Budget Battery Capacity Readout

[Landshark]

Good day all,

I finally found some time to make some changes to the analysis focusing on forecasting an accurate capacity reading based on all the data (which can now be found on the second page).

This method takes the average capacity readings for each user in each mileage category (every 5,000 miles) that has at least 3 entries and performs a linear regression to predict what the capacity will be. (nerd note: the correlation coefficient of this data is 0.97 so it's very well attuned to the linear regression method). if there were any categories that did not meet the criteria, the forecasted capacity is used to maintain the continuity of the forecast. This can be seen for reference in the upper left hand chart.

In short, you can now enter any mileage you wish and see what the projected capacity in Ah will be based on the aggregate data. You can also enter your current capacity in Ah and your current mileage to see how you compare to the the total population.

in the example below, I've entered my current capacity (40.38 Ah) and my current mileage (55,000) and see that my projected capacity should be 47.06Ah
This is depicted on the chart on the lower left as follows:
Yellow = Projected capacity/Entered mileage
Green = Your entered capacity in Ah
Red = the 36.6Ah warranty threshold

You can also see where you current capacity should fall on mileage based on the projection. In my case my capacity is far below what should be expected. What this shows is that based on the cumulative data the average person wouldn't expect to see my capacity until at least 125k miles (don't worry, I picked mine up used a couple months ago and am definitely an outlier in the data).

View attachment 19870

As MrFixIt mentions, this data will only become more accurate the more datapoints are entered, so please continue to contribute!

I’ll project that your methodology will produce inaccurate results. Your own vehicle is a perfect example. Why would you dismiss actual data as being an outlier? Sure, since you recently purchased the vehicle, you have no history of how it was operated. That unknown history holds the secrets of why the capacity is what it is today.

Battery degradation isn’t linear. Additionally, using a mileage interval as a variable to project future battery capacity in a PHEV is futile. One owner might drive 5000 miles on gas and another might drive the same distance on batteries. The difference between the two cars could be hundreds of charge cycles and you’d have no idea of discharge rates, operating temperatures, age of the batteries or storage conditions.

I get it that with enough data, there will eventually be an average to which others can compare, however, that average will have been derived from a simplistic methodology that isn’t likely to produce meaningful results.

 

[Thomas Clauser]

Hello Thomas - Thanks for making your excellent analysis tool even better !!

I would like to ask your opinion about some of the data (which I consider to be clearly anomalous) and may be having an influence on the statistics. There are 13 of these anomalous users currently. Some of them may already be excluded by some of your existing filtering rules. I consider these anomalous users to fall into 3 main categories. To find them, I went through all users and viewed each of their curves individually.

Category 1 - These users have had battery capacity "Resets". Something (perhaps a 12V battery discharge or long disconnect) caused these resets, and it takes a very long time for the reading to come back to 'normal'.
CLARITYDAVE1
LINCOMATIC
DANKS1
RANGAVASU1
KLIM1
JL-EV1 (probably a reset, but only 2 points so reset not completely apparent)

Category 2 - Implausibly high reading (typically only one reading that is implausibly high for the mileage)
EVA FARKAS1
JIMK1
VINCES1
DONNYD1
STACEY BURKE1

Category 3 - Questionable because of a 'perfect' starting measurement even though mileage was 'significant'
CIVAN1 ('perfect' at start, but with 12K miles)
ALTER1 ('perfect' at start, but with 22K miles)

Here is a graph of just these users (not all show well because of the default colors):
View attachment 19871

In an ideal world, I might propose something like this:

Category 1 users - should be ignored while they are 'recovering' from the 'reset', but that is often an extended period and hard to identify unambiguously.
Category 2 users - should be always excluded (perhaps any single-point users with capacities greater than 54 and miles > 10K)???
Category 3 users - Not sure about these. Keeping them probably doesn't hurt much because there are only 2 of them...

What is your opinion about the impact of these users who's data clearly seems 'wrong' , and whether additional filtering makes sense?

Hello Thomas - Thanks for making your excellent analysis tool even better !!

I would like to ask your opinion about some of the data (which I consider to be clearly anomalous) and may be having an influence on the statistics. There are 13 of these anomalous users currently. Some of them may already be excluded by some of your existing filtering rules. I consider these anomalous users to fall into 3 main categories. To find them, I went through all users and viewed each of their curves individually.

