Safety Score

[22522]

Thank you all for the tips. I was able to pull my 2 day score up to 95 by making todays score a 99.

I had a following distance issue to fix, so I drove to Greenville and back with 3 to 4 seconds following distance the whole way with no Autopilot. Score now looks like this. Thanks again!

 

[fiveten]

I drove too many miles the first day…if I want to bring my model Y score up to a 99…I need to drive about 200 miles the next 5 days getting 100 each day…challenge accepted!!

 

[PagodaY]

Boy, I'm jealous of all you IOS people. Can't wait for the Android app to update to 4.1

 

[Driver08]

The street up to my house is kind of narrow and it curves, with cars parked on both sides. I would always get false positive FCW with the medium setting so I changed it to late. Now I’m getting penalized on my safety score for these false positives. I’m not sure what to do other than to try driving extra slowly in the middle of the road. Any suggestions?

https://twitter.com/x/status/1442197732882632707

that was hard braking with AP, but perhaps the same applies to false FCWs

 

[nvx1977]

Boy, I'm jealous of all you IOS people. Can't wait for the Android app to update to 4.1

You can sideload 4.1.0.

Tesla 4.1.0-663 (160-640dpi) (Android 6.0+) APK Download by Tesla, Inc. - APKMirror

Tesla 4.1.0-663 (160-640dpi) (Android 6.0+) APK Download by Tesla, Inc. - APKMirror Free and safe Android APK downloads

www.apkmirror.com

after downloading the apk bundle, use this to install it:

APKMirror Installer (Official) - Apps on Google Play

APKMirror Installer helps you install APKs and apkm, xapk, and apks app bundles

play.google.com

 

[Lexdysic]

Driving on a side street near my house, I was forced to brake hard because a child started to run into the street from between 2 parked cars. If I had chosen to hit the child and continue, my score would've remained 100. Instead I'm at 98.

This is insanity.

 

[johnmodely]

Yea these metrics are so stupid.

should be driver attentiveness via camera, running red lights, and crossing yellow divider as better three metrics than these dumb hard braking and aggressive turning.

you can do those things without being a bad driver, but crossing median or a red light I would think is a way better predictor.

guess they figured everyone would pass if that’s what they tracked so using silly metrics to create a smaller group.

 

[MP3Mike]

Driving on a side street near my house, I was forced to brake hard because a child started to run into the street from between 2 parked cars. If I had chosen to hit the child and continue, my score would've remained 100. Instead I'm at 98.

I think you are overly obsessing about the score. We don't know the criteria that Tesla will use to select testers, but my guess is that it will be something like:

• Everyone with a score between 80 and 92 will be put in the list as a potential tester.
• That list will then be sorted by a combination of:
  • Average miles driven daily
  • % miles on AP
  • % highway miles
  • % city miles
  • % miles driven at night
  • Geographic location
  • etc.

Then they will pick off the top 100 to get access to the beta and monitor them to see how it goes. If all is well after a couple weeks they will re-do and add a couple hundred more, lather, rinse, repeat.

Note: IMHO: I think they will exclude people with a score above 92 because they know in normal driving you are going to have "events" and it shows they are trying to game the system, and you don't want those kind of people testing your beta software. (Maybe the cut-off will be 95, or 97, but I really think there will be a cut-off.)

 

[thewishmaster]

After the “unsafe“ trip I took a scenic non-hilly route home and scored perfectly on every criteria but my average score went down

 

[Dan D.]

The street up to my house is kind of narrow and it curves, with cars parked on both sides. I would always get false positive FCW with the medium setting so I changed it to late. Now I’m getting penalized on my safety score for these false positives. I’m not sure what to do other than to try driving extra slowly in the middle of the road. Any suggestions?

Reverse up the street to your house?

 

[AlanSubie4Life]

IMHO: I think they will exclude people with a score above 92

I think that is really unlikely. Good drivers will want to optimize their score and they have the car control to do it (it’s not all that difficult, does not require an expert driver, just a careful one), even if what is being measured is somewhat loosely correlated with actual safe behavior.

