New Scoring

Currently i am also working on a better scoring.

As you might have noticed, SPOJ has a pretty weird scoring, and Caesum got a nice fix for us:

Currently we have an expotential scoring like 0.50 * 0.50 * site_score. whereas 0.50(p) equals 50% solved.

Caesum suggested to change p^2 to p^(1+100/site_challcount), and it really works charmy.

Before we go live with this change, i want to ask you what you think about it.
Please try to spot anomalies for your account in the new rankings.
wc4scoring.
Especially take a look at the totalscores
There are indeed some changes in the rankings, let us know if you are not happy with the new scores coming up.

I'd like to thank you in advance for your feedback,
gizmore

This new system doesn't use 'site_score' at all? That is a somewhat arbitrary value chosen on a site by site basis if I remember right. If so, I'd support eliminating it.

The new scoring does shuffle things a little bit, but not too much from what I can tell, but I am only really aware of the changes near my own rank, for the most part.

The new scoring is the same as the old, except the pow(p, 2) is now pow(p, 1+100/challcount).

Here is the whole old formula to calculate a users score on one site again:

site_score = site_basescore + (avg*site_basescore) + (challcount*basescore_per_chall)
userscore = site_score * pow(percent_solved, 2);

Now in the new scoring, we will change the userscore formula as follows:

userscore = site_score * pow(percent_solved, 1+100/site_challcount);

Also new is that i can adjust basescore_per_chall for each site individually (lowered SPOJ there for example)
and i can adjust the "100", but all sites seem to work well with this default value.

I hope this explained it more deeply and removed your concerns

I would jump 107 places upwards, right now I'm ranked 729 and on the new total scores I would be ranked 622. Is that right? I would presume it is because I get most of my point from 1 site?

Yeah, your score would double for CSTC, you seem "in luck" for the new scoring

Just one stupid question: why is that scoring is better than the old one?
Anyway, I belive the most of us playing these games for fun, and not for points, so in my opinion the scoring system is not so important, however, WeChall is about to make an overall chall site scoring system, and if the new pointing system could show an exacter ranking I think it's cool and should be used.

Best regards
CHItA

Thanks for your feedback.

The main problem with the current/old scoring is that SPOJ scores really "suck".
In the current/old scoring, SPOJ has a basescore of 100,000 (100k), and all but one player have around 2% which does generate almost no points.
In the new scoring it looks way better balanced. The basescore of SPOJ has been "lowered by hand", but because it has so much challs the % really score high.
Hispabyte still has a low basescore, and also you need a lot % to get high scores.

In summary, the new formula seems to reflect challcounts way better, which fixes several minor issues.

Hi

I'd prefer the new scoring system. In the current scoring you could get a lot of points by solving only a few challenges if the corresponding site has only few challenges in total or you can do a lot of challenges and earn nearly no points (like on SPOJ as rather extreme example). So I think each single challenge becomes more valuable in the new scoring.

As far as I could see at a first glance, picking some challengers at random, most of them seem to get more points in the new scoring, but strengthening the lower ranks a bit.

Quote from Gizmore

Jan 19, 2011 - 18:38:18

userscore = site_score * pow(percent_solved, 1+100/site_challcount);

Ok. Got it. Not what it seemed to me at first, but still probably better than the current system.

Its a complicated thing you are trying to do here. I'm not really going to complain if you want to play with it over time. I probably would.

When should we expect this implementation?