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

Discussion of statistical analysis of vehicle fires as it relates to Model S

This site may earn commission on affiliate links.
...what I will accept is that it's a relatively low percentage of all crashes that are caused by vehicle-related road debris, the same way I'd accept that it's a relatively low percentage of collisions with road debris that cause vehicle fires, which I'd not have required a 148 page study to come to that conclusion. It's simply a matter of common sense, a little intelligence and life experience.

well we're in agreement on this.

when you say "it's a relatively low percentage of all crashes that are caused by vehicle-related road debris", would you be willing to put an upper bound on what that means to you? does it mean at most 1/2% of all crashes? at most 1%? 2%?

and, would you be willing to put an upper bound on the"it's a relatively low percentage of collisions with road debris that cause vehicle fires"? does that mean at most 5% of these collisions with road debris cause fires? 10%? 30%?
 
@luvb2b
First of all, nice work.
I haven't read the 11+ pages prior to this post so please forgive me if this has already been discussed.
Assuming that some auto buyers shop within (or near) a vehicle class, is there data to support a comparison by class and/or model year?
For example, is data available to compare luxury sedans, sport sedans? 2013 models vs. older models? This question is more from a buyer/shopper perspective, less from an investor context.

Again, nice work!
 
Guys, this whole conversation is completely off the rails at this point.....

Ha- True, there is a bit of derail-age. I have to express my gratitude for the thread and luvb2b work on it though. Just from my seat here (not imposing this on anyone of course), it's enlightened a dark corner. The statistical conclusions are sound in my mind after checking the work extensively. My background for reference includes Aerospace engineering with some substantial work in predictive signal processing and similar- so not statistics but some related math.

That said, I believe Bonnie also nailed it here; The focus is better served on what possible conclusions there are for each investor- Mine seem to be at odds with most of the posters, (and strike me as perhaps a source of mis-communication - not sure).
For me the conclusion (and I know many are not willing to form this conclusion currently) is that ModS currently is more susceptible than ICE to large road debris, undercarriage, collision fires that apparently produce no significant additional risk of injury (due to the excellent fire suppression rate design elements) - This in my mind is an extremely positive one. It identifies for both the consumer and investor the issue to consider and it fails as a long term issue for Tesla, in my mind (where before the there was uncertainty). The same data shows ModS to be very much less susceptible to fires from other, more common causes in ICE. If the risk of a non-injurous fire is greater than ICE for a narrow event category then no worries (I lived happily in FL knowing my risk of a lightening strike was an order of magnitude greater than anywhere else for example). This will all take some time to play out, some education, etc thus I simply stayed long and moved LEAPS out further (as I also belief most of this has already been discounted in the stock price).

I really see this as a positive enlightening analysis that (to luvb2b's) point would have better served at an earlier time- but given the stock price now, it really serves a positive conclusory in my mind. In fact I disagree with luv's decision to remain uninvested at these levels. I see downside risk here 20% of the upside potential, including with my belief there will be another collision fire in next months- (thanks for that forward projection work luvb2b)... those are my 2c
 
Last edited:
Ha- True, there is a bit of derail-age. I have to express my gratitude for the thread and luvb2b work on it though. Just from my seat here (not imposing this on anyone of course), it's enlightened a dark corner. The statistical conclusions are sound in my mind after checking the work extensively. My background for reference includes Aerospace engineering with some substantial work in predictive signal processing and similar- so not statistics but some related math.

