Tesla vs NY Times: A deeper look at the data
redmonk.com
redmonk.com
If I drove like a typical New Jersey-ian then my range would have been somewhere around 450 miles. With some adjustments (namely going around 65mph at most), I could regularly attain 700+ range. This is an adjustment many people absolutely refuse to make. Once I really learned how to get the most out of my car the milage numbers were unbelievable. When I sold the car my lifetime milage was 68 mpg over 130,000 miles (in nj climate).
I can see why Musk is so frustrated... he gets a reviewer that has no interest whatsoever in understanding the vehicle's prime operating conditions and instead is driving the car like a Hertz rental. When going on a long range trip, it should be obvious that you need to drive conservatively. Use the car's incredible performance for shorter trips so that you are not at risk.
I have no problems with these limitations, and I think they are acceptable. I was quite impressed that the range numbers for Tesla were over 500 miles... my current kia rio can't touch that despite getting nearly 50mpg. I have a feeling the media is never going to "get" electric cars because they don't get conservation.
I have to ask about this. The highest EPA estimates (which are notoriously generous) for the 2013 Rio are 37 MPG [1].
[1] http://www.fueleconomy.gov/feg/bymodel/2013_Kia_Rio.shtml
It's quite fun, actually!
Of course, if you get really creative... https://secure.drivers.lexus.com/lexusdrivers/magazine/artic...
There may well be something to say about Tesla's estimations being off, but this isn't it.
Did he? Because from this figure, it appears the estimated range is NOT dependent on speed (as, of course, it should be). Notice that the lines are extremely linear, with identical slopes throughout.
If he sped for part of the trip, why don't we see steeper slopes during those times?
So Broder "sped illegally fast." What does this even mean? The speed limit is 55 mph on most highways, but nobody goes at that speed. Can the Tesla range estimator not acknowledge speeds about 55 mpg?
Really? In this part of the country (pacific northwest) the speed limit on interstates is 60 mph in the city and 70 mph in the country.
But I think this is incorrect:
> In nearly every case, the Tesla gets about 20% fewer miles out of each charge than it estimated.
Musk and Broder 'agree' that the vehicle's actual range was greater than it estimated. Musk's wrote,
> Then he bizarrely states that the screen showed “Est. remaining range: 32 miles” and the car traveled “51 miles," contradicting his own statement... The car actually did an admirable job exceeding its projected range.
As an aside, please disable Tynt on your blog.
Precisely, the slope indicates the same thing. Unfortunately the vast majority of a general audience has forgotten what a slope is, so rather than confuse them, I explained it differently.
> Musk and Broder 'agree' that the vehicle's actual range was greater than it estimated.
Isn't that explained by the reserve power noted in Musk's blog?
The point, imho, is that the press is always trying to find a story also where there isn't. Sad for them, this time there's the hard data to support Tesla's facts.
I read a recent interview with a well-known landscape photographer who shoots in the Arctic - he carries a whole bandolier of batteries, kept close to his chest at all times to keep them warm. Apparently in the extreme Arctic cold the batteries are reduced to literally a few dozen shots before dying.
Deleted comment
> Estimation theory is an established field, and this is the most basic type of problem.
All estimations in this class must make assumptions about future behavior. Changes in behavior change the estimate.
> The reason for this is that the range estimator can't predict things the driver might do to spoil the forecast
Notice that all of the lines are extremely linear, meaning that Broder's behavior (or a miscalibration) caused a consistent bias in the estimate -- for 500 miles! It's very possible to design an estimator that can take this into consideration and make more accurate predictions.
> All estimations in this class must make assumptions about future behavior. Changes in behavior change the estimate.
Obviously. So what assumptions is it making? "My estimate has been consistently off for the last 250 miles, but it's going to be correct in the next 250 miles.... I can just feel it!"
> Changes in behavior change the estimate.
They should, but apparently they don't in this case.
They may make the estimate conservative when there is little charge left so people are more apt to charge up.
You'd think you wouldn't want to do this with a range estimation because the error's more obvious, but I'm also sure they don't want to sandbag its estimate from a full charge because people will be wondering why they're not seeing the advertised range. I'm sure they do want the car to be able to keep going a little past empty, though, and those spare miles have to come from somewhere.