This YouTube Channel Achieves Fractal Wrongness in Data Misuse

This one. The channel of “GOOD Magazine”. I take it it’s one of those ironically named things, like how Fox News’ Slogan is “Fair & Balanced”, because the bad data is strong with this one.

Let’s take a look at one of their YouTube videos titled “Top 10 Most Powerful Hurricanes Ever“.

Let’s start at 0 seconds in. Do you see a problem?

I see a problem. The title of the video is “Top 10 Most Powerful Hurricanes Ever”. What does that mean? What is their metric for “Most Powerful”? Hmmm, maybe we’ll find out after watching this video for a while. Let’s see…

Okay, 15 seconds in, the text on screen tells us that they’re clarifying that “most powerful hurricanes” means “biggest hurricanes”.

…really? Because I think for most people “biggest” defaults to meaning “biggest by area”, and I’m not quite convinced at all how big a hurricane is tells anything about their power. (Watching the video a bit, one can see that in fact this is not the case and the video uses “biggest” in a more metaphorical sense.)

All right, let’s roll their top 10 through.

Done watching it? How many things bothered you? If your answer is zero, you should go and rewatch it again, until you find something fishy. No really, there are things in this top 10 that are fishy that do not require knowledge about hurricanes to detect.

Let’s start with realizing what their metric of hurricane power is: it’s the smallness of the central pressure. But even they decide to put this metric second in their data display, so certainly it should also be nonobvious to them that this is a nontrivial clarification to make?

But second of all, how reliable is their data on central pressure (which, incidentally, appears like it should have been labeled minimal central pressure—the central pressure of a hurricane changes over time)? Notice that they report one hurricane (the “Labor Day” Hurricane) from Gregorian calendar 1935, long before the other hurricanes mentioned. What was the quality of meteorological instruments back then?

It turns out the “Labor Day” Hurricane occurs significantly before accurate hurricane data became available, and “892 mb” is merely the hurricane’s pressure at landfall. As hurricanes tend to weaken around landfall, the storm was likely more powerful (and its central pressure a lower reading) at some point prior at sea. (Alternatively, since they didn’t explicitly say “minimal” on the data label, one could suggest that maybe they are reporting landfall central pressure rather than minimal central pressure, but in that case the landfall central pressures of other hurricanes are incorrect, so this is a theory of more discrepancy.) This fact was ignored simply because data was unavailable because of the state of meteorological technology eighty years ago, and the data point that was the strongest data available was presented as the strongest data. Other hurricanes from way-back-then were strong enough to potentially be close contenders for this list, but their data is unavailable, so their possibility is silently discarded.

They also present wind speeds. Like how they left out the “minimal” before “central pressure”, they also left out the “maximal” before “wind speed”. But also notice that all the wind speeds presented are divisible by 5 except for Katrina’s. Did it just happen that among the top 10 only Katrina’s maximum wind speed was not divisible by 5?

It turns out that hurricane wind speeds are typically reported to the nearest 5 mph in the US. I have no idea where they got their 174 figure for Katrina from, but it’s probably not from the same source as their figures for the other hurricanes on the list.

Speaking about mph and the US, isn’t it odd that all the strongest hurricanes decided to pick on North America? Do hurricanes have a thing against imperial units? (If this helps the US get off the imperial system, I don’t think I actually mind that much.)

It turns out they get away with a technicality here, because “hurricane” is a term that’s local to North America. There are more powerful cyclonic storms than the ones presented in the Western Pacific, but there they are not called “hurricanes”; they’re called “typhoons”. Still, this is a technicality against intuitive interpretation, and storms of the same meteorological nature that just happened to be called different names should not be discounted in the creation of such a list.

Okay, so circles appear on a map above this presented quantitative data. It looks like they mean to present the path of travel that each of the hurricanes took. But first of all, the paths are wrong. They do generally have the shape that the actual paths of the hurricanes had, but the paths are quite shifted from the actual paths the respective hurricanes took.

But also, one would suppose that the circles are equidistant in time. (They’re certainly not equidistant in space—note Gilbert’s path versus the paths of the others, for instance.) That is, one would think the amount of real-life time it took for the hurricane to move from one plotted location to the next location is the same as the amount of real-life time between the hurricane being at that second location and being at the one after that. But three days pass between the second and third circles shown for Hurricane Wilma whereas only six hours separate consecutive circles at the end of the path presented for Wilma (that is, if we assume that the presented path is just a position-shift from the correct path, and not something crazier).

And that brings us to the next questionable piece of data, the dates shown for each hurricane. One date is presented for each, but all of these hurricanes lasted longer than one day. What’s the displayed date?

Is it the date of landfall? This seems like the most reasonable of dates to use as a “date of a hurricane”, but several of these hurricanes had multiple landfalls. And also the date presented for Hurricane Wilma does not correspond to any of its landfalls…or for several other hurricanes. It’s not date of greatest power by any of various reasonable metrics of power, either: the date given for Rita is off by one day, for instance. Now one day isn’t much, but the dates given for Katrina and Allen are off so much they’re dates during which the storms haven’t even strengthened enough to achieve hurricane status yet.

So, uh, points for spelling the name of each storm correctly, because that seems to be the only category of data presented that is not shoddy, including the name of the presentation itself. This video is Category 5 bullshit.

But actually, so is the channel in general.

In the video “Mass Shootings in the US During 2015”, why was the Waco shootout omitted from the May map? If the channel considers 9 deaths too small to be considered a mass shooting, why does the Colorado Springs Planned Parenthood shooting, with 3 deaths, make the list? (Or is there another mass shooting with more than 9 deaths that happened in the area that the news never knew about?) Also, the fact that the video shows only 16 days for December confirms that the video was in fact made before the year in question ended, but it does not explain why October had only 30 days, nor does it explain whether even if October had only 30 days why the sum of the number of days in the months presented is greater than the number of days in the year it claims to report on (31+28+31+30+31+30+31+31+30+30+30+16=349>348).

And in the video “Vaccine-Preventable Disease Outbreaks Around the World”, at what resolution are the bubbles presented in. There seems to be many bubbles for the US, suggesting city-level resolution, but if the data is presented at the city-level, do nearly all of the outbreaks in China happen to occur in some city in central China that is probably not one of its largest? And what’s with the number of copies of the world presented? It’s clearly more than one, but it’s not quite two either: notice most places are shown twice but Europe isn’t.

But here’s the saddest thing. The videos of this channel seem to do verzero downvotesy well in votes. That first video I mentioned, the hurricane video? As of now, it has zero downvotes, despite 23 upvotes. You know how YouTube is a place where you expect anything about a video that’s slightly suspect to get angrily shouted at in the comments? This video has none. It goes to provide evidence that the majority of humans out there are completely credulous as to unacceptable practices in data, how unreliable the foundations of most people’s beliefs are, and how easily people can be fooled by the malicious.