Welcome to Tesla Motors Club
Discuss Tesla's Model S, Model 3, Model X, Model Y, Cybertruck, Roadster and More.
Register

Mid Range Battery Math

This site may earn commission on affiliate links.
Yet, we have fleet data that looks like this for the LR RWD and AWD (credit @KenC and the Stats app), showing that 6% “degradation” is within the normal distribution (though modestly towards one tail). Not an MR, but presumably the behavior of that very small fleet is similar, and aging may be slightly accelerated with a smaller battery, given the same number of miles accrued. Both time and discharge cycles are factors of course:

This is just data from people reporting typical range. Most people don't even charge to 100% and report data or don't take into account that different temperatures affect the battery. This data is not informative and is zero scientific

I can tell you for certainty that 6% so soon is not normal nor ok. I am at 2% and there are people with 40,000km+ like Bjorn Nyland who have around 4%. There are actually at least 3 reports of people with the same actual degradation at above 45000km.
 
I have to observe it. Looking at my screenshots from SMT they are always matching, even when there is a 1.5kWh difference between nominal and expected. Could it be that the IOS app delivers different CAN Bus values and calls them the same as SMT?

I believe they both probably provide the same data assigned to the same names, but that SMT just does the calculation for SOC wrong. Maybe you can provide some more screen shots to verify this.
 
This is just data from people reporting typical range.

It’s actually gathered automatically from Stats from what I understand. Exactly how Stats conditions the data (whether it’s an average of a few charges and limited to charges above a certain %, controlled for temperature, etc), I do not know. But they are data that are gathered dispassionately.

I can tell you for certainty that 6% so soon is not normal nor ok. I am at 2% and there are people with 40,000km+ like Bjorn Nyland who have around 4%

Everyone has to be part of the distribution! Plenty of data points exist. Some people get lucky and are on the high side! There are plenty of data points in the image above showing 2% “degradation” from 310 is quite normal. But there were a lot of people who started at 312-313 rated miles and a few look like they even had 315 (this is for dual motor). So it’s not quite as impressive if those same people are retaining 305 miles now. What did your 100% charge start at?
 
Last edited:
310/500km as everyone else...

That STATS app is so flawed, especially since it reports based on estimation from 80-90% and doesn't factor in temperatures, but to each their own.
Do you know if at the current state of my battery I pulled the stats app? It will probably show 10% degradation. But that is because the battery is at 6°. When I warm it up and push it to 100% I am at around 75-75.3kWh even in winter...

6% Degradation at 15,000km is not normal, for any battery. Not even the stupidest smartphone batteries after 50 cycles. Let alone a highly sophisticated ev battery with BMS and cooling.

But you can believe whatever you like.
 
Do you know if at the current state of my battery I pulled the stats app? It will probably show 10% degradation. But that is because the battery is at 6°. When I warm it up and push it to 100% I am at around 75-75.3kWh even in winter...

Of course. Temperature is important.

When looking at the data above, I am cognizant of these factors. But I am willing to look at the data with those limitations in mind.

I’m not sure what your resistance to the extrapolation from 90% is. I’ve always found it to be very accurate (I’ve charged to 100% quite a few times for road trips and I’ve never been surprised). I concede that there are cases where it may not be accurate due to BMS confusion. And it is not accurate to extrapolate when at a low SoC since the effect of rounding error is amplified.

Anyway, what I see in these distributions is that over time and distance there definitely is significant loss of available capacity taking place. Whether it is permanent or excessive, I do not know. Since this plot is done by odometer reading, the effects of temperature are actually randomized (people buy cars at different times and drive them different amounts). So they’ll create spread on the data independent of the odometer value. It’s convolved with the “degradation” distribution. Probably it’s the primary reason for the spread at low odometer readings (though that could also be the extrapolation error distribution).

It’s probably reasonable to look at the spread on the 0mi data point as mostly the overall data gathering error due to Stats not controlling for all of these factors. But it’s pretty easy to see that something else happens as the odometer reading increases.

I do wonder how many data points are in these distributions, and whether this plot tracks the same cohort of cars from 0 miles, or whether it is gathering different cohorts of vehicles for each range of miles traveled plotted.
 
Last edited:
Actually, I have the snowflake now. If I pull the tab on the new app to 100% charhe it shows me an estimated 455km. If I had the stats app imagine the offset...

Of course. No surprises there. I imagine that may be the reason for some of these tail data points. But not clear why such data points would not show for a vehicle with 1000mi or whatever. Makes me think some conditioning of the data may be done by Stats. Or maybe there just aren’t enough data points. Doesn’t really matter that much.

I certainly would not attempt to quantify firm expectations based on these plots. Too many unknowns. I’d just conclude that you’re likely to see your range decrease somewhat as mileage increases, and then level off at some point.
 
I believe they both probably provide the same data assigned to the same names, but that SMT just does the calculation for SOC wrong. Maybe you can provide some more screen shots to verify this.
Do you have screenshots where expected and ideal are different on your app? I am observing them right now and they move at the dame rate - identical. Not sure if SMT is reporting the wrong values. Anyways expected or ideal, at least we know it is not nominal

What I also observed is that when I charge on 11kW 3 phases it also pulls 7kW from the stators when the battery is cold up until around 13°. The motors are warming up the battery. So the car is actually pulling 18-20kW from the grid. This is almost 100% overhead. I don't have a reader for that, but will try to test it on a public charger with a meter.
 