Category 1 - These users have had battery capacity "Resets". Something (perhaps a 12V battery discharge or long disconnect) caused these resets, and it takes a very long time for the reading to come back to 'normal'.
CLARITYDAVE1
LINCOMATIC
DANKS1
RANGAVASU1
KLIM1
JL-EV1 (probably a reset, but only 2 points so reset not completely apparent)

Category 2 - Implausibly high reading (typically only one reading that is implausibly high for the mileage)
EVA FARKAS1
JIMK1
VINCES1
DONNYD1
STACEY BURKE1

Category 3 - Questionable because of a 'perfect' starting measurement even though mileage was 'significant'
CIVAN1 ('perfect' at start, but with 12K miles)
ALTER1 ('perfect' at start, but with 22K miles)

Here is a graph of just these users (not all show well because of the default colors):
View attachment 19871

In an ideal world, I might propose something like this:

Category 1 users - should be ignored while they are 'recovering' from the 'reset', but that is often an extended period and hard to identify unambiguously.
Category 2 users - should be always excluded (perhaps any single-point users with capacities greater than 54 and miles > 10K)???
Category 3 users - Not sure about these. Keeping them probably doesn't hurt much because there are only 2 of them...

What is your opinion about the impact of these users who's data clearly seems 'wrong' , and whether additional filtering makes sense?

Here's what's great about what you guys have built/accumulated. With so much data (and hopefully more to come) any outliers become largely irrelevant to the trend.

Just for example purposes, here is that same analysis with ALL of the users you mention removed. The same inputs caused a change in results of less than 1% from the previous.

 

[Thomas Clauser]

I’ll project that your methodology will produce inaccurate results. Your own vehicle is a perfect example. Why would you dismiss actual data as being an outlier? Sure, since you recently purchased the vehicle, you have no history of how it was operated. That unknown history holds the secrets of why the capacity is what it is today.

Battery degradation isn’t linear. Additionally, using a mileage interval as a variable to project future battery capacity in a PHEV is futile. One owner might drive 5000 miles on gas and another might drive the same distance on batteries. The difference between the two cars could be hundreds of charge cycles and you’d have no idea of discharge rates, operating temperatures, age of the batteries or storage conditions.

I get it that with enough data, there will eventually be an average to which others can compare, however, that average will have been derived from a simplistic methodology that isn’t likely to produce meaningful results.

I'm not quite sure how any data can produce inaccurate results...it's just setting the expectation of what it produces...This isn't an analysis on battery degradation in a stable environment. This is an analysis of a data accumulated with many variables and projecting that out. It just so happens to fit a linear trend (because of all of those variables).

The only hard datapoints being accumulated are capacity, mileage, and date. There is EV%, that is fairly subjective. So those are the only datapoints being taken into consideration.

In my example what is shows me is out of 500+ datapoints, my car (and it's unknown history) are out of line from what the general population experiences on average. That is all this analysis will show and that's all that should be taken from it. If you track enough of your own datapoints, you will have a forecast for yourself and your driving habits/environment (on the individual results tab)

 

[MrFixit]

Your own vehicle is a perfect example. Why would you dismiss actual data as being an outlier?

My interpretation is that @Thomas Clauser isn't 'dismissing' his data, but he is simply pointing out that his vehicle is statistically farther from the average than most others.

Battery degradation isn’t linear. Additionally, using a mileage interval as a variable to project future battery capacity in a PHEV is futile. One owner might drive 5000 miles on gas and another might drive the same distance on batteries. The difference between the two cars could be hundreds of charge cycles and you’d have no idea of discharge rates, operating temperatures, age of the batteries or storage conditions.

Interestingly, the correlation coefficient of 0.97 does seem to imply that the data collected so far is looking fairly 'linear'. Not sure I would have expected that either, but the math is implying that it is pretty linear.

Using a mileage interval to project future battery capacity is not 'futile'... To be sure, it does represent a composite viewpoint of the entire population of Clarities, but that information is still quite helpful (particularly to a potential buyer who has yet to establish a modus operandi). If you have established your own trend (by contributing multiple data points to this effort), then you will see your own trend relative to the composite which enables you to mentally refine the composite projection to better represent your own expectations.

If Honda had provided a way to separate EV from HV miles, then it would probably be better to use EV miles as the variable rather than just the total, but sadly they didn't. The spreadsheet asks contributors to estimate EV vs. HV but few do that, and there is no standard methodology so I feel that trying to use estimated EV miles would only cloud the analysis more.

I get it that with enough data, there will eventually be an average to which others can compare, however, that average will have been derived from a simplistic methodology that isn’t likely to produce meaningful results.

In my opinion, this is neither simplistic or meaningless. You are certainly welcome to draw your own conclusions

 

[MrFixit]

Just for example purposes, here is that same analysis with ALL of the users you mention removed. The same inputs caused a change in results of less than 1% from the previous.

