Jump to content
Ensign_Cthulhu

Lies, Damned Lies and Statistics - an illustrative example.

6 comments in this topic

Recommended Posts

11,985
[ARGSY]
Members
20,067 posts
14,307 battles

The statement is attributed to various authors, but its usefulness is not in doubt. 

https://en.wikipedia.org/wiki/Lies,_damned_lies,_and_statistics 

 

It would be the cold hard truth to tell you that in the last month, my World of Warships views have risen by a whopping NINE HUNDRED AND NINETY NINE PERCENT.

Actually it's more than that, but 999% is the highest that YouTube is actually capable of reporting to me.

This is a level of viewership increase that is God Tier. Who could possibly hope to match it? 

 

Of course this grandiose claim immediately comes undone with the reveal of the actual raw data... 

 

image.png.a06a68de339164cf455b203f04676175.png 

...and the fact that I can still count my number of publicly posted vids without taking my socks off... but hey, I'm not actually lying, right? :Smile_trollface:

  • Funny 1
  • Meh 1

Share this post


Link to post
Share on other sites
2,252
[WOLF5]
Supertester
4,103 posts
4,084 battles

My dad used to say every time I did a household job well that he would double my salary....

Of course double of nothing is still nothing:Smile_teethhappy:

 

You can make numbers say whatever you want if you do enough math with them. As part of my job I have to analyze experimental results for the type of work we do this requires a lot of workup. While it's accepted that a lot of analysis needs to be applied to our data, we always have to check that we're not over analyzing things and skewing them to want we expect or want. I've been trained to look at the raw data and do my own analysis. If you just look at statistics you have no idea what's been done to make that pretty graph. The coronavirus is a great example. A bazillion models have been made, anywhere from the end of the world to don't worry about it. And they've all pretty much been wrong. But they're all based on pretty much the same data set. And whatever you want to believe there's a study to back you up.

So yeah, you can make statistics say whatever you want. That's why in game statistics are useful, but to judge a person solely on stats or even one stat is usually grossly incorrect.

Share this post


Link to post
Share on other sites
11,985
[ARGSY]
Members
20,067 posts
14,307 battles
4 minutes ago, AJTP89 said:

You can make numbers say whatever you want if you do enough math with them.

That's what scares me about "Statistics" in the sense of the mathematical subspecialty. Once you move away from means, standard deviations and some of the more basic statistical tests, I begin to lose confidence that what the mathematics is doing to the raw data is actually reliable. I'm actually far more trusting of intuitive deduction based on a clear presentation of the raw data. 

9 minutes ago, AJTP89 said:

That's why in game statistics are useful, but to judge a person solely on stats or even one stat is usually grossly incorrect.

Hell, yeah. I won't judge a long-term player who has a recent surge of Random battles in a Hosho, because we all know that some people came to carriers late (I know I did) and others do the reset-regrind thing. But if the Hosho numbers keep rising even as they accumulate battles in their Hakuryu, I start worrying for the baby seals and paying less attention to their overall winrate. 

Share this post


Link to post
Share on other sites
3,906
[WOLF9]
Privateers
13,767 posts
4,630 battles

You gots YouTube vids?

Share this post


Link to post
Share on other sites
11,985
[ARGSY]
Members
20,067 posts
14,307 battles
9 minutes ago, iDuckman said:

You gots YouTube vids?

I does now! Just this weekend I joined the secret cabal, bled the blood, made the sign, spoke the oath and took up the dark art. :fish_haloween: 

Like many things, including playing WOWS to start with, it's something I wish I'd taken up a lot earlier.

 

Share this post


Link to post
Share on other sites

  • Recently Browsing   0 members

    No registered users viewing this page.

×