Username: 
Password: 
Restrict session to IP 

Sitescore adjustments

1 2
Global Rank: 252
Totalscore: 87267
Posts: 1635
Thanks: 1336
UpVotes: 885
Registered: 16y 42d




Last Seen: 2d 19h
The User is Offline
Sitescore adjustments
Google/translate1Thank You!0Good Post!1Bad Post! link
I adjusted the base-score for 3 sites:

Yashira scores 4000 now (old value 2000)
Brainquest scores 3000 now (old value 2000)
infomirmo scores 3000 now (old value 2000)

We might soon start a rescoring of all sites with user votes as suggestive values.
Note that the sitevotes will not affect the decisions, there will be a seperate voting for adjusting the scores.

Happy Challenging
Gizmore
The geeks shall inherit the properties and methods of object earth.
Global Rank: 1847
Totalscore: 13882
Posts: 17
Thanks: 20
UpVotes: 16
Registered: 15y 154d

Last Seen: 94d 22h
The User is Offline
Sitescore adjustments
Google/translate1Thank You!0Good Post!1Bad Post! link
-5 places Sad
Oh well...
Global Rank: 43
Totalscore: 271164
Posts: 60
Thanks: 60
UpVotes: 31
Registered: 15y 361d
mirmo`s Avatar








The User is Offline
Sitescore adjustments
Google/translate1Thank You!0Good Post!1Bad Post! link
It's a very good idea Smile

I think that the number of chall must count in the base-score.
1 chall of TBS count less than 1 chall of any other chall, i think it's a little unfair. Isn't it ?

Global Rank: 252
Totalscore: 87267
Posts: 1635
Thanks: 1336
UpVotes: 885
Registered: 16y 42d




Last Seen: 2d 19h
The User is Offline
Sitescore adjustments
Google/translate1Thank You!0Good Post!1Bad Post! link
Yes the scoring is kinda unfair imho, and we urgently need to rescore sites better.

How about the following way to rescore the sites ?

-----------------------------------------------------------
Put a poll at each site comments, where all members can vote what the score should be.
Maybe 5 score options are ok. 5000, 8000, 10000, 12000, 15000.

We should vote for each site this way, and decide afterwards which site gets how much points.

I think it might be more accurate to let the users vote than a formula.
Of course the sitescore should only depend on the challenges...
amount, difficulty, work involved.


For the non-english sites we should settle a smaller malus from 0% to 25% (currently its like 50%).

All in all we really need to adjust scores to make it more fair.

The final rescoring will be decided by the Admins and Beta-Testers.

Of course we will need slight adjustments once in a while... for example rankk gained quite a lot new challenges since joindate, but the score did not raise.
We could put _that_ into an algorithm and increase the basescore for sites dynamically.

Greetings
Gizmore
The geeks shall inherit the properties and methods of object earth.
Global Rank: 73
Totalscore: 213061
Posts: 148
Thanks: 206
UpVotes: 107
Registered: 16y 42d
Kender`s Avatar



Last Seen: 2y 13d
The User is Offline
Sitescore adjustments
Google/translate1Thank You!0Good Post!1Bad Post! link
We need to be careful with this.

Letting the users vote just means that people will vote sites they have a high % on high and sites at which they suck low.

Currently users can directly influence a sites score by solving challenges. The more you complete on a site, the lower its score gets.
This is so a site with 10 easy chals will score lower than a site with 10 hard chals.

I do agree that not taking the nr of challenges into account is a bit unfair.
Solving TBS challenges doesn't get you a lot of points because it has so many challenges.
Maybe simply adding challcount*10 to the site score?
On the other hand, TBS has a lot of "duplicate" or "stupid" challenges, while Electrica has very few easy ones.
Do we have automatic challenge counting for all sites yet Gizmore?

I do prefer to come up with some algorithm or automatic system. The scoring now seems too random.
Why is the base score for HackThisSite almost twice that of Lost-Chall? I spent a lot of time on Lost and got to 39%, I spent a single day on HTS and got to 18%..

Perhaps we should also rethink the idea of exponential score keeping.
I feel that solving a site 100% deserves more than twice the points as solving it 50%.
Perhaps percent = (percent*percent+100*percent)/200 ?
That in combination with incorporating challenge-count into the sites score might work.

