where I live.

Waiting for Elon to post his Safety Score.....that should be a real laugh.

I'll just wait 2 more months anyway when City Streets driving is available to everybody like my App and their website says....not point in this stupid game anymore.

With that said, if Elon/Tesla really want to make their system learn real world driving and teach it, their neural network needs better real world environments like large multi-million people areas where driving is a necessity like Tampa/ATL. We have no city mass transit, everybody has to drive, roads are so under built and planning commissions (government for you) are so far behind where they should have been literally 20yrs ago in this area that it is just a recipe for utter garbage scores.

Feel free to come down during morning/afternoon rush hour and let us know how long your 99-100 score lasts. I give it 5 minutes...rofl.

My daily scores are Atlanta I-85 from Hall county to Fulton county rush hour to and form work. Not for the faint of heart and definitely not, from what I am reading, what you are referring to my 99 total score possibly relating to. (maybe your comment "Feel free to come down during morning/afternoon rush hour and let us know how long your 99-100 score lasts. I give it 5 minutes...rofl." wasn't in reference to my score tho?

 

[PagodaY]

Finally Google has the updated app. Installed it and found my score to be 98, without any special monitoring! I'm happy with that. Will it get me into the first batch of release? I've waited this long, so no biggie if I don't get selected.

 

[Zythryn]

This whole safety system is such a sham and destined to fail too make the daily commute to/from work where I live.

I’ve been to Florida many times.
Frankly, I think 95% of Florida drivers should have their driving privileges revoked ;-)

 

[nvx1977]

Y'all probably don't care, but I'm excited this morning! Finally averaged back up to 100! (technically 99.52 over a total of 538 miles)

Also, I'd like to say thank you to everyone who posted tips in this thread! Over the weekend I figured out that I was actually decelerating too slowly in a lot of cases which, as I've seen a few people say here, wasn't counting towards my braking time between .1g and .3g. Because of this I swapped from chill to sport and have found it much easier to get good braking scores. Not having a delay of regen coming on makes it much easier to slow down smoothly without going so slow you don't actually reach .1g of braking. So actually the past few days, I've driven what I'd consider pretty normally, just looking way ahead and trying to avoid "bad" situations before they happen, and it's been tons easier to get consistent 100's.


View attachment 718031

So many people throwing their hands up and complaining about inexplicable dings, esp the AP users. Good advice is out there on these forums (albeit spread around in gigantic threads), and it's not that hard, if you have a bit of drive, to dig into the mechanics and figure out how to minimize those dings.

People's sour grapes assumption of needing to drive like a granny are exaggerated and incorrect. Actively trying to drive like that can very likely hurt your score, as you said, when good braking and cornering aren't given credit. If you learn how to avoid the dings, and stay in the "good" ranges, you can put a lot of spirit back into your daily driving. Some situations do require slow speed, and it's all about situational awareness (which while not scored, is probably a bigger predictor of safe driving). No one around me would suspect I'm taking special effort to score high for this "game." Just another tesla that can crush anyone at a light.

 

[blecchus_rex]

Fellow FSD beta candidates, you’ve been at this for several days by now. You’ve reverse engineered the algorithm, your approach to stops and turns transcends mere safety, you define driving par excellence, collectively this esteemed group has achieved Sunday driving ne plus ultra.

So it’s time to live a little. You purchased (or potentially leased) a high performance vehicle that demands to be driven… so careen into that turn, press forward into traffic and urge them forth from the near confines of their rear bumper, feel the exhilaration of emergency braking, and explore the performance envelope for FCW. Bravo, brave beta candidates - you deserve this. Today’s your day to reward great behavior w/ a little bit of naughty reckless abandon - you know you want to!!!

…

And for those cynics among you that might think this post is driven by ulterior motive… shame on you. I definitely don’t see you as vying for the same spots in the beta. And I would never encourage behavior that might elevate my score vis-à-vis.

 

[tivoboy]

Finally Google has the updated app. Installed it and found my score to be 98, without any special monitoring! I'm happy with that. Will it get me into the first batch of release? I've waited this long, so no biggie if I don't get selected.

showing up here now too, google must have finally opened the flood gates.

 

[dcesq06]

The work needed is training the neural nets which requires vast amounts of data gathered from a large number of beta testers. Neural nets get better and better as you feed them more data. But it takes a lot of data. Hopefully FSD will improve quickly making it safer to increase the pool of beta testers which will help FSD improve more, etc.

