Codechef Rating System

Can anybody explain me how exactly the rating of any user is calculated ?

I did’t get it from this link also

Please Explain us with some example.
I will really appreciate your effort.

and Please explain me some think about ELO seed based rating system.

I really want to know about it.


Admin EDIT: We have introduced a new rating system. Please have a look at this.

1 Like

What exactly you want to understand?

There is simple rule: more problems solved in contest -> higher rating (your rating will grow with solved problems in a rated contests)

You can also try to read about Elo rating at wikipedia.

@theshubhamgoel ELO rating is relative rating. If an opponent has a high rating then he is expected to solve more number of questions, basically get a good rank.In the case of kgcstar, he was ranked number 3 in the world previously. So because to his high rating he should have gotten a good rank this time. However his rank must have been around 100 which was less than expected. Hence his rating reduced.

If a person with low rating gets a good rank then his rating will increase.

Is this rule apply for both long and short contest ???

1 Like

No yet told me about ELO seed based rating system.

Please explain me some think about ELO seed based rating system.

1 Like

Codechef’s rating system is different from other programming competitions (Topcoder, Codeforces) because it also includes an additive term equal (more or less) to (the total number of competitors in the contest / your rank). Let’s consider @kgcstar’s case which was mentioned. Normally, because of his low rank in this contest (compared with his overall rank according to rating) his rating should have decreased more than it did (for instance, this happens in TopCoder when a top 10 contestant gets something like place 60 or 70). But he also got a term which was equal to approximately 4000 / 88 (there were about 4000 participants in the contest and 88 was @kgcstar’s rank in the contest), so his rating decrease was very low. Let’s consider @ACRush now, who won the contest and was also the best rated coder. The fact that he won the contest was to be expected according to rating, so his rating shouldn’t have necessarily increased by very much. But, because he got rank 1, his rating also increased by about (4000 / 1) (and you can check to see that his rating indeed increased by around 4000). You can also see that the top ratings for short contests are lower than top ratings for long contests - this is because fewer people participate in the short contests (cook-offs) and, thus, the ratings of the top contestants do not increase as much as in the case of long contests.


@theshubhamgoel Like i told you ELO is a very complex rating system and very few people can explain it in the sort of detail you want. If you are not satisified with the explanation of @mugurelionut ,who is one of the most qualified persons on this forum ,then i believe no one here is qualified enough to answer your queries.

I would suggest you try to google it and look from explanation from other sources.
Here are some links i found interesting : 1 , 2 ,3 , 4 .I would suggest you try to google it and look from explanation from other sources.

Renewing this thread again and again and asking the same question will not help in any way. I am sorry we could not answer your doubt despite our efforts. Best of luck.

1 Like

Well that additive term is added and then averaged with all the values? is that how ratings are calculated??

But how can someone get negative rating if he solve 7 problem in jan14 long contest
see this users profile

1 Like

I see… The idea of the rating is that is describes your experience. And rating changes like (simple version): contestants are ordered by their rating decreasing and this expected order is compared to real order after the contest. If your standing is better then expected your rating grows if worse it drops.

Thanks for explanation.

Can you explain me how to calculate expected rating and what is the expected ranking in my case

1 Like

Yes, but there are two independent ratings…

Think about it in relative terms… If you solved 10 problems in a month and on the next month you only solve 7, then your relative performance dropped… So your rating will also drop…

Not exactly, but basically correct.

If the best coders solve only 7 out of 10 (not likely to happen), then it should be same as solving 10 in previous contest…

Is there any effect of solving practice problem on expected ranking

1 Like

Believe me, you do not want to know that :smiley: Besides that formulas are not known here (AFAIK), see formulas for TopCoder to imagine the complexity -

1 Like

Not directly, but more problems you solve (here or at another contest page) it is more likely you solve more in contest too…

Additional article for TopCoder rating -

Ok but explain me some think about ELO seed based rating system.
I really want to know about it

1 Like

ELO is commonly used in chess. You might want to google it as we are not experts and have just a basic understanding of it.