That said, I believe Bonnie also nailed it here; The focus is better served on what possible conclusions there are for each investor- Mine seem to be at odds with most of the posters, (and strike me as perhaps a source of mis-communication - not sure).
For me the conclusion (and I know many are not willing to form this conclusion currently) is that ModS currently is more susceptible than ICE to large road debris, undercarriage, collision fires that apparently produce no significant additional risk of injury (due to the excellent fire suppression rate design elements) - This in my mind is an extremely positive one. It identifies for both the consumer and investor the issue to consider and it falls as a long term issue for Tesla, in my mind (where before the there was uncertainty). The same data shows ModS to be very much less susceptible to fires from other, more common causes in ICE. If the risk of a non-injurous fire is greater than ICE for a narrow event category then no worries (I lived happily in FL knowing my risk of a lightening strike was an order of magnitude greater than anywhere else for example). This will all take some time to play out, some education, etc thus I simply stayed long and moved LEAPS out further (as I also belief most of this has already been discounted in the stock price).

I really see this as a positive enlightening analysis that (to luvb2b's) point would have better served at an earlier time- but given the stock price now, it really serves a positive conclusory in my mind. In fact I disagree with luv's decision to remain uninvested at these levels. I see downside risk here 20% of the upside potential, including with my belief there will be another collision fire in next months- (thanks for that forward projection work luvb2b)... those are my 2c

+1 I agree Ken, thanks for your input.
 
You seek to call into question the integrity of the car, the company, the CEO, and everyone who is favorable toward them. You are working overtime here to convince the world that the car is statistically prone to fire and therefore ultimately unsafe.

i will defend myself on this one. from the very first post, i have said that the data support elon's conclusions regarding overall safety. i dug through all of the source documents, retrieved the source statistics, and ran the calculations myself.

i have shown how it can be demonstrated to a 99% confidence level that model s is safer than ice from the standpoint of mechanical or electrical fires.

i have mentioned that i own a model s, that i am very happy with it, and that i am not concerned about a fire (unless of course i am in a collision).

the one angle i am questioning is the risk of collision related fires. that's the line of research people have disagreed with, and i have posted numerous rebuttals. maybe that's why it dominates the discussion, because it's the point of disagreement.

You seem to further the implication that this fire issue (and the accompanying "lies" of the CEO) may be a problem which cannot be feasibly addressed, let alone "fixed" by Tesla Motors, even with a recall.

i said no such thing. in fact, i mentioned that the worst case scenario i envision is 25,000 teslas are recalled and have to go through $10,000 of retrofits. that's a ** worst case ** dollar estimate of $250 million.

you're questioning my motives and saying i must be a short. i've mentioned repeatedly that the stock has already lost several billions of market value since this incident, and that in my opinion the shorting opportunity has long since past. when the mit article came out yesterday, and there was no reaction, i mentioned that the market is reacting as if there's nothing new there (meaning, the information is already priced into the stock).


i have offered pretty much zero in the way of investment implications long or short, other than saying i don't feel compelled to be a long or a short in the stock at the moment.


what i sense here is there is some bizarre dread of short-sellers. have you forgotten how much money we made on the back of the shorts in from april to september? do you really think we could have done that well if there weren't mountains of shorts who were forced to drive the stock higher? for as much money as all of us have made at the shorts' expense, we should welcome them with open arms.


now those of you who want investment implications, you could look at it a number of different ways.


1. you could say, yes, there is a higher risk of collision related fires, but overall the data shows the tesla is definitely safer. the stock has lost multiple billions of market cap and so it's more than discounted this problem related to the collision related fires. so tesla is a buy.


2. you could say, the statistical significance of collision related fires is high enough that nhtsa might find a real problem. perhaps the perception of this will keep the stock from going up a lot until the investigation is complete. so maybe short term call options are the wrong play, maybe it's better to be in leaps or just hold the stock.


3. you could say, if nhtsa finds a real problem, the stock will go down on that news. but i already know the car is safer overall, so if that happens i'm going to be ready to buy when it does.


4. you could say well teslas are more likely to catch fire in an accident, and because of this people will stop buying them, and so i should short.


5. you could say, luv is just wrong, those numbers are all flawed, and i am a long term holder anyway, so i don't do anything but hold.

6. or something else. that is entirely up to you.

