FSD beta lowers safety score

[uscbucsfan]

I mean I'm not confused at all, and I have a 99 SS score to back that up... What I know for certain is that "days" don't matter at all. All that matters is total miles driven in the last 30 days.

All days matter...that's how they calculate the 30 day score.

We combine your daily Safety Scores (up to 30 days) to calculate the aggregated Safety Score

I had 100 on 3 different cars, but it's written pretty clearly by Tesla. Believe what you want.

 

[cbdream99]

I mean I'm not confused at all, and I have a 99 SS score to back that up...

What version of SS you are using? In 1.2 and up, FSD miles are not used to calculate the 30 days weight average not it will affect the score, no matter how close it follows and how hard it brakes. You cannot use AP miles to help to bring bad days back.

• Updated Safety Score to use the miles-weighted average of the last 30 day’s Safety Scores, ignoring any miles driven on Autopilot.

 

[clydeiii]

The OG Google Docs spreadsheet to game out how various things affects your score. Not how everything is weighted by miles: Tesla Safety Score - Tracking & Planning v10

 

[clydeiii]

What version of SS you are using? In 1.2 and up, FSD miles are not used to calculate the 30 days weight average not it will affect the score, no matter how close it follows and how hard it brakes. You cannot use AP miles to help to bring bad days back.

• Updated Safety Score to use the miles-weighted average of the last 30 day’s Safety Scores, ignoring any miles driven on Autopilot.

"Ignoring any miles driven on Autopilot" doesn't mean what you think it means. It absolutely counts factors in miles driven on AP/FSD.

 

[clydeiii]

The verbiage is confusing and should be revised, but "ignoring any miles driven on Autopilot" simply means that if bad things happen while on AP, it won't count against you.

 

[uscbucsfan]

"Ignoring any miles driven on Autopilot" doesn't mean what you think it means. It absolutely counts factors in miles driven on AP/FSD.

It means what I'm saying it means. That's how people are getting horrible daily SS when they drive a few miles without AP and the rest with AP. It's not a bug, it's how it's designed.

Again, you seem like you will believe what you want regardless of what Tesla and others say.

 

[clydeiii]

I do agree with this guy saying that Hard Braking issues can't be fixed by AP/FSD miles though: 95% of my drives, Safety Score reports "Hard Braking".

That's true.

 

[clydeiii]

Thread discussing how to game the system by driving long stretches on AP: Disappointed with "safety score"

 

[uscbucsfan]

Thread discussing how to game the system by driving long stretches on AP: Disappointed with "safety score"

It seems you are still missing the point...I fear there's no hope.

Again, during the SS for FSD Beta you could drive a long drive with just FSD and your daily average would be 100...that daily average is calculated to give you your 30 day score, BUT...FSD/AP miles do not add to average out within your daily score as miles that you drive in a day aren't factored in. It's % of non-AP drives that the incident occurred.. So if you had an incident, like hard braking or close following, in a single day on a short distance...the AP miles won't erase that. That's why people got 80% close following or 60% hard braking when they used AP for hundreds of miles, but had an incident in the 1 or 2 miles prior or after engagement.

This is what Tesla is saying with..."Your daily Safety Score is not impacted by the number of miles you drive. We combine your daily Safety Scores (up to 30 days) into a mileage-weighted average to calculate the aggregated Safety Score, which is displayed on the main ‘Safety Score’ screen of the Tesla app."

 

[clydeiii]

It seems you are still missing the point...I fear there's no hope.

Again, during the SS for FSD Beta you could drive a long drive with just FSD and your daily average would be 100...that daily average is calculated to give you your 30 day score, BUT...FSD/AP miles do not add to average out within your daily score as miles that you drive in a day aren't factored in. It's % of non-AP drives that the incident occurred.. So if you had an incident, like hard braking or close following, in a single day on a short distance...the AP miles won't erase that. That's why people got 80% close following or 60% hard braking when they used AP for hundreds of miles, but had an incident in the 1 or 2 miles prior or after engagement.

This is what Tesla is saying with..."Your daily Safety Score is not impacted by the number of miles you drive. We combine your daily Safety Scores (up to 30 days) into a mileage-weighted average to calculate the aggregated Safety Score, which is displayed on the main ‘Safety Score’ screen of the Tesla app."

Mileage-weighted average tells you all you need to know.

 

[uscbucsfan]

Mileage-weighted average tells you all you need to know.

You obviously didn't read what I posted. I'm not going back and forth on this anymore. You don't seem to grasp what Tesla, myself, or others have said.

We will agree to disagree.

 

[clydeiii]

You obviously didn't read what I posted. I'm not going back and forth on this anymore. You don't seem to grasp what Tesla, myself, or others have said.

We will agree to disagree.

I read every single word. The more miles driven, the more that day’s score weighs over the last 30.

 

[cbdream99]

"Ignoring any miles driven on Autopilot" doesn't mean what you think it means. It absolutely counts factors in miles driven on AP/FSD.

If you look read the FAQ in each version of SS, and read the answers for FCW, then you will know.

 

[clydeiii]

If you look read the FAQ in each version of SS, and read the answers for FCW, then you will know.

"It is important to note that miles driven while Autopilot is engaged are not used to determine the Forward Collision Warnings per 1,000 Miles. Additionally, driving on Autopilot (including 5 seconds after Autopilot is disengaged) will not be factored into the Safety Score formula. Still, the miles driven while on Autopilot are included in the total."