@TimothyHW3 I'm confused.... if I take the range displayed and divide it by the percent battery remaining, I get the same number as when I actually charge to 100% (I've only done this ~3 times, so sample size is small). That's all Stats (and TeslaFi, etc.) are doing. Is there a better way?

One caveat - I live in Los Angeles, so my winter ambient temps are 10-15C. So although the battery does show some temp lockout if you look at the API data, it's usually about 1%
 
@TimothyHW3 I'm confused.... if I take the range displayed and divide it by the percent battery remaining, I get the same number as when I actually charge to 100% (I've only done this ~3 times, so sample size is small). That's all Stats (and TeslaFi, etc.) are doing. Is there a better way?

One caveat - I live in Los Angeles, so my winter ambient temps are 10-15C. So although the battery does show some temp lockout if you look at the API data, it's usually about 1%
Yes, you can take the dash range displayed and divide it by the dash percentage remaining and get the correct result, bounded by the rounding errors due to whole number rounding of the rated miles and % value.

What @TimothyHW3 and I are discussing is the fact the the Scan My Tesla app uses the wrong formula for calculating SOC %. The formula SMT uses for SOC is: (nominal remaining - buffer)/(Nominal Full - buffer), when it should be: (ideal remaining - buffer)/Nominal Full - buffer).
According to Tim, the ideal remaining value always seems to match expected remaining. If that is true, then of course you could use expected remaining value and get the same result.
I have always observed only small differences between ideal remaining, expected remaining, and nominal remaining values, maybe about 0.2 kWh spread among the 3 values, but they can all be different values in my experience. Tim has observed much bigger discrepancies between the values, which will result in a much bigger error in the SOC value in that case.

Bottom line is, if you want to get a more accurate SOC value than the whole number dash display, do the calculation from the SMT values yourself rather than just using the SOC reported value from Scan My Tesla.
 
@TimothyHW3 I'm confused.... if I take the range displayed and divide it by the percent battery remaining, I get the same number as when I actually charge to 100%

As far as I can tell, Stats uses three digits of precision for the % internally when projecting the full range. So from that perspective it seems like this method would be reasonably accurate. Assuming Stats is using the “right” calculation for SoC of course.

This projection/extrapolation method for full miles has better accuracy the higher the SoC %. That is one issue with that method. But as long as you only pay close attention to the number when your charge is above 50%, the precision error won’t be inflated too much. Once your rated miles drop below 100, for example, you only have two significant figures, so you can only say what your 100% range is within 10 miles or so.

Other than that, this method is pretty much equivalent to reading back the “fullkWh” from SMT or whatever.

However, with all of these methods, without charging to 100%, you are susceptible to BMS “confusion.” The BMS might be inaccurate in its estimate of what “fullkWh” is until you charge to 100%. There is not a way to know until you get to 100%.

Even if you do that, it may still be inaccurate until you discharge the battery to near 0% and then recharge it. How much error there could be there, I do not know. I only routinely discharge to 5-10% or so and I’ve never seen any significant adjustments after doing that.

As you say, I too have always found the projections of miles to be fairly accurate, and once you know the vehicle charging constant you can always calculate what a CAN bus reader would tell you within about 0.5%, without actually having one. The CAN bus reader doesn’t provide significant additional information on battery capacity - though it does provide a lot of other information of course. Just saying you don’t need one to be able to say what your fullkWh is - the screen in your car effectively tells you that. Can see that earlier in this thread, where we were able to figure out your battery capacity and your approximate buffer size before you received your CAN bus reader.
 
Last edited:
As far as I can tell, Stats uses three digits of precision for the % internally when projecting the full range. So from that perspective it seems like this method would be reasonably accurate. Assuming Stats is using the “right” calculation for SoC of course.

However, with all of these methods, without charging to 100%, you are susceptible to BMS “confusion.” The BMS might be inaccurate in its estimate of what “fullkWh” is until you charge to 100%. There is not a way to know until you get to 100%.

Even if you do that, it may still be inaccurate until you discharge the battery to near 0% and then recharge it. How much error there could be there, I do not know. I only routinely discharge to 5-10% or so and I’ve never seen any significant adjustments after doing that.


As you say, I too have always found the projections of miles to be fairly accurate, and once you know the vehicle charging constant you can always calculate what a CAN bus reader would tell you within about 0.5%, without actually having one. The CAN bus reader doesn’t provide significant additional information on battery capacity - though it does provide a lot of other information of course. Just saying you don’t need one to be able to say what your fullkWh is - the screen in your car effectively tells you that. Can see that earlier in this thread, where we were able to figure out your battery capacity and your approximate buffer size before you received your CAN bus reader.
That is why no one should get too concerned over a few miles of rated range variation over the short term. You really need to look at long term trends. With my car, after quickly stabilizing within a month or so to about 5% degradation, I went about two years or more before seeing any noticeable degradation again.

