Блог пользователя pani.natalia

Is it true that getting red can be easier in certain periods?
Автор pani.natalia, 3 года назад, По-английски

The current color scheme was introduced in October 2015 and updated in May 2018. One of the reasons was that getting red had become “much easier”. Is it true that at some period it could be easier? Certainly it is not my practical concern, but I still wonder ) So how can we measure this ease to get red?
It looks like it will be several iterations but for base line I would like to know what percent of CF users are red. I ran the script and found number of active users of every colors and every title for the end of 2018, 2019, 2020 and 2021. I mean number of users who were active in the corresponding period. Feel free to download .csv files and play with them, but it is really convenient to do it in Tableau – this is the dashboard with selective options. Its left part represents percent distribution by colors, the right one is by titles. You can select options for left and right charts independently and for example discover that you had to be in top 0,67% — 0,65% — 0,69% — 0,67% to be red and in top 6,61% — 6,31%-5,91%-8,47% of reds to be legendary grandmaster. Is it getting easier to become a legend? )

  • Проголосовать: нравится
  • +92
  • Проголосовать: не нравится

»
3 года назад, # |
  Проголосовать: нравится +35 Проголосовать: не нравится

Nice visualization! It seems like the requirements to reach the top (masters and above) seem to have remained relatively constant over the years. What's more interesting to me is the sudden explosion of newbie accounts from 2018 to 2019. Presumably this was due to the change in rating calculation for new accounts (not sure if there might be another reason), but that just goes to show how many of the so-called "active accounts" probably participated and were abandoned after a few bad contests. On the other hand, this could be seen as an affirmation that the new system worked well in revealing the actual distribution of rating in Codeforces.

→ Ответить
»
3 года назад, # |
← Rev. 2   Проголосовать: нравится +6 Проголосовать: не нравится

Difficulty can't be measured by the percentage of reds, since today many users open accounts in order to train for job interviews (and reaching GM requires much more effort I assume). I don't know if it's easier nowadays to become GM as I started CP only a year ago, but a better way to measure might be to convince a large group of high rated users to participate in an old contest and compare their performance.

→ Ответить
  • »
    »
    3 года назад, # ^ |
    ← Rev. 3   Проголосовать: нравится +13 Проголосовать: не нравится

    Imo it isn't an accurate measure either. The bar is always increasing. Many old hard Div1 problems would be too well known these days. At least their ratings would be less compared to what there are now if one uses them in contests these days. I haven't done a lot of old cf VC (< 20 ig) but can say it with confidence after doing a lot of old ARC VCs and looking at the rating of those problems. I had the same feeling when I did those odd 10-20 old cf round VCs.

    Spoiler
    → Ответить
    • »
      »
      »
      3 года назад, # ^ |
        Проголосовать: нравится 0 Проголосовать: не нравится

      If a problem requires some certain technique that has become popular since (like the DP optimization trick aka Aliens trick) then your statement is correct. However, I think the majority of problems still require the same problem solving skill. I actually have a conjecture that older problem ratings are underestimated because the contestant ratings were much lower back then.

      → Ответить
      • »
        »
        »
        »
        3 года назад, # ^ |
          Проголосовать: нравится -7 Проголосовать: не нравится

        Usually, when I solve old problems with rating X, they feel much easier than modern problems with rating X

        → Ответить
  • »
    »
    3 года назад, # ^ |
      Проголосовать: нравится 0 Проголосовать: не нравится

    Doing cp for job interview is not a new thing afaik. I know people who did cp in early 2010's for getting job interview from bigtech. I remember reading a basic programming language book long time ago where the author said he started cp in university to get interview from bigtech. He started university in 1999.

    → Ответить
»
3 года назад, # |
← Rev. 2   Проголосовать: нравится 0 Проголосовать: не нравится

It seems strange that I am already in top 20% in all codeforces accounts, but still just specialist.

→ Ответить
  • »
    »
    3 года назад, # ^ |
      Проголосовать: нравится 0 Проголосовать: не нравится

    You are not "stuck" yet. You are on an upward trajectory right now. A single contest rating delta doesn't move you directly to the rating of your real performance, but only adjusts it closer. For example, if you have a 1400 rating and performed similar to 1600-1700 rated people in a contest, then your rating will adjust by maybe something like only up to +100 points in one go. Multiple contests are required for your rating to converge. You will be stuck after your rating starts going up/down in a certain range.

    → Ответить