(cite)

All I've ever been saying here. More importantly, it's impossible to get a FCW while on AP/FSD, so it's definitely smart to stay in that mode for as much of the time spent driving as you possibly can, if all you care about is getting a near-perfect score. Also put your FCW into "early" mode to avoid getting to the "medium" FCW and you'll be fine.

 

[DrChaos]

I think what you are confused about is that Tesla gives you a score for a single day. If you have bad miles on that day and then drive on AP the rest of the day...the bad miles will be heavily weighted on the day....BUT if you drive 3 days in a row all on AP with nothing bad, the days are averaged together weighted by total miles...

So the 'warnings per day' is computed over non-AP miles only, but to compute the aggregate for the the month they do count AP-miles, inconsistent from the per-day statistic? That's really strange.

On the app it says "Rate of Forward Collision Warnings per 1,000 Non-Autopilot miles". It's unclear what that refers to and how that counts for the FCW score.

And what would be most predictive of risk rate? To me (yes I am a data scientist) the most predictive would probably be something as simple as "forward collision warnings per 30 days". More warnings == more driving, and more driving off AP. If there is any normalization, normalize by hour of moving vehicle.

Tesla: Still, the miles driven while on Autopilot are included in the total.

That might not be for the FCW Rate subscore (wouldn't make sense to me to be inconsistent in handling AP vs non AP). It could be for the usual insurance pricing where more miles means more premium, even in California where the SS itself isn't used for setting premiums.

so yes over the course of 30 days AP miles with no incidents with non-AP miles will level out the bad, but not on a single day or drive.

There are maybe 1,000 examples of this being true and Tesla explains that's how it's structured.

 

[DrChaos]

Exactly, they assist in drowning out the bad miles.

I see the opposite on the Tesla site:

Safety Score Beta | Tesla Support

The Safety Score Beta is an assessment of your driving behavior based on several metrics called Safety Factors. These are combined to estimate the likelihood that your driving could result in a future collision. We combine your daily Safety Scores (up to 30 days) to calculate the aggregated...

www.tesla.com

The Safety Score formula ignores any events that occur while Autopilot is engaged (including 5 seconds after the Autopilot is disengaged). This does not include Forced Autopilot Disengagements or Late Night Driving, which will impact your Safety Score. Any events that occur while a driver is manually engaging the acceleration pedal while on Traffic Aware Cruise Control will count towards the Safety Score.
Miles driven while Autopilot is engaged are not used to determine the Forward Collision Warnings per 1,000 Miles. Miles driven on Autopilot are not considered when calculating the mileage-weighted Safety Score.

Interesting to the nerds: the model of is exponential in the various subscores, or more likely (how it was developed), linear in the logarithm of the prob of collision (might actually be log odds but for small probs they are equivalent).

The coefficients show you the relative importance (take log for relative, and log(1+x) =approx= x for small x)

1) Forced AP disengagement: 1.40
2) Hard braking 1.16
3) Late night driving 1.05
4) Aggressive turning: 1.015
5) FCW: 1.0125
6) Speeding: 1.009397
7) Unbuckled: 1.00901
8) Unsafe following: 1.0024

I'm really surprised about how low speeding is. Speeding might not be correlated to frequency of collision but it's hugely correlated with damage/liability per collision and premiums should be priced accordingly.

 

[clydeiii]

“Not considered” does not mean “thrown out”. It means they are never going to impact your score.

 

[DrChaos]

“Not considered” does not mean “thrown out”. It means they are never going to impact your score.

I don't understand which distinction you're making, can you elaborate?

The question here as I understand it: "to improve Safety Score after FCW events should I drive on AP for many miles later in the month, or should I drive off AP for many miles". Some people say the first, "drive ON AP". I think the second may be correct, "drive OFF AP". (Assuming no violations).

What do you think?

 

[cbdream99]

"It is important to note that miles driven while Autopilot is engaged are not used to determine the Forward Collision Warnings per 1,000 Miles. Additionally, driving on Autopilot (including 5 seconds after Autopilot is disengaged) will not be factored into the Safety Score formula. Still, the miles driven while on Autopilot are included in the total."

(cite)

All I've ever been saying here. More importantly, it's impossible to get a FCW while on AP/FSD, so it's definitely smart to stay in that mode for as much of the time spent driving as you possibly can, if all you care about is getting a near-perfect score. Also put your FCW into "early" mode to avoid getting to the "medium" FCW and you'll be fine.

I am just mentioning how the score is calculated NOT whether FCW will be encountered when in FSD/AP.

From 1.0 FAQ
"Miles driven while Autopilot is engaged is used to determine the Forward Collision Warnings per 1,000 Miles. Miles driven on Autopilot are also considered when calculating the mileage-weighted Safety Score."

From 1.2 FAQ

"Miles driven while Autopilot is engaged are not used to determine the Forward Collision Warnings per 1,000 Miles. Miles driven on Autopilot are not considered when calculating the mileage-weighted Safety Score."

 

[clydeiii]

I am just mentioning how the score is calculated NOT whether FCW will be encountered when in FSD/AP.

From 1.0 FAQ
"Miles driven while Autopilot is engaged is used to determine the Forward Collision Warnings per 1,000 Miles. Miles driven on Autopilot are also considered when calculating the mileage-weighted Safety Score."

From 1.2 FAQ

"Miles driven while Autopilot is engaged are not used to determine the Forward Collision Warnings per 1,000 Miles. Miles driven on Autopilot are not considered when calculating the mileage-weighted Safety Score."

Yes, proving all other things count when in AP mode. Plus you can’t get FCW on AP so just drive on AP. So simple.