Rating upgrade

Post Reply
Field Mouse
Posts: 13
Joined: Sun Oct 12, 2008 1:47 am

Rating upgrade

Post by Field Mouse »

Hello,

I think the rating is not solved well now.
The problem of the current system is that the rating depends much on how many people tried to solve the problem - if a problem with 70% difficulty has 1000 tries and a problem with 95% difficulty has 100 tries, then the first of them will have rating much much bigger, even though it is easier. So after all problems with the highest ratings are those which are *easy* enough to attract people to try to solve, but *hard* enough to have more than something like 60% difficulty so their rating grows in time as more and more people tries to solve. Rating does not say much about the problem and it does not even say much about the user.


The solution I recommend is this:

The rating should be fix and should be set by a problem difficulty (for example each 1% of difficulty could be worth 27 rating points to make the maximum rating 2700 like the GoR for 7d). The changes in players rating could be counted in a similar way like the GoR them.


Advantages:
- the rating will express much better players strenght
- players will know that if they, for example, have rating 2000, that their strenght is equal about 1k Goproblems
- the huge amount od older problems whith only a few tries after the new rating thing was launched will get correct ratings
- this system would look professional and would attract much more people to buy the pro accounts :)

The disadvantages:
- I guess this would be quite hard to implement
- more motivation to cheat for some users

:-)

~~~(,_,_*>
gj5
Posts: 3
Joined: Wed Oct 29, 2008 7:52 am

Re: Rating upgrade

Post by gj5 »

Yep, I fully agree. The contribution of a problem to a user's rating needs to be weighted by how many people have tackled the problem. In addition, it would be even better to also log the strength of the user when he/she does the problem. For example, if I solve a problem that means that it's dead easy, but if fieldmouse does _not_ solve a problem, it means it's serious :-)
I think this would only work based on data from the pro accounts. So you could have a pro rating (which would also attract people) and then you could compare the other users to the data from the pro rating. That may or may not be doable. I'll have to think some more about this rating thing...
Post Reply