After the big “dropping of ratings due to plagiarism” incident, it seems that Codechef cannot calculate the ratings of people accurately now. Why? Glad you asked. Here a few rating changes of people:
Rank Username Initial Rating New Rating Change
19 vaibhavvg 1546 1741 +195
159 sid_008 1546 1742 +196
21 notira 1500 1775 +275
130 reaper_mist 1500 1749 +249
So, firstly, if 2 people have the same initial rating, the person having a shittier rank in the contest gets a better final rating(1546==1546, 1741 < 1742, implies, rank 19 gets a lesser boost as compared to rank 159, given same initial rating).This means, I just wasted my time solving the 3rd problem and should’ve just let my rank slide to the bottom, because sid_008 solved just 2.
Secondly, @vijju123 was saying something about the codechef system being unique about its rating in a previous post. I also agree. Only on codechef can one perform worse and still get a better rating than his peers. Check out notira(rank 21). He had a rating of 1500 initially. After contest, he is now at 1775. I think we all can spot a basic logical error here.
Let there be 2 people A and B having ratings r1 and r2 respectively. Given r1 > r2 initially. A performs better than B in a contest. So, obviously A should have a better final rating than B (r1’ > r2’) . I agree that sometimes B can get a better boost(+ delta) than A here. But the final rating of B should never exceed A’s here.
This is precisely what has happened here. Rank 21, rank 130 and many more.
So one way I see is for codechef to rectify the error here, or correct me if I’m wrong in saying that a newer account should always be the alternative when your rating decreases.
I suspect this effect has to do with volatility: new members have higher volatility, so their rating change (both up and down) is larger than for old members. https://www.codechef.com/ratings
The maximum rating change , in your case before the cook-off was capped by,
100+75/23+500*100/(46+500)=194.83 which corresponds to the rating change you actually got.
Similar situation happened with me, I had rank 13 in May cook off but got only +170, with initial rating of just 1260.
To me, this seems a bit unfair, my ratings should have gone up much faster.
Codechef, please do see if you can do something about this, as the volatility factor is helping people who score low(so the capping doesn’t come into picture) but for people with very good ranks, it is in no way helping us, the max cap rating rule is overriding everything else.
But I think that resolves your question.
This is common in most of the rating systems due to high volatility of new participants. Not only here, but also in other games, for example, in Chess, because the system is trying it’s best to predict your actual rating. But it goes either way. If you perform poor than your current rating, then your reduction will also be huge as compared to the one who has been participating for a long time. So here relative ranking comparison does not play a direct role but most importantly how well you are performing w.r.t your current rating and for how long you have been playing. In case of the penalty profiles, probably codechef is increasing their volatility limits. Hence you observed what you have asked.