It remains difficult however. Is getting 10% on TBS worth almost 8 times more than 10% on Electrica? How about 100%?
Is solving 20% on 4 sites worth the same as 60% on one site?

The end goal is to have the ranking reflect some form of respect for the challengers. The question remains what deserves more respect?

Global Rank: 252
Totalscore: 87267
Posts: 1635
Thanks: 1336
UpVotes: 885
Registered: 16y 42d




Last Seen: 2d 19h
The User is Offline
Sitescore adjustments
Google/translate1Thank You!0Good Post!1Bad Post! link
Currently most sites submit their challengecount.
I know that rankk.org is not sending challenge count atm, but i think it will soon.
Tommorow i will check all other sites if they make use of it. (all others do that already iirc)

For the rescoring i was thinking about a formula too.
My idea for base-score was also something like challcount * difficulty (the computed solving difficulty), but it wont work very well i suppose.
For example i think that electrica, although it has only a few challenges, deserves a higher basescore than my above formula would calculate. Maybe rethink the difficulty formula ?


Your idea of expotential scoring on the other hand might work very well:
Quote from Kender

Is solving 20% on 4(typo?) sites worth the same as 60% on one site?

I think solving 20% on 3 sites is much easier than reach 60% on a single site,
thus i vote with yes for your formula.

Edit: I think the main problem is to adjust/calculate the basescores
The geeks shall inherit the properties and methods of object earth.
Last edited by gizmore - Jan 25, 2009 - 01:25:14
Global Rank: 170
Totalscore: 115606
Posts: 166
Thanks: 162
UpVotes: 119
Registered: 16y 34d
Z`s Avatar



Last Seen: 210d 5h
The User is Offline
Sitescore adjustments
Google/translate1Thank You!0Good Post!1Bad Post! link
I like Kender's idea with
percent = (percent*percent+100*percent)/200

But don't know, which one is better, because I had a suggestion about this topic in the past:
final_score_at_a_site = actual_score *(1+(1/(rank+a)))^b
with a and b we can tune this formula, maybe a=5 and b=2 is a good start.

And with both formula I bet I will loose a lot of rank here, but I don't care Smile

Global Rank: 73
Totalscore: 213061
Posts: 148
Thanks: 206
UpVotes: 107
Registered: 16y 42d
Kender`s Avatar



Last Seen: 2y 13d
The User is Offline
Sitescore adjustments
Google/translate1Thank You!0Good Post!1Bad Post! link
20% on 4 sites was not a typo. With the formula I posted getting 20% on 4 sites would equal 60% on one site.

Suggestions:
1. Add challengecount * 10 to the base score.

2. Incorporate (x*x+100*x)/200 into the scoring.

3. Set all english sites to 10000, all non-english sites to 5000.


Global Rank: 170
Totalscore: 115606
Posts: 166
Thanks: 162
UpVotes: 119
Registered: 16y 34d
Z`s Avatar



Last Seen: 210d 5h
The User is Offline
Sitescore adjustments
Google/translate1Thank You!0Good Post!1Bad Post! link
I agree with Kender
Global Rank: 252
Totalscore: 87267
Posts: 1635
Thanks: 1336
UpVotes: 885
Registered: 16y 42d




Last Seen: 2d 19h
The User is Offline
Sitescore adjustments
Google/translate1Thank You!0Good Post!1Bad Post! link
All sites but rankk and electrica are sending challenge count.
----------------------------------------------------------
For scoring adjustments i will try to implement the new formulas soon.
Before it will go online i will post some results here, so we can check again if it would work well.

Greetings
Gizmore
The geeks shall inherit the properties and methods of object earth.
1 2
tunelko, quangntenemy, TheHiveMind, Z, balicocat, Ge0, samuraiblanco, arraez, jcquinterov, hophuocthinh, alfamen2, burhanudinn123, Ben_Dover, stephanduran89, braddie0, SwolloW, dangarbri have subscribed to this thread and receive emails on new posts.
1 people are watching the thread at the moment.
This thread has been viewed 6579 times.