Meta: I urge you to post this again in a few days. Very sad to see it never “got heat” and its reach was limited. I think this is one of the more valuable posts this year and wish more folks had seen it.
It may have started equal at founding and changed over subsequent rounds. I know of a case where 4 founders started equal and differed by order of magnitude by the end.
I had VIN #150,000 18 months ago, and they crossed 200,000 in May 2017 so it’s likely sitting at just shy of 300,000 right now. The 150k figure you are citing is US only (where there is a race to 200k when the federal credit expires).
Isn't this particular test susceptible to confounding effects? Traffic fluctuates day-to-day, week-to-week, and month-to-month, so how can you be sure it was the presence-or-absence of ads and not something else? If you randomize at the visitor level, you are sampling from both high-and-low traffic days, and control for any external fluctuations.
Because you're randomizing at the 2-day level, on average there will be just as many advertising/high-traffic days as advertising/low-traffic days, and as many no-advertising/high-traffic days as no-advertising/low-traffic days. The randomization is unaffected by traffic and uncorrelated with it. The unit of analysis is each day, not each visitor. This is why it has to be run for several months, otherwise you don't wind up with a decent n=50 pairs.
That's the tradeoff here: it lets you look at the totals, but it takes a lot longer than if you randomize per visitor in which case you could finish the test in a few days, often.
The stats went well over my head but as a web analyst I thought maybe you could have asked a simpler question. Pick a segment of your site such as visits from Google Search who landed on your homepage (most likely people who searched "gwern") which should reduce a lot of those spikes.
Subsetting will also increase the variance of each datapoint (consider the extreme case of picking a subset which was 0 or 1 visits per day), so is probably not a win. It's also hard to imagine what subset properly reflects all sources of traffic and so is informative about the total effect of advertising. Search queries definitely is not it.
As such, the x-axis is a metric about the population, not the individual. I happen to use them a fair bit as I find them easier to intuit density from than the alternative, which is to plot on a single line with transparency.
> The electric drivetrain, when coupled with a vast supercharging network...
I took that to mean that the car exclusively refueled on Superchargers, so the fuel cost was $0. That is a legitimate apples-to-apples comparison to the ICE vehicles.
I don't think that's apples-to-apples -- you have to go out of your way to always use a supercharger, so you'd have to account for the value of the time lost in traveling to a supercharger station compared to doing all the same journeys and using nearby gas stations. (Refueling at a supercharger is slower too, so that's extra fillup time.)
Alternatively you could use the electricity cost of recharging it at home, plus the cost of the occasional missed journeys that are out of range from not always being full. (Say, book their cost at what you'd pay for an Uber for those legs.)
(I don't know if charging a Tesla at home overnight is enough to get to 100% -- anyone know?)
This is a comparison of the vehicle for their business, which is a "city mobility service". I imagine any business like this is either going to have on-site fuel, or an agreement with a fuel depot with good rates, and drivers would stop by that location on the way home anyway, so this is probably comparable for this type of service use.
> I don't think that's apples-to-apples -- you have to go out of your way to always use a supercharger
For average person - yes, Tesloop, however, is Culver City - based, also known as home to the Culver City Supercharger, so there's a bit of a hometown advantage for them.
In fact, Tesla's recent penalty fee for hogging a supercharger spot long after the vehicle has finished charging is related to Tesla owners' complaints (on TMC, Twitter and Facebook groups) about Tesloop vehicles parked overnight in Culver City supercharger spots.
The other thing is this is just a data point for one car. If the results were for a dozen cars or more it would be useful. Imagine if Backblaze gave failure rate based on 1 HDD.
It is very misleading - I would call it deceptive - for the article to say "total combined maintenance and fuel costs of the Tesla Model S..." [my emphasis] when the fuel costs are covered by Tesla, especially as this is a loss-leader that you can no longer take advantage of.
Not sure if it's related to load, but whenever I try it on a long-tail site (i.e. not one you would have precached) it comes back with information about google.com instead of the domain I'm on.
Very strange. It is working fine on our side: https://cl.ly/1x150t1m0922
I'll check this with our developer and get on it right away. Thank you so much for taking time to report this!