That's good news, and I believe it indicates that we have already gotten enough data (and more will continue to roll in) to minimize the effects of anomalous (or just plain 'bad') data. Thanks for your insights.

 

[Thomas Clauser]

To be sure, it does represent a composite viewpoint of the entire population of Clarities, but that information is still quite helpful (particularly to a potential buyer who has yet to establish a modus operandi).

Case in point......if I had known then what I know now I definitely wouldn't have bought my car (or at the very least, would've kept looking with the right expectation for capacity).

But then I wouldn't be here with all of you fine people! So it all worked out for the best.

 

[Landshark]

I'm not quite sure how any data can produce inaccurate results...

Not too long ago, a tremendous amount of data was analyzed. From that data, some of the nations greatest minds concluded that inflation was transitory. That data produced an inaccurate projection.

Manufacturers put together focus groups to collect data from consumers on what sort of products they’d like to see in the future. Typically, applicants fill out a questionnaire which reveals their age, gender, likelihood to buy a certain product, etc. The manufacturer/marketing team, carefully selects a group from the applicants that will likely produce the desired result.

There used to be a term called “Garbage in, garbage out.” Maybe that problem was fixed?

 

[ClarityBill]

Case in point......if I had known then what I know now I definitely wouldn't have bought my car (or at the very least, would've kept looking with the right expectation for capacity).

But then I wouldn't be here with all of you fine people! So it all worked out for the best.

Interesting idea that buying a used EV should be based on battery capacity, rather than odometer miles.

Seems obvious now, but I think it is an excellent outcome of the study.

Welcome to the family!

 

[MrFixit]

Case in point......if I had known then what I know now I definitely wouldn't have bought my car (or at the very least, would've kept looking with the right expectation for capacity).

Of course there is always a chance of the Silver Lining... If you go below the warranty threshold, you may wind up with a replacement and be better off than had you been able to 'cherry pick' a different vehicle. It will be interesting to see how this turns out for you !

 

[MrFixit]

Interesting idea that buying a used EV should be based on battery capacity, rather than odometer miles.
Seems obvious now, but I think it is an excellent outcome of the study.

With a true EV, all the miles are EV so miles may more directly relate to capacity.
It is possible for them to be quite different with the PHEV.

 

[Thomas Clauser]

Not too long ago, a tremendous amount of data was analyzed. From that data, some of the nations greatest minds concluded that inflation was transitory. That data produced an inaccurate projection.

Manufacturers put together focus groups to collect data from consumers on what sort of products they’d like to see in the future. Typically, applicants fill out a questionnaire which reveals their age, gender, likelihood to buy a certain product, etc. The manufacturer/marketing team, carefully selects a group from the applicants that will likely produce the desired result.

There used to be a term called “Garbage in, garbage out.” Maybe that problem was fixed?

Yep...You nailed it! That's why the second part of the sentence was so important:
"I'm not quite sure how any data can produce inaccurate results...it's just setting the expectation of what it produces..."

I am setting the expectation that the combined capacity and mileage data provided from people on a forum, hand typed on a spreadsheet can produce enough information to make a valid comparison of your personal data and also provide a projection of what you could expect in the future.

Then it's up the viewer of the data to figure out the 'why' if there is a difference. In my case, my wild guess that my previous owner never used a drop of gasoline and rode that battery hard every single day for 50,000 miles. I will get to see how that changes over time with me driving 60-70% of the time on battery.

As the data continues to grow and changes, so will the results it projects (and yes, perhaps even the methodology being applied).

Conversely, because there are so many unknowns and the P-value is so wildly high, I am not drawing the conclusion that says "this analysis proves that all Clarity's will reach 36.6Ah at 160k miles". That would be akin to your examples of inflation analysis.

The problem with many analysis (including the ones you mention) is that their P-value get manipulated (intentionally or not) so that a conclusion is drawn even when all the variables are undefined or discarded. That is not the case here. The analysis is not conclusive. It's indicative (or cohort) analysis.

Long story short.....data doesn't lie....but analysts sometimes do.

 

[Landshark]

I’ll project that your methodology will produce inaccurate results.

I'm not quite sure how any data can produce inaccurate results

To be clear, I stated that your methodology will produce inaccurate results, not that the data would produce inaccurate results. Interesting how the human mind can interpret things.

The data on its own can’t do anything. It certainly can’t lie. That is a uniquely human characteristic, although some species do engage in deceptive behavior.

I had intended to use either, meaningless or useless, either of which would have been a more accurate description, but being polite got the best of me. You can collect data and create a hypothetical average. A person can also use that average for comparison purposes. How meaningful is that average?