I don't know much about the Stats app, but if it is an API app, then the reading of SOC is the same as the car shows, i.e., rounded to a whole number. But it should be able to provide decimal point precision to the rated miles value.
 
But it should be able to provide decimal point precision to the rated miles value.

Ah that explains it. I noted that the displayed miles and % did not align with what Stats projected for 100%. For example (see below), 182/0.61 = 298 which is not 299. But 182.4/0.61 = 299. So Stats must be using that extra decimal precision on the rated miles.

So, that checks out. But it is really only two significant figures, technically. It’s about 299+/-3 since percentage could be 60.5 to 61.5.

Anyway, you can see the error % is reduced the higher is the value of the divisor. But in any case it’s about 1%, so not a big deal in this context. Or any context, really.
79339A08-DC0A-408E-B6BA-80D8D20076F9.png
 
I have always observed only small differences between ideal remaining, expected remaining, and nominal remaining values, maybe about 0.2 kWh spread among the 3 values, but they can all be different values in my experience. Tim has observed much bigger discrepancies between the values, which will result in a much bigger error in the SOC value in that case.
You are in SoCal, normal. Go to Colorado and leave the car in the cold - I had a difference of 1.5kWh today. I observed a charging session and both numbers expected and ideal moved together. Not sure what Amund is doing and wether the values are different, might need an update.
Amund knows about this bug in SMT, but not sure if he is planning on fixing it and change the calculation to expected/ideal instead of nominal remaining.
 
Of course. Temperature is important.

When looking at the data above, I am cognizant of these factors. But I am willing to look at the data with those limitations in mind.

I’m not sure what your resistance to the extrapolation from 90% is.
It is from experience. This guy here had a video that went viral just 2 months ago - claiming he has 2% degradation at 50,000 miles. Turns out he has 5-6% real degradation when he read the BMS.

Granted his battery is cold so it might be closer to 5%, but still.

It is better to read the data directly.

And this further proves the theory that having 6% after 10,000km is normal. It is not. All the data points towards maximum of 5% above 50,000km . At least from all the people that have good and balanced BMS and read it with SMT.
I am pretty sure that the 5-8% keep steady until about 100,000-150,000km and diverge from there
 
6% Degradation at 15,000km is not normal, for any battery. Not even the stupidest smartphone batteries after 50 cycles. Let alone a highly sophisticated ev battery with BMS and cooling.
I guess I could ask Tesla to take a look, but they will just look at the 6-7% of mine and say it is "within spec" since it's not more than the 30% warranty and the battery is showing no faults. Maybe when I get my HW3 upgrade I will have them look then.
 
This guy here had a video that went viral just 2 months ago - claiming he has 2% degradation at 50,000 miles. Turns out he has 5-6% real degradation when he read the BMS.

I couldn't find his video talking about his 2% degradation. But overall he does not appear to understand much about the car's battery (he seems to be under the impression it had 75kWh originally...you'd think if you were going to talk about degradation you'd figure out how much you started with...). And then in the end while he talked about his "fullkWh," he never mentioned what his projected 100% rated miles was! He just said he likes to leave it in %. :rolleyes: Which is fine, in general. But not if you want to quantify loss of capacity.

And this further proves the theory that having 6% after 10,000km is normal. It is not. All the data points towards maximum of 5% above 50,000km .

There's not really any information in the video about when his 5% loss occurred. For all we know it could have occurred in the first 15k miles.

In any case, there are plenty of examples here of people with more than this much loss of capacity. It's entirely possible these cases are BMS related, of course, but that does not change the fact that they have less usable capacity than others, and that there's no clear proven fix for these issues that works reliably.

Single data points are not really definitive either way. It's better to look at the distribution data we do have access to, to get a rough idea of what to expect. Obviously you can exclude the outliers, but the trends are relatively clear. The data says that if you expect 10% loss of capacity after 1-2 years & 20k miles, you have a very low probability of being disappointed, and will likely be happy with where you end up.
 
The video of his 50,000 is a little while back, just search his videos. It was even covered in some websites. He turned the comments off for obvious reasons.


He doesn't seem to understand a lot of things, but he will learn it. The projected is mentioned in the comments I think it was around 308 on a software upgraded RWD(325 update)
The 5 % can't happen so soon, no real BMS data has such indication. I have seen some people with apperent 10% at around 25,000km, but they had a lot of cell imbalance so either their BMS is miscalculated or their packs are faulty. Regular, normal working, batteries will see degradation after 20,000 miles or so and it will be around 2-3%.
 
The projected is mentioned in the comments I think it was around 308 on a software upgraded RWD(325 update)

So, 308/325 = 94.7%, so 5.3% loss of capacity. It seems like his video could have been a lot shorter, at least in regards to the loss of available capacity discussion. CAN bus reader seems useful for diagnosis and other power tracking, but no need for it to see what your available capacity is, so not sure why he was focused on that. And he never needed to make the 2% claims, since obviously he could see that it was higher than that (it would be kind of silly to do 308/310 for the math).
 
Last edited: