Блог пользователя kostka

Автор kostka, 7 лет назад, По-английски

We probably all noticed that the rating inflation still occurs [citation needed].

In several games, there's a thing called "rating decay", when your rating decreases when you don't play ranked games (in our world: if you don't participate in contests). Please compare:

What do you think about implementing such a solution to the Codeforces rating? It should slowly drain some of the unnecessary rating (especially from people who do not participate anymore). Similar to many sources above, I think it should affect only highly rated users (maybe with 2100+ rating?). It can also serve as an encouragement to participate in contests more often.

  • Проголосовать: нравится
  • +43
  • Проголосовать: не нравится

»
7 лет назад, скрыть # |
 
Проголосовать: нравится +65 Проголосовать: не нравится

It might serve as an encouragement to participate more, but

  1. I am assuming one can play a game at any time, but for contests you can't just sit at your preferred time.

  2. If rating decays below a certain level then, there might be many coders who are actually at a high level and they will dominate a div2 contests since they did not / could not participate in previous contests.

Just my thoughts though

»
7 лет назад, скрыть # |
 
Проголосовать: нравится +27 Проголосовать: не нравится

rating is just a number. who cares!?

»
7 лет назад, скрыть # |
 
Проголосовать: нравится +37 Проголосовать: не нравится

It should slowly drain some of the unnecessary rating (especially from people who do not participate anymore).

If they don't participate, then they are not causing rating inflation. So it doesn't matter in this case.

encouragement to participate in contests more often

This does the opposite — it causes rating inflation.

I see no reason why this should be considered to help at all. Even in your final starcraft2 link they removed rating decay 4 years ago.

And a big problem in these sort of games is some dude who gets deranked from rating decay basically ends up getting a couple of smurf matches and dominating everyone else (aka: 10 reds coming back to get their rating again and crushing a bunch of div2 people). You know how there are always a ton of unranked shady guys winning div2 contests? It will be like that, but with more people.

So basically, nah.

  • »
    »
    7 лет назад, скрыть # ^ |
     
    Проголосовать: нравится +3 Проголосовать: не нравится
    This does the opposite — it causes rating inflation.

    Could you please elaborate? I never looked into it closely, so the rating system of CF isn't intuitive enough for me to be confident about this claim being correct (or about it being wrong).

    I used to think that the main source of inflation is new accounts that "give" rating to the system and get abandoned with a rating lower than their initial rating. Is this idea wrong? Or is there a different key reason? Sum of rating changes for arbitrary CF round is negative, so it sounds like by participating there you are "expected" to decrease your rating (though it would be wrong to assume that things are this simple: it is not clear how the fact of your participation affects overall sum of rating changes in the round).

»
7 лет назад, скрыть # |
 
Проголосовать: нравится -24 Проголосовать: не нравится

Plot twist : You are paid by Heroes of the Storm . :D

»
7 лет назад, скрыть # |
 
Проголосовать: нравится +38 Проголосовать: не нравится

Next blog be like: Let's hide rating numbers from people and show color only (CS:GO), and devs do what ever they want.

(Actually calibration (In CS:GO you play 10 matches before your rank gets defined) makes sense, most of one-time accounts will be eliminated)

»
7 лет назад, скрыть # |
 
Проголосовать: нравится 0 Проголосовать: не нравится

I fully agree your opinion. Like other video games, if you don't do competitive programming for a long time, then your cp-optimized observation skill is going down.

»
7 лет назад, скрыть # |
 
Проголосовать: нравится +3 Проголосовать: не нравится
(especially from people who do not participate anymore)

But why would we care about them?

They are effectively not a part of the rating ecosystem if they don't participate; they don't affect rating re-distribution in any way.