Hacker News Reader

Comment by shawn-furyan | Hacker News Reader

How do you think that decisions are made under uncertainty? You can't always wait until you have very solid numbers to come up with an analysis. It's not "ridiculous" to come up with an order of magnitude estimate when better information is unavailable, it's a critical step in solving new problems.

A more productive criticism would have been to offer an improvement, rather than to just point to the imperfections in an attempt to estimate the future impact of a new technology.

edit: added second paragraph

View on HN

OK, but in 2016 we don't have to guess how many cars Tesla built in 2015. It's historical fact.

What is it? You could actually improve the estimate rather than /just/ pointing out that it has flaws. And your original criticism was vague. At least this one moves things forward.

After a quick Google search, the number of vehicles Tesla sold in 2015 does in fact seem to be a bit uncertain. It seems they sold over 50,000 vehicles, but I've not found a more precise number, in a few minutes of looking.

Sales from previous years seem to be:

2012 - 2,650

2013 - 22,477

2014 - 31,655

2015- ~50,000

though I got these numbers from a quora post, so take them with an adequately large dose of salt.

edit: added google findings, tweaked wording to be more specific

There is no reason to guess at these numbers OR to Google for other people's guesses. The sales figures to 5 significant digits are in Tesla's 10-K report. For example:

During 2015, we achieved significant efficiencies in Model S production and produced 14,037 vehicles in the fourth quarter of 2015. We began production of Model X in the third quarter of 2015. We plan to deliver 80,000 to 90,000 new Model S and Model X vehicles in 2016.

So... we're about where we were at with the very first estimate...

Precision isn't particularly helpful for this sort of rough projection. Which highlights that your original criticism wasn't particularly productive.

clarification: In claiming that we're about where we started, I'm saying that a 10-20% correction doesn't significantly impact a rough order of magnitude estimation. The estimate was in the precision bounds of its stated goal. Therefore the estimate wasn't "ridiculous" at all, or at least, you haven't shown that it was.

That's not why I thought it was ridiculous. I said it was ridiculous for using projected 2016 Tesla production and dividing that by actual 2013 battery production, when battery production is rapidly expanding. You either have to use 2016 projected battery production on the denominator, or actual 2013 Tesla production (a number very close to zero) on the numerator.

You've been criticizing the rough order of magnitude estimate for an hour and still haven't offered a significant improvement. I find that more more ridiculous. If the improvement is so obvious that not including it is ridiculous, then why not just produce the improvement you're talking about?

Note: This is rough, back of the envelope projection meant to improve on a 5% figure that someone pulled out of the air.

I think you're confused. _I_ am the one who said 5% and I think it's a reasonable working number. mikeash then replied, saying that Tesla will consume 22GWh (2018) and the battery market is only 35GWh/year (2013), implying that Tesla's share is so large it violates normal rules of market liquidity. Then I objected to that implication.

I don't think 5% is right. You seemed to think that 30% year-over-year growth in the industry was a reasonable estimate. If it was 35GWh/year in 2013 then that puts it at about 77GWh today and 130GWh in 2018. That puts Tesla at around 10% of total production now, and 23% in 2018.

I can't figure out which part of this you object to. You seem to accept all the numbers, but continue to argue about the shape of the box they're presented in.

I'll spot you 10%, but I think that's still within the realm of linear market behavior. I really don't agree with the original idea (now far upthread) that it required a visionary to lay the path for making all of Tesla's batteries. There are numerous firms around the world totally dedicated to market forecasts for major industries like batteries, car parts, and raw materials.

That's fine, I just objected to the "small fraction" bit, not the rest.

> I think you're confused.

How delightfully condescending of you. Yes, I didn't realize that the 5% number came from you, that hardly makes me "confused". Taking that into light, it's not obvious to me how a very rough back of the envelop projection is significantly more ridiculous than a number pulled out of thin air. My whole problem is that you called it ridiculous when it was clearly meant to be a rough order of magnitude projection. You can't really criticize a number pulled out of the air, because it goes without any attempt at justification. But a back of the envelop calculation can be improved in a piece-wise fashion to hone in on a better result. So any attempt to model a phenomenon (as long as it doesn't come with unjustified claims of precision and certainty) is an improvement over a number pulled out of the air. But that process is short circuited when people just dismiss the projection out of hand as you did. This is why I appealed for an improvement in my first comment. It's really easy to shit on someone else's effort, but it is completely unproductive. If you see a problem, then fix it. Then you're adding to the conversation, rather than just shitting on a little bit of effort someone put out there.

I wouldn't have taken the "I think you're confused" to be condescending. I mean I wouldn't interpret it as condescending and I don't think it was meant in that way.

Reply on news.ycombinator.com