mukel's blog

About HackerRank rating system
By mukel, 11 years ago, In English

Because HR doesn't have a public forum/blog like Codeforces I post this here, anyway the community is shared between both sites.

The issue is with rating calculation on HR. I've participated on 2 contests recently; in the Functional Programming Contest I've got full score 8th place overall and in the Weekly contests full score again 4th place overall, which I consider are quite good results, but my "rating" just increased a few points, and to my surprise I went down in the global Leaderboard, from 163 to 168, how is this possible? I tried to find some information in the Scoring page, but there is no clue about the "weight" of different kinds of contests, eg Functional, 101, 20/20, Weekly .... In Codeforces for a mere 107 place I got +88, which doesn't mean that is better or worse, on Codeforces I remember once that the rating system was changed when tourist got 1st and his rating decreased, indeed there was something wrong with the rating calculation (someone correct me if I'm wrong), TopCoder implemented a quite sophisticated rating system... Could someone explain how the global rating works on HR?

UPDATE: It seems like the global Leaderboard becomes crazy, users are placed multiple times with different ratings.

  • Vote: I like it
  • +27
  • Vote: I do not like it

»
11 years ago, # |
  Vote: I like it +19 Vote: I do not like it

It seems like my rating did not change after weekly contest at all.

→ Reply
  • »
    »
    11 years ago, # ^ |
      Vote: I like it 0 Vote: I do not like it

    same here. my guess is that he have to wait for a few more days.

    → Reply
»
11 years ago, # |
  Vote: I like it +10 Vote: I do not like it

And non standrad color rank:(

→ Reply
»
11 years ago, # |
  Vote: I like it +5 Vote: I do not like it

Seems like weekly challenges are not counted into rating yet

→ Reply
»
11 years ago, # |
  Vote: I like it +5 Vote: I do not like it

I also don't understand how the rating system (global leaderboard) works on Hackerrank. Anyway, the idea of mixing together ratings for different types of contests (which require different types of skills) doesn't seem like a great approach to me (e.g. "long" contests, "short" contests, functional programming contests, "real data" contests, etc.). I participated mainly in the 20-20 hacks (i.e. "long" contests) and I usually got one of the top places, yet my rating increased by small amounts after the first few such contests. My rating increased more when I got one of the top places in Codesprint (another type of contest), which made me believe that if I want my rating to increase more then I should participate in more types of contests. For instance, I've never participated in 101 hacks (Hackerrank "short" contests), but my guess would be that, unless I do something really bad in my first such contest, my rating will increase more than by getting a top spot in a "long" contest (because for the 101 hacks I am currently "unrated"). I don't know if my guess is correct, but I hope that I will have time to test it soon :) My hypothesis could also be tested sooner in case they decide to rate the first weekly challenge contest (which I'm not sure will be the case).

→ Reply
»
11 years ago, # |
  Vote: I like it 0 Vote: I do not like it

I think HackerRank recalculates your rating for each rated event even if you don't participate in this event

→ Reply
  • »
    »
    11 years ago, # ^ |
      Vote: I like it 0 Vote: I do not like it

    Nah, it's only for the contests in which you made at least one submission. My rating never went down for contests in which I didn't make submissions, even though I generally registered for every contest in the past.

    → Reply
»
11 years ago, # |
  Vote: I like it 0 Vote: I do not like it

What's also funny is that in the rules there is no mention about the method in which ties are resolved. For example in the last Infinitum Math contest there were at least 20 full scores, all ranked first place. But only the top-10 of those got a T-shirt (me included, luckily). One of the problem setters later explained to me that they used an ICPC-style penalty system, but there's no mention of that anywhere and the penalties are also shown nowhere.

For the new Infinitum they used the time of the last submission as the penalty, which makes more sense but it's still not documented.

→ Reply
»
11 years ago, # |
  Vote: I like it 0 Vote: I do not like it

After double check the contest history of tourist, I can surely tell you there's no contest of codeforces where tourist got 1st place and his rating decreased!

→ Reply
  • »
    »
    11 years ago, # ^ |
    ← Rev. 3   Vote: I like it 0 Vote: I do not like it

    Codeforces probably updated the rating calculation from that round onwards.

    P.S. there may not be such a round as u said, but there are two rounds where tourist finished second and yet his rating didn't increase! :D

    → Reply
    • »
      »
      »
      11 years ago, # ^ |
        Vote: I like it 0 Vote: I do not like it

      I think it's reasonable that his rating decreased, because he is definitely the best. No reason to blame the rating system. :D

      → Reply
      • »
        »
        »
        »
        11 years ago, # ^ |
        ← Rev. 2   Vote: I like it 0 Vote: I do not like it

        i never blamed the rating system!
        i was just saying that tourist is so good that sometimes even finishing second is not enough for his rating to increase! whereas for ordinary users like me (hopefully i'll also become red soon :P) get rating increased for finishing ~300!! :)

        → Reply
  • »
    »
    11 years ago, # ^ |
      Vote: I like it +8 Vote: I do not like it

    It happened indeed, take a look to Rating Formulas Changes

    → Reply
    • »
      »
      »
      11 years ago, # ^ |
        Vote: I like it 0 Vote: I do not like it

      You are right! My bad. Three years ago I can't solve the A+B problem not to mention codeforces. I didn't know that happened. Sorry.

      → Reply
»
11 years ago, # |
  Vote: I like it +8 Vote: I do not like it

Hi mukel & everyone, thanks for writing in. This is one of our top priority items for this week. We're fixing it this week and everyone's rating will work fine.

How are we going to fix? There are two types of leaderboards for every domain (AI, math, ML, FP, etc.) — classic & contest; The contest leaderboard = function (performance of contests) — similar to elo rating; Classic leaderboard = cumulative points of problems solved.

Would love to hear your feedback.

coming soon: *A blog feature *We hired a full-time person to take care of content (challenge problems, rules, etc.) *Secret network feature (we are excited!)

→ Reply
»
8 years ago, # |
← Rev. 2   Vote: I like it +14 Vote: I do not like it

Hi, not sure if this is still relevant but my friend kobortor and I were confused by the rating changes for the recent Week of Code. Even though my initial rating was less than his and I placed higher than him, my rating change was lower. My profile.

→ Reply
  • »
    »
    8 years ago, # ^ |
      Vote: I like it +5 Vote: I do not like it

    They said me, that they are going to change rating system in the nearest future.

    → Reply
    • »
      »
      »
      8 years ago, # ^ |
        Vote: I like it 0 Vote: I do not like it

      Will they recalculate all the previous contests using new rating system?

      → Reply
      • »
        »
        »
        »
        8 years ago, # ^ |
          Vote: I like it 0 Vote: I do not like it

        I don't know, but I suppose, that they are not going to recalculate them.

        → Reply