Блог пользователя RP-1

Автор RP-1, история, 31 час назад, По-английски
Disclaimer

Looking at the top-10 users and Recent Actions bar right now, two things become apparent:

  1. Since LGMs cannot change their rank to a higher one, more people in, say, top-100 have changed their rank to a lower one than to a higher one.
  2. Codeforces-wide, more people have changed their rank to a higher one than a lower one.

So there has to be an exact cut-off point, above which the same number of people have changed their rank to either side. I decided to do some scraping to find out exactly where it is.

Lo and behold:

Rank changes (leaderboard position)

Of the top 15k users, about tenth changed their rank to a higher one, and the same number to a lower one.

Now, I'm no physicist, but I do know how to draw a line. I honestly have no explanation as to why the blue graph is so perfectly linear. If anyone can explain this kind of behavior, please leave a comment below. ;)

Now let's scale the x-axis by rating:

Rank changes (rating)

Looks like people who have almost reached a higher rank are less likely to change it. Also, I think that the orange curve becomes linear over time? I don't know. I'm certainly in no position to try and do some advanced statistics on this data, so I won't.

Rank changes difference (rating)

What I find curious is that Experts are almost evenly split into two groups. The difference almost doesn't change until around 1800 and then plummets.

The exact cut-off point in question, according to my calculations, is 1579. I really wanted to shout out a user with this rating, but there's quite a few, so congrats to all of you!

Tell you what, have some more graphs.

Most popular rank changes

Rank changes by target rank

Keep in mind that I only fetched 40k users, so the histogram might look a little different in reality, but the trends are clear.

Cyan is easily the most popular before 2000 (looking at you, Um_nik), but then nutellas and grays skyrocket, and after 1500 LGMs take over. Purples and oranges are tied for third, but I hope you'll agree that they're the prettiest of all.

The code I used is available here.

Anyway, thanks for reading, and I wish you all the best for the New Year!

  • Проголосовать: нравится
  • +325
  • Проголосовать: не нравится

»
31 час назад, # |
  Проголосовать: нравится +16 Проголосовать: не нравится

Upvoted

»
30 часов назад, # |
  Проголосовать: нравится +30 Проголосовать: не нравится

Too many words in blog but pictures are cool!