The rest of the criteria and general strategy, you’re probably right. I think they’ll also aim for geographic diversity I would think.

No idea what the qualifying level will be though.

I agree that a score of 90 or even lower is probably more typical if you are not gaming it. But I assume Tesla wants people who can get a high score. Otherwise why would they have this system?

 

[azred]

Game the system? I am simply honoring my deceased grandmothers with my driving the past two days.

 

[EVNow]

Game the system? I am simply honoring my deceased grandmothers with my driving the past two days.

Or go back to Leaf like hypermiling.

 

[MP3Mike]

I think that is really unlikely. Good drivers will want to optimize their score and they have the car control to do it (it’s not all that difficult, does not require an expert driver, just a careful one), even if what is being measured is somewhat loosely correlated with actual safe behavior.

A score of 100 may indicate that you are driving in an area/times where nothing interesting is happening. (No traffic, unexpected drivers getting in your way, pedestrians running in to the street, twisties, etc.) Tesla may not be interested in getting more uninteresting footage/testing. (It already does well in uninteresting/no traffic situations.)

 

[AlanSubie4Life]

After the “unsafe“ trip I took a scenic non-hilly route home and scored perfectly on every criteria but my average score went down

Yep. Such is the danger of increasing mileage…

 

[EVNow]

Boy, I'm jealous of all you IOS people. Can't wait for the Android app to update to 4.1

Side load.

 

[Daniel in SD]

Do we know that they have collected the data that they are using for the score all along without telling us? I would consider that a massive breach of privacy. If they haven't, they probably don't have a whole lot of data yet. Can Tesla Insurance customers outside of California opt in to be scored?

No I don't know, I just assumed I clicked on a EULA at some point that allowed it. haha.
They say it's 6 billion miles of data which seems like enough and more than would get from Tesla Insurance?
It would be interesting to know how well their formula works.

The current formula was derived based on statistical modeling using 6 billion miles of fleet data.

 

[AlanSubie4Life]

A score of 100 may indicate that you are driving in an area/times where nothing interesting is happening

Definitely. But I assume they are also recording location and times of drives (though it is not clear how comprehensive the data collection from The Button is - the verbiage associated is extremely broad though).

You’d think Tesla would look for good scores in somewhat difficult circumstances. They’ll probably look for a variety of factors and pick the best scores for that given set of factors.

I just don’t see what benefit there would be in picking lower scores, given the other information they presumably have.

 

[Rsandy]

My wife drives an S and I have a model 3. Both are in the queue to get the FSD beta. I didn’t tell my wife that her driving is being monitored. She is blithely unaware of the entire FSD beta button. Her score for two days is 100%. Obviously she isn’t gaming the system. My score is 96%. I was trying to be careful. I found that her results were better amusing.

I think Tesla wants beta testers who will use FSD, plus they claim that using it way raises safety. That is why they mask miles on autopilot but penalize getting kicked out for not holding the steering wheel. As for their other metrics, the warnings for imminent collisions and high g’s on turns and braking make sense to me as predictors of accidents. The angst over these metrics is misplaced. Tesla wants to accumulate more FSD beta testing miles and still have zero accidents in the testing program. They are not trying to provide FSD beta to everyone who paid for it and signs up.

Lastly, the responses of people who would like to gain access is a great example of how people respond to incentives. Tesla has kicked testers off of the program for being inattentive. If you want to stay on testing program you will have to stay on your good behavior.

 

[Daniel in SD]

Everyone with a score between 80 and 92 will be put in the list as a potential tester.

Nice try trying to lower people's scores. 92 is the median driver! It's funny how they scaled it so half the people get A's.
Maybe they want people who can learn the game and have the driving skills to get a perfect score? I question whether Tesla wants people using FSD Beta in difficult situations right now anyway. It doesn't look like they're having any problem finding bugs these days. Maybe they just want to roll it out very slowly and maintain a good safety record.