- - - Updated - - -

Ha- True, there is a bit of derail-age. I have to express my gratitude for the thread and luvb2b work on it though. Just from my seat here (not imposing this on anyone of course), it's enlightened a dark corner. The statistical conclusions are sound in my mind after checking the work extensively. My background for reference includes Aerospace engineering with some substantial work in predictive signal processing and similar- so not statistics but some related math.

That said, I believe Bonnie also nailed it here; The focus is better served on what possible conclusions there are for each investor- Mine seem to be at odds with most of the posters, (and strike me as perhaps a source of mis-communication - not sure).
For me the conclusion (and I know many are not willing to form this conclusion currently) is that ModS currently is more susceptible than ICE to large road debris, undercarriage, collision fires that apparently produce no significant additional risk of injury (due to the excellent fire suppression rate design elements) - This in my mind is an extremely positive one. It identifies for both the consumer and investor the issue to consider and it falls as a long term issue for Tesla, in my mind (where before the there was uncertainty). The same data shows ModS to be very much less susceptible to fires from other, more common causes in ICE. If the risk of a non-injurous fire is greater than ICE for a narrow event category then no worries (I lived happily in FL knowing my risk of a lightening strike was an order of magnitude greater than anywhere else for example). This will all take some time to play out, some education, etc thus I simply stayed long and moved LEAPS out further (as I also belief most of this has already been discounted in the stock price).

I really see this as a positive enlightening analysis that (to luvb2b's) point would have better served at an earlier time- but given the stock price now, it really serves a positive conclusory in my mind. In fact I disagree with luv's decision to remain uninvested at these levels. I see downside risk here 20% of the upside potential, including with my belief there will be another collision fire in next months- (thanks for that forward projection work luvb2b)... those are my 2c

+1

ken, thank you very much for posting that. i was hoping to provide a helpful objective analysis of the situation where i saw one lacking. i worked very hard to be objective in carefully evaluating as many reasonable sources of data as i could find. i tried to present my work so that someone knowledgeable or motivated could independently verify it.

it's sad to watch people getting blown away in short term options week after week when they don't have a handle on the data. i was hoping posting a good objective analysis would allow people to come to their own conclusions, just as i came to mine. i'm glad you found the research helpful.

as i said in the first post: "to my numerous acquaintances on the board, i apologize that my vacations have kept me from doing this analysis sooner, as i'm sure it would have saved many people much money since the third fire was reported. but hey, you get what you pay for... i've been lurking on the forums and i've seen some good statistical information posted and discussed. my purpose here is to present the various points of view, point out the subtle differences, and present statistical data to support."

thank you for a fabulous positive note, on your closing note i feel it is time for me to stop posting to this thread for the near future.
 
... i was hoping to provide a helpful objective analysis of the situation where i saw one lacking. i worked very hard to be objective in carefully evaluating as many reasonable sources of data as i could find. i tried to present my work so that someone knowledgeable or motivated could independently verify it.

... i was hoping posting a good objective analysis would allow people to come to their own conclusions, just as i came to mine.

mission accomplished...
don't stay away too long and have a great vacation!!
 
I think point #1 is what I was trying to make at around page #41 or smth ;) With two or three fires there just is no statistical way to claim either. The car could be far far safer or far far worse, but at this level of statistics both hypothesis are consistent with observation and therefore no conclusion can be drawn.

THANK YOU!

And yet the conversation drags on, as apparently this point is dodged time and again.
 
Last edited:
MIT vs Musk: The case of the missing numbers

(originally posted on the Tesla Motors forum)

Both MIT Technology Review's numbers and Tesla's Elon Musk's numbers leave a lot of room for improvement.
So had some fun with them.
My improvements are marked with [!]

Important Notice: 2 or 3 fires is really to small a sample to deliver reliable statistics! But I did an attempt anyway :)

Elon Musk's Numbers, Fine-tuned

Actually, these numbers originate from this Automotive News article:
http://www.autonews.com/article/20131108/BLOG06/131109827/tesla-firetraps-numbers-dont-back-it-up#axzz2lOMLFo26


My improved(?) calculation:
Facts:
250,000,000 vehicles total
190,000,000 passenger cars (short wheel base) [!]
20,000 Model Ses

=> 1 in 9,500 cars is a Model S

190,000 car fires / year
3 Model S fires / year
Chance to catch fire (not related to collisions):
Other cars => 190,000 / 190,000,000 = 0.001 = 1 in 1,000 cars catches fire
Model S => 3 / 20,000 = 0.00015 = 1 in 6,667 Model Ses catches fire

=> Model S is over 6 times less likely to catch fire than another car


MIT's Numbers, Fine-tuned

This article factors "collisions" in the equation. Let's see what we get.
http://www.technologyreview.com/view/521916/early-data-suggests-collision-caused-fires-are-more-frequent-in-the-tesla-model-s-than/

My corrected “MIT” numbers:

MIT Technology Review asserts that only 4% of fires are caused by a collision. However, what they do not mention is that there is no proof that driving over debris is categorized as a collision in the source article they're using. They are just assuming it is! Well, I'm assuming that the word "collisions" actually means "accidents". Let's see what we end up with, using this fresh look on the case.

First, let's look at other cars:
6,000,000 accidents / year [!]
190,000 car fires / year [!]
7,600 car fires / year caused by a collision (4%)

Chance that an accident causes a fire in other cars:
=> 7,600 / 6,000,000 = 0,001267 = 1 in 789 accidents cause a fire

Then let's look at the Model S:

=> 6,000,000 / 9,500 = 632 Model S accidents predicted per year (linear extrapolation based the numbers above)​

632 accidents / year [!]
3 car fires / year
1 car fires caused by an accident (33%) [!]

Chance that an accident causes a fire in a Model S:

=> 1 / 632 = 1 in 632 accidents cause a fire

=> Model S is almost equally likely to catch fire in a collision than other cars

Note 1: These are US-only numbers.
Note 2: The last year, there have been only 3 Model S fires. I know they all happened within 2 months, but that's how statistics are sometimes :) (see important notice above)
Note 3: As others have remarked, it would be better to compare numbers of equally new and expensive cars instead of using the complete US-car park as a reference. Sadly, I don't have those numbers, although I tried to get them.

References:



 
Last edited:
IMO the Elon's numbers are correct rather than MIT's numbers for the following reason:

The probability that a car catches fire should be considered in absolute and not only fires starting after collisions should be considered. This hypothesis should be done in first place because otherwise we should have two statistical calculations, one for cars catching fire spontaneously (meaning reasons not related to collisions) and one for car catching fire after collision.
And in the first statistical calculation the Model S would be the absolute winner
 
Last edited:
well we're in agreement on this.

when you say "it's a relatively low percentage of all crashes that are caused by vehicle-related road debris", would you be willing to put an upper bound on what that means to you? does it mean at most 1/2% of all crashes? at most 1%? 2%?

and, would you be willing to put an upper bound on the"it's a relatively low percentage of collisions with road debris that cause vehicle fires"? does that mean at most 5% of these collisions with road debris cause fires? 10%? 30%?

LOL! You are a funny man.

The actual relatively low percentage number is of no consequence because by its very nature it's relatively low. In other words, in the scheme of things it's irrelevant and therefore has no significance.

See where I'm going with this now?
 
The number of car fires per thousand accidents is well documented at 2.9/1,000. No need to caculate it, the NHTSA has done it for you. If people wreck Teslas at the same rate they do other cars, then the Tesla numbers ocome ut to about 3 fires for 472 accidents. That is 6.36 fires/1,000 accidents. So Tesla is over 2x more likely to catch fire in an accident than a comparable ICE car using the limited data available. If we had solid data on underbody strikes, then we could do that comparison as well. Hopefully the government has those numbers and can do a full analysis.

- - - Updated - - -

IMO the Elon's numbers are correct rather than MIT's numbers for the following reason:

The probability that a car catches fire should be considered in absolute and not only fires starting after collisions should be considered. This hypothesis should be done in first place because otherwise we should have two statistical calculations, one for cars catching fire spontaneously (meaning reasons not related to collisions) and one for car catching fire after collision.
And in the first statistical calculation the Model S would be the absolute winner

Musks numbers aren't correct for several reasons. First, they include all ice vehicles and not just passenger vehicles. Second, they don't account for intentionally set fires. Third, they don't account for vehicle age. Fourth, they compare general risk of fire and the NHTSA is interested only in a very specific case, whcih is why they are doing their investigation. You are right that we should have multiple sets of stats for multiple different situations. In fact we do have that and in the situation of underbody strike, the NHTSA and others think there is cause for investigation. You are also right that the MS is the winner in the first caculation. Overall, risk of fire in ICE is probably quite low, but that doesn't mean there isn't a unique situation that needs to be addressed. Ergo, I disagree with most of your assessment, but agree that the stats should be viewed in multiple ways in order to gain a full understanding of the situation.
 
Last edited:
Musks numbers aren't correct for several reasons. First, they include all ice vehicles and not just passenger vehicles. Second, they don't account for intentionally set fires. Third, they don't account for vehicle age. Fourth, they compare general risk of fire and the NHTSA is interested only in a very specific case, whcih is why they are doing their investigation. You are right that we should have multiple sets of stats for multiple different situations. In fact we do have that and in the situation of underbody strike, the NHTSA and others think there is cause for investigation. You are also right that the MS is the winner in the first caculation. Overall, risk of fire in ICE is probably quite low, but that doesn't mean there isn't a unique situation that needs to be addressed. Ergo, I disagree with most of your assessment, but agree that the stats should be viewed in multiple ways in order to gain a full understanding of the situation.

In a statistical calculation of any kind main thing is the starting hypothesis. Elon's statistical calculation is done considering ALL fires and, starting from this hypothesis, his statistical calculation is correct.
Then you can agree or disagree with the starting hypothesis, but that it's another matter. Of course if you want to change the starting hypotesis another statistical calculation has to be done.
In this particular case the difference between Elon's statistical calculations and MIT's statistical calculation consists in the fact that, while Elon considers all fires, MIT considers only fires starting after collisions and I pointed out that, since cars catch fires for many reasons besides collisions, the MIT starting hypothesis doesn't match to the reality and for this reason Elon made a better statistical calculation.
Then of course you can do other statistical calculations starting from other hypothesis. I am happy that you agree with me on the fact that if we consider as starting hypothesis fires starting spontaneously the Model S is the winner.

- - - Updated - - -

The number of car fires per thousand accidents is well documented at 2.9/1,000. No need to caculate it, the NHTSA has done it for you. If people wreck Teslas at the same rate they do other cars, then the Tesla numbers ocome ut to about 3 fires for 472 accidents. That is 6.36 fires/1,000 accidents. So Tesla is over 2x more likely to catch fire in an accident than a comparable ICE car using the limited data available. If we had solid data on underbody strikes, then we could do that comparison as well. Hopefully the government has those numbers and can do a full analysis.

In this case, since you are doing a very specialized starting hypothesis, IMHO two cases should be considered.

1) Cars catching fire after collisions. In this case Tesla numbers is 1 fire for 472 accidents. In fact only the accident in Mexico should be considered in this case to be fair because the other two fires started because of underbody strikes (road debris) and not for normal collisions. This would mean a probability of 2.11 fires/1000 accidents and Tesla is always ahead ICE cars.

2) Cars catching fire because of road debris (underbody strikes). Sorry not to have numbers to compare for this case.

I would like to add that if you thought that fires starting because of normal collisions and fires starting because of road debris (underbody strikes) should be considered in the same statistical calculations I would disagree with you.
 
