I've been getting back into Magic the Gathering and listening to podcasts and it is wild how sophisticated the pro Magic community has gotten about data.

I'm listening to them talk what a particular measure "really" means and why it might or might not be valid and how they're using qualitative data to evaluate its validity and it's like, damn, ya'll are better at this than most engineering managers I know.

#17lands

They really deeply grasp that that measures are measuring something very specific ("the number of matches won by anyone who had a particular card in their deck and was reporting their data") that isn't necessarily the thing they care about ("will it increase my chance of winning if I put this card in my deck")

And while it seems obvious when I put it like that I can probably count on one hand the number of times that I've seen numbers used in a business context with that level of sophistication.

Usually people get stuck at "this number literally means what it's labeled on the chart" and just don't think about how the way the number is measured affects it unless they're specifically asked.

And then even if they *do* think about it they completely forget about it as soon as that conversation is over.

Anyway if you're hiring for data analysts you could do worse than get in touch with 17 Lands and see if you can put a job ad on their homepage.

https://www.17lands.com/

@nat Totally agreed. I've seen very little in a business context as careful with numbers as the x-wing dice study: https://hackaday.com/2019/06/20/automated-dice-tester-uses-machine-vision-to-ensure-a-fair-game/

@williampietri You can see here exactly why this might be, too.

You can generate data very quickly, the event you're interested in is relatively short and has a very unambiguous win condition you can optimize around (winning!), and even when there are a lot of decisions that feed into it it's easy to model how those decisions/input feed into the result you're analyzing.

@nat This article is my go-to when I need to explain this to people: https://ferd.ca/plato-s-dashboards.html