Statistics and a more correct way of percentiles

Posted by
September 16th, 2013 11:54 pm

I pythoned again to create a full spreadsheet of LD27 games, with real percentiles*. :)

Links:

Compo (CSV) – 1210 Rated Games
Jam (CSV) – 620 Rated Games

Note that all the coolness data was taken directly from the site, and not processed.

* When you divide your position by 1437 you include 227 games that haven’t received any ratings at all, giving you a brighter look at reality. Moreover near the mean score there are quite a lot of ties, the position you have on your page is the top of the tie range (should be middle for calculating percentiles) which can brighten your day even further.

 

Might make some graphs later of number of ratings vs. score (other suggestions are more than welcome).

 


11 Responses to “Statistics and a more correct way of percentiles”

  1. Larzan says:

    How did you get the data from the site, did you parse it yourself? I am asking because i couldn’t find my entry in the docs and would like to see the original file

  2. Gabriel says:

    Dude, this is seriously awesome! Congrats and thanks for posting it!

    As a funny sidenote, your spreadsheet had quite the opposite effect on me than it was designed to do: it actually brightened my day! hahaha

    I think we tend to look at the scores and ranking without a context of how much Ludum Dare has grown. I mean, if someone positioned at 10th 3 editions ago and now positions in 20th, it is actually the same level of performance, right?

    BTW, is there any easy way to apply this same calculation for other editions? I was wondering if I could check my games one against the other looking at percentages instead of absolute numbers…

  3. vrld says:

    How did you get the data? And more importantly: Is there a way to get variances / ‘error bars’ on the data?

Leave a Reply

You must be logged in to post a comment.

[cache: storing page]