Last edited:
1) Cars catching fire after collisions. In this case Tesla numbers is 1 fire for 472 accidents. In fact only the accident in Mexico should be considered to be fair because the other two fires started because of underbody strikes (road debris) and not for normal collisions. This would mean a probability of 2.11 fires/1000 accidents and Tesla is always ahead ICE cars.

I would like to add that if you thought that fires starting because of normal collisions and fires starting because of road debris (underbody strikes) should be considered in the same statistical calculations I would disagree with you.

Even if you don't want to include the fire in Mexico, which is not correct given where the damage occurred that caused the fire, you have 2/472, which is 4.24. Your last point makes sense and would have the effect of greatly magnifying the risk of fire to Tesla from underbody strikes.
 
Even if you don't want to include the fire in Mexico, which is not correct given where the damage occurred that caused the fire, you have 2/472, which is 4.24. Your last point makes sense and would have the effect of greatly magnifying the risk of fire to Tesla from underbody strikes.

Sorry maybe that I didn't say it good. English is not my mother tongue. It's the other way around. I consider the fire in Mexico as the only accident in the kind of collisions.
The other two fires originated because of road debris as you know.
That's why IMO my calculation is correct.
As I said the calculations for the road debris case cannot be done because I have no numbers to compare with ICE cars. If you manage to get such numbers we could compare Tesla with ICE cars also in this case.
 
Sorry maybe that I didn't say it good. English is not my mother tongue. It's the other way around. I consider the fire in Mexico as the only accident in the kind of collisions.
The other two fires originated because of road debris as you know.
That's why IMO my calculation is correct.
As I said the calculations for the road debris case cannot be done because I have no numbers to compare with ICE cars. If you manage to get such numbers we could compare Tesla with ICE cars also in this case.

I see. Thanks for clearing that up. I don't think it is your English, which seems darn good. I think it is just standard internet communication difficulty. Thanks for clearing up you point.
 
The number of car fires per thousand accidents is well documented at 2.9/1,000. No need to caculate it, the NHTSA has done it for you. If people wreck Teslas at the same rate they do other cars, then the Tesla numbers ocome ut to about 3 fires for 472 accidents. That is 6.36 fires/1,000 accidents. So Tesla is over 2x more likely to catch fire in an accident than a comparable ICE car using the limited data available. If we had solid data on underbody strikes, then we could do that comparison as well. Hopefully the government has those numbers and can do a full analysis.

There has been 1 Tesla crash that resulted in a fire. So that would be 1 fire from 472 crashes, which is not significantly different from 2.9 out of 1000.
However that 1 fire was in Mexico not the US, there is no data on the crash to fire ratio in Mexico, it is possibly different from the US. Perhaps emergency services response time is different, and response time affects the risk that a puddle of gasoline ignites. We just don't know. It's not quite apples and oranges, but its certainly 2 different varieties of apples.

There have been 2 Tesla fires as a result of underbody strikes, but if Tesla had not investigated them nobody would have known they were due to road debris and they would have just been another unexplained fire at the side of the road. As you just said, there is no publicly available data on fires resulting from underbody strikes, so there is no way for us to know if it is abnormal.

It is possible that no data exists to make that comparison at all, however I expect that insurance companies have very detailed statistics on all forms of risk. They probably don't have data on fire risk from road debris because they have no reason to do a detailed investigation of a cause unless they think they can find someone to bear financial responsibility.
If they insure a car that is destroyed by a piece of steel with a big "How's my driving" sticker on it, they have someone to go after. However if that piece of steel caused a fire, then the sticker is gone - so even if they found it, it wouldn't do them any good, so they won't even bother to look for it.
But they will have fire risk for cars by make, model and year. Comparing the Tesla to similar new vehicles would be the best way to tell if the Tesla has abnormal risk, but that information is proprietary and they won't give it away. Tesla and/or the government can likely get that data and decide if there is something to do, but I doubt any of us will ever see it.