Sjors Provoost @sjors@m.sprovoost.nl

@michaelfolkson @stephanlivera for your reading pleasure 🙂

I guess @birdsite would have been better than a link.

It uses https://gitlab.com/ayush-sharma/tweet-toot by @ayushsharma22 and a simple cron job every minute.

If you use PyEnv then you can make a bash script like so, and have the cron job call it:

#! /bin/bash
cd tweet-toot
source venv/bin/activate
cd tweet-toot
python run.py

You can have cron log the output:
* * * * * ~/toot.sh >> ~/bird_toot.log 2>&1

I added a separate account to relay my bird site remarks. Just follow it here: https://m.sprovoost.nl/web/accounts/3875

My favourite feature is an RSS feed per municipality . It lets you know when to return your bunker 🙂

Or in my case: when to come out of it. A green two-week streak is my personal threshold for going to indoor settings like a supermarket or hair dresser. That's not happening any time soon in my area.

I occasionally nag my municipality about the latest numbers, but they don't offer any explanation. Hopefully that changes if enough people start asking questions based on this data.

I'm glad the health authorities are publishing some useful data, but availability and format are terrible. I ended up using a GeoServices REST endpoint provided by Esri. They in turn download and process CSV (!) files from government. There's gaps in the early data. The daily numbers are historical totals, which I then subtract to figure out the daily number of positive test results. Negative numbers mean a correction, but historical totals can also change. Anyway, it seems to work now.

Allow me to plug https://pletdecurve.nl here as well. It displays a "fire chart", as Yaneer Bar-Yam from endcoronavirus.org calls them, for all Dutch provinces. You can drill down on a province to see how its municipalities are doing.

Built with Ruby on Rails, initially hosted on Heroku, but I moved it to my own Ubuntu server, which was easier than I expected.

It's open source, and should be easy to adapt to other countries; happy to accept PR's to make maintaining a fork easier.

I have a one month gap in my timeline. Looks like some daemon got ill and recovered once I upgraded Mastodon.

Dutch PM just announced that he embraces the bat shit crazy plan to create herd immunity by controlled infection of the population. Because the "experts" say so.

Fortunately a combination of contact tracing and genetic forensics mean we can keep track of how many infection he caused down the pipeline, and hopefully hold him accountable.

Meanwhile Draper is still actively putting lives at risk, both through his direct actions and insane policy recommendations. But clearly getting people killed is fine by Twitter policy. As long as there are no guns involved.

I visited the bird site Kangaroo court. You get exactly 200 characters to appeal their flagging, and explain why they're idiots. Then you get an email without explanation that they deny your appeal.

Essentially they just want you to delete the tweet. There's probably no human involved in the process. No replies means you can't do a Turing test.

Weirdly enough you can just login and fill out the appeal form again. So I guess I'll just keep exchanging insults with a robot :-)

Which brings us to the S&P chart. If it was Gaussian, it should stay under the blue line. It's not even close, so it must a scarier distribution. But what's in the chart?

Well we can plough through 3 pages of Taleb's math, or check Wikipedia for (non nuclear war) MAD: "In other words, for a normal distribution, mean absolute deviation is about 0.8 times the standard deviation"

That's where the sqrt(π/2) blue line comes from. QED-ish; not a normal distribution.

https://en.wikipedia.org/wiki/Average_absolute_deviation#Mean_absolute_deviation_around_a_central_point

First thing to understand is that there's hierarchy of statistical distributions, our high school friend Gaussian being near the friendly bottom.

Unless you understand the underlying process, like in some areas of physics, you can measure all you like, but you can't prove something is in the lower left. You can however *disprove* it, often with a single observation (after you check your instruments). Falsification at work.

As you disprove friendly distributions like Gaussian, you move up.

Why the S&P 500 is not Gaussian

I'm trying to keep up with Taleb's new technical book, but I have neither his background in trading nor mathematical rigor. In a previous life I was a physicist; they are notorious for skipping math steps based on intuition, expecting mathematicians will prove their intuition correct.

I may do some out loud thinking here, starting with a TL&DR of this chart.

The book Statistical Consequences of Fat Tails is here: https://www.academia.edu/37221402/STATISTICAL_CONSEQUENCES_OF_FAT_TAILS_TECHNICAL_INCERTO_COLLECTION_