The average Bull Market lasts for about 5 years and 2 months, yet no Bull Market has ever lasted for 5 years and 2 months. The average doesn’t exist and has never occurred. While the data and analysis has produced accurate and factually true results, those results are useless to an investor.

The flip side of all this could be that there just isn’t that much variation in battery degradation between vehicles, regardless of how each owner operates their vehicle. So all this analysis may be simply confirming what battery manufacturers and Honda already know. They set the warranty threshold after all. I’ll project that the average Clarity will have a battery capacity of ~75% at the 10 year mark and that ~95% of those cars will have a capacity between 70-80%. Let’s revisit this in 2030 and compare notes.

Out of curiosity, when does your projection indicate that your battery capacity will drop below the warranty threshold? Would you still regret the decision to buy the car if you end up with a new battery at the 7-8 year mark?

 

[Landshark]

Interesting idea that buying a used EV should be based on battery capacity, rather than odometer miles.

I would recommend including battery capacity on the checklist along with total miles and all the other factors that one takes into consideration when shopping for a used vehicle.

 

[Thomas Clauser]

Out of curiosity, when does your projection indicate that your battery capacity will drop below the warranty threshold? Would you still regret the decision to buy the car if you end up with a new battery at the 7-8 year mark?

Current projection says 90k miles(+/-5k)....But to your point, that will depend on how my driving differs from the previous owner.

My concern is that until then (however many years that takes) I do seem to be plagued with extreme versions of what others report. I have trained myself never to deplete the battery. If I do, I am guaranteed to get the "angry bees" without fail every time. In some cases (like if I forget to switch to HV mode before the range hits 0) the battery has almost completely drained (0 bars instead of 2) and I literally have to pull over for fear of that engine noise when it goes to jet engine takeoff status.

What seems like the occasional nuisance to others, is my consistent experience so far. While I can't really conclude whether or not that is related to the fact that I am one of the lowest capacities in all of this data, it makes me wonder if I'll even make it to that warranty cutoff.

Time will tell.

Hey, thanks for the back and forth. I appreciate good skepticism and will certainly keep my eyes peeled for anything I can do to make the analysis as accurate as possible.

 

[Landshark]

Hey, thanks for the back and forth. I appreciate good skepticism and will certainly keep my eyes peeled for anything I can do to make the analysis as accurate as possible.

Thank you. I’ve always found the conversation to be more engaging when the room isn’t full of narrow minded people. Of course the acceptable term is “like minded”. So many people have become ostriches, where they refuse to talk about anything that makes them uncomfortable.

 

[coutinpe]

Question: My car (2018, 32K miles) kept capacity consistently at 51.26 since I joined this experiment until yesterday, when it suddenly dropped to 50.4. Is that the usual way you have seen? Thank you.

 

[MrFixit]

Question: My car (2018, 32K miles) kept capacity consistently at 51.26 since I joined this experiment until yesterday, when it suddenly dropped to 50.4. Is that the usual way you have seen? Thank you.

You only have 3 data points (the first not until 29K miles). You can't evaluate a trend over an interval of only 3K miles... Often the longer term curves are not totally 'smooth' and the behavior depends on driving patterns (for instance if someone goes on a 1000 mile trip, all on HV, it will create a flat spot in the curve).

Keep monitoring and an actual trend will emerge. You can look at MANY others to see what might be considered 'normal'. Your curve, given it's very short duration, is not abnormal at all.

If you go to the battery analysis tool CLICK HERE
Choose Page 3 with the arrow at the very bottom, you can compare individual curves. On the upper right is a pull-down to add or delete users from the curve.

 

[coutinpe]

You only have 3 data points (the first not until 29K miles). You can't evaluate a trend over an interval of only 3K miles... Often the longer term curves are not totally 'smooth' and the behavior depends on driving patterns (for instance if someone goes on a 1000 mile trip, all on HV, it will create a flat spot in the curve).

Keep monitoring and an actual trend will emerge. You can look at MANY others to see what might be considered 'normal'. Your curve, given it's very short duration, is not abnormal at all.

If you go to the battery analysis tool CLICK HERE
Choose Page 3 with the arrow at the very bottom, you can compare individual curves. On the upper right is a pull-down to add or delete users from the curve.

That was amazing! You have done a heck of a work there. Thanks!

 

[ClarityBill]

I made it past 200,000 miles, and wanted to get my battery information out there...

Looks like my battery is at 44.06 ah

 

[Domenick]

I made it past 200,000 miles, and wanted to get my battery information out there...

Looks like my battery is at 44.06 ah

It looks to me like you're at around 80% of your original battery, then. Thinking this is pretty decent for 200,000 miles.