Recently, I'm interested in the rating distribution and I searched this.
Method of searching
- Searching the place of people (in active users) who has rating 999, 1049, 1099, 1149, 1199,... , 3049, 3099.
- Let pn be the place of people who has rating n.
- The number of people in range [1000, 1050) is p999 — p1049, the number of people in range [1050, 1100) is p1049 — p1099, ...
However, there was some surprising thing about it.
Look at following graph:
- The number of people who have rating [1900, 1950) is especially larger than [1800, 1850), [1850, 1900). ( twice larger approximate )
- The initial rating is 1500, but rating [1350, 1400) is about twice larger than rating [1450, 1500).
- The average rating is about 1450. This is lower than the initial rating: 1500.
Why these things can be occur?
UPD: The graph was extended.
UPD2: The average rating was calculated.
UPD3: The distribution and searchings about codeforces contribution is Here.
[1350, 1400] is probably the range your rating will fall into if you perform poorly in your first contest and people would create another account instead of keeping that one.
[1900, 1950] is where Div 1 clones can't participate in Div 2 contests.
I agree with you, if he considered only people with lets say 5 or more contests, the results would be more like the expected.
About [1900, 1950]: there are many people who once got into div1 and after that they don't participate in rated contests just because they don't want to lose their color.
Why should they want to stop once they become purple? Why not to stop at any other color? What is so exclusive in the purple? The fact that it is the first div1 color is not enough explanation, I think. It rather must be some oddity of the rating system or competition format on Codeforces.
There are people who do that, I know some examples. But I agree that it isn't the main reason, the main reason is fake accounts. I think if you take users with at least 5 contests, the graph would look much better.
Can you please explain, how fake accounts affect that?
When there is a Div.2 only contest, many Div.1 users make new accounts to participate officially. Since they usually do well in these contests, the account becomes Div.1, so they make a new one and so on.
Why would div.1 guys want to participate officially in div.2 contests? Is it because of the goodies awarded to the winners?
Just like ania said below:
Okay. Thanks. :)
I think there have only been like 3 or 4 rounds where the Div 2 winner got anything...
It's not prize, it's pride.
Isn't purple div 1 cutoff? I think reaching div 1 is the most common (realistic) goal. Some people feel like they will spoil the achievement if they compete more and fall down to div 2. You can notice that there seems to be a slightly more people in 2400-2450 than 2350-2400 (although it's hard to be sure since both bars are small). I suppose this comes from the goal to get red.
I still can't imagine, why would someone want to get to div 1 if he is not going to participate in div 1 contests? If think a real goal, which most coders consider worthy, is to become red, and that indeed can explain the small spike in 2400–2450, but it looks very small relative to the huge spike in 1900-2100.
For the sake of achievement I guess. If you are close to the cutoff and plan to retire because you have other things to do, you may want to feel you accomplished something.
And the spike is so huge probably because of the fake accounts, i.e. some people from div 1 want to feel the thrill of winning a contest but are unable to do so in div 1 so they create a new account and (usually don't) win div 2 contest.
Why no orange and red and nutella part of distribution that's discrimination fix plz :((((
UPD: Updated version of the plot is even more offending!!1 Why everybody with 2400+ are joined together that's terrible i'm leaving
You wrote ɹǝqoqoןz, not Zlobober.
I give up :D
Because of a lot of div1 guys who just make accounts before div2 rounds and get a place in top 10. Otherwise, the distribution is normal it seems.
This graph is made with all users or only with the active users?
Only with active users. It is written (or might be edited) in the blog.
Thank you for many comments. :D
I'm interested about the contribution distribution too, so I searched and summarized in this blog.
Thank you! If you want, let's see the distribution and searchings about contribution.
amazing blog
If you want to know why this comment is downvoted, see the revision (If you don't know how to, click "<-" of "<- Rev.8" like strings, which there are left of vote rating like "-18".)
If you see Rev.7, I think you can see that he want to steal the username and password, for creating fake account that has zero submission. (If I'm wrong, I'm sorry) BTW, that's amazing blog :D
I cant seems to see the image. Can anyone share it?
Seeing revision 7 and see the source of this page, you can see that the image (?) is connected to www.897theriver.com/admin.
yes, but it seems like authentication is required to see that page?
If it is, I'm sorry my bad. But just seeing Rev.7, you can see the same form. I think he/she is intended to do something.