I understand the concept of feeding the NN, but why expand the beta group by leaps and bounds now? Wouldn't it make more sense to start gathering that data after the FSD stacks are merged?

Dont get me wrong, I am AMPED about Tesla expanding the beta group (and I am fighting daily to keep my safety score in check!!!!), but it just seems like an unusual time to go from 1k beta testers to 20k+ by the end of October (based on purportedly adding 1k new testers per day starting this weekend). Just seems like the more prudent approach would be to gradually expand the beta group (maybe by 1k with each new release). And, my gut feeling tells me that Tesla won't come anywhere close to expanding the beta group by 1k per day. In the next day or two, Elon will tweet some reason as to why they are going to take a slower approach.

Just my $.02. Friday/Saturday can't get here soon enough!!! As I work from home and I'm happy with where I stand with my safety score, I am conflicted as to whether to drive some every day; and, if so, how much.

 

[TofOK]

You're right, then.

In that case, as you said, you have enough total miles (and enough miles today already) that it just makes sense to drive as much as you can today, until you get to around 160 miles. It only makes sense to stop driving in very low distance (like, 1-2 miles) days (assuming there is a cap on the max value used for FCW/1000 miles). Because then that can be very low weight (1 mile with 1 FCW should result in 51 (assuming the score simulator reflects reality), rather than something lower, due to the cap) and you can easily make it better, with 50-100 miles the subsequent day. But that ship has sailed (and was never possible for you).

So you're right that given the situation, you do need to drive more miles today.

It's diminishing returns after about 160 miles though. At about 160 you'll be at 98.5 (99), and to make that to 99.5 you'd have to drive another 340 miles (500 total). So you may as well (given the way the rounding calculations work currently) just drive those miles the following day and get them to count and average to 100 (160 at 98.5, rounding to 99, plus 160 at 100, which averages to 99.5, which rounds up to 100). (That would be 320 miles total rather than the 500 on a single day, which is better.)

Anyway with all your prior 100 scores you just have to get to 160 miles today (you need to see the score go to 99) and then you should be ok.

For example, if you did just 26 total today, you'd be at 89 for the day, and to get that to average out you'd need to add 600 miles at 100 to get it to average out to 100 (99.5). That's worse (626 miles total). But if you did just 80 miles today (gives 97 score), you'd need to add 400 in the future at 100 to balance it to average out to 99.5. (480 total)

Thank you for the detailed explanation.

I did go on a long drive last night, around 90 miles altogether, which I believe was flawless, and was very excited to see how this would affect the safety score only to find out that I had another phantom FCW and unsafe following during this drive

Anyway, I decided not to go out of my way anymore to get this fixed ... just hoping Tesla vision isn't as bad driving as it is detecting safety incidents!

 

[Daniel in SD]

The work needed is training the neural nets which requires vast amounts of data gathered from a large number of beta testers. Neural nets get better and better as you feed them more data. But it takes a lot of data. Hopefully FSD will improve quickly making it safer to increase the pool of beta testers which will help FSD improve more, etc.

Tesla collects data from the entire fleet already. From what I can tell the FSD Beta testers just show them what bugs to work on and it sure looks like they have enough to keep the team busy right now.

 

[jebinc]

You have less than 3 hours to improve your daily score: 1. to fix following too close, you need to drive over 50 mph and follow a car within 3 seconds without AP (but more than 1 second, or you will get another ding). 2. To fix aggressive turn, you can drive on a roundabout with G Sensor (app on Play store) and make sure the G is bigger than 0.2 but less than 0.4. If you have enough time you can get 100 for today. I had a similar situation last Friday: drove 90 miles but was dinged when existing freeway before a red light for hard breaking (1.3%) which made my daily score to 97. I found a quite road and keep accelerating and regen braking while watching G sensor to make sure it more than 0.1 less than 0.3. I got dinged again once, but I managed to lower the hard breaking score from 1.3% to 0.4% by midnight and daily score from 97 to 99. If I have time I can bring it down to 0.2 and fix my daily score to 100. Then I made another 90 mile drive next day with 100 to bring my overall score back to 100 (from 99.41 to 99.51). Also if you know you are dinged, you can reset to discard that trip (I didn't know I was dinged)

They have medications for compulsions like this...