Tesla’s Safety Score Algorithm Is **Extremely** Sensitive

Sign up for daily news updates from CleanTechnica on email. Or follow us on Google News!

Like seemingly everyone who has purchased the Tesla “Full Self-Driving” package, I’ve been driving as extra cautiously as possible since Saturday — “like a grandma” as many of us have been saying. I think I already drive extra cautiously anyway, but I’ve been driving extra extra cautiously and watching every potential way I could lose points on my Safety Score on a drive. The aim is to be at 100 or as close to 100 as possible by the time Tesla starts rolling out FSD Beta access to more drivers.

On the first day, as I already wrote, a light that turned yellow right at the worst time and I had to stop quickly or risk running through the light at about the time it’d turn red. I chose to stop quickly, immediately expecting it might mess up my Safety Score a bit, and it did. I got a score of 97 instead of 100 on Saturday because of that. I’ve since built my average up to 99, but I’ve also seen a lot of complaints on Twitter about people being penalized for almost nothing. In fact, some have not even seen how they’ve lost points. One of the most notable things I’ve seen is people complaining that any non-regen braking got them to lose points. I wasn’t sure whether to believe that — they seemed to be certain of it, but that would be so extreme and illogical. And then …

This morning, after dropping my daughters off at school, I was coming up to a light that I know well. It’s a turn light and I know it is only green for a short moment. It was green and I knew there was a very high chance it would turn yellow too soon for me to go through, so I slowed down a little bit, but not so slow to cause trouble or piss off anyone coming up behind me. Sure enough, it turned yellow and I had to stop. I only slightly put on the brakes myself, mostly relying on regen braking but putting my right foot down a tiny bit to stay behind the white line. It was not a sudden stop at all. Nonetheless, I immediately wondered if I would lose points for it and growled (or worse). Yep. Dinged! I lost 2.1% for braking naturally — not even suddenly — for a yellow light I was completely ready for. Grrr.


On the way home from picking my daughters up, after making a left turn across traffic, I came upon a crosswalk where some kids and parents were on the left side of the road and starting to cross the street. I could have driven through without hitting anyone or even making them stop, but it would have been rude, unpleasant, and illegal of course. They might have paused crossing the first lane going in the opposite direction if they didn’t see me slowing down. I let the regen do the job a bit in order to use as little of the brake as possible, and then I barely skimmed the brake pedal with my right foot to stop well in advance of the crosswalk and make it clear to them that I was not moving.

Algorithm! Why do you forsake me so?! I lost 2.7% on my Safety Score on that drive because of that. (Ironically, I almost never have to stop at that crosswalk, and never for so many people as today, so I was smiling and in fury that this happened this week.)

Anyway, the point is not only to complain (or not entirely to complain). What is clear is that Tesla’s Safety Score algorithm is super sensitive. It’s looking for perfection, perfection that is basically disconnected from reality at times. It is also encouraging people to be exceptionally aware of the situation around them and ahead on the road in order to prevent even the slightest risk of accident (unless you count encouraging people to run yellow lights and drive through crosswalks). You need to be a bit lucky, but you also need to be tremendously aware and cautious in order to use regen braking ~100% of the time, not to mention avoiding the other things that can cut your score down (I have no experiences to share losing points on them, though). To keep the FSD Beta program accident free and to include more and more people who can help train the system in good driving, this is the population you want to add to the program next.

That said, the algorithm should also be adjusted a bit. While I am joking about it, for someone trying to get 100%, you are essentially encouraged to drive through scenarios where it’s better to just step on the brake a bit. In my opinion, the system should take into account to some degree what Tesla FSD Beta would do in various situations. Of course, deciding when to trust FSD Beta and when to just use a hard rule like “no braking” is surely quite difficult and complicated. However, I hope and expect Tesla will keep refining this system and will use it indefinitely to encourage safe driving. It is very effective — as gamification is often shown to be — at getting people to modify their behavior in order to get a high score. Also, after just a couple of weeks, some of those changes could become second nature and lifelong habits. That would be a big win beyond simply evaluating who to let into a beta program first.

In the meantime, though, I’m just going to cuss out the algorithms any time I have to touch the brake for something.

Have a tip for CleanTechnica? Want to advertise? Want to suggest a guest for our CleanTech Talk podcast? Contact us here.

Latest CleanTechnica.TV Video

CleanTechnica uses affiliate links. See our policy here.

Zachary Shahan

Zach is tryin' to help society help itself one word at a time. He spends most of his time here on CleanTechnica as its director, chief editor, and CEO. Zach is recognized globally as an electric vehicle, solar energy, and energy storage expert. He has presented about cleantech at conferences in India, the UAE, Ukraine, Poland, Germany, the Netherlands, the USA, Canada, and Curaçao. Zach has long-term investments in Tesla [TSLA], NIO [NIO], Xpeng [XPEV], Ford [F], ChargePoint [CHPT], Amazon [AMZN], Piedmont Lithium [PLL], Lithium Americas [LAC], Albemarle Corporation [ALB], Nouveau Monde Graphite [NMGRF], Talon Metals [TLOFF], Arclight Clean Transition Corp [ACTC], and Starbucks [SBUX]. But he does not offer (explicitly or implicitly) investment advice of any sort.

Zachary Shahan has 7344 posts and counting. See all posts by Zachary Shahan