Authority Pumping
Back to General discussions forum
Currently the Authority in ranking table is just the sum of votes received by user.
When I created it I suspect that such a simple formula is vulnerable to Authority Pumping - i.e. two users can vote for every each other's solution with no bad purpose but just out of sense of gratitude :)
Possible solution is to change formula so that N votes from the user Anna to user Bob will give not N points of authority but only:
P = 3-2 * exp((1-N)/2)
It will give the following results:
N P
1 1.000
2 1.787
3 2.264
4 2.554
5 2.729
6 2.836
7 2.900
...
So if Eve have 5 votes from Anna and 2 from Bob here authority would be
2.729 + 1.787 = 4.016
instead of 7.
Does it make sense? What do you think?
Well, it was just the goal - to limit the total amount of authority which user can get from a single other user.
Though the result would be roughly similar with sqrt, so I do not insist...
Ha OK, I did not get it.
Voting 100 times with a square root will bring only 10 points of authority.
Or a base 10 logarithm scale: to get 3 points, you must earn 1000 votes from the same person. Ouch.
How about a "Code was not supplied" button that applies a -1 multiplier to the score for that problem? :)
I find it mysterious that some participants can apparently code solutions to many problems, but have not yet mastered cut-and-paste to show their code.
As for authority, I don't really care about that stat. It's too "free-form" to have any meaning IMHO.
By the way, would it be possible to get rid of the "Confirm Navigation" dialog when leaving a code listing?
> How about a "Code was not supplied" button that applies a -1 multiplier to the score for that problem?
Well, I thought of it but could not came to conclusion whether it will work properly. Not sure we want groups of revengeful people to compete in downvoting :)
What will prevent from pressing this button even if the code is supplied?
And if code is later submitted... Well, I have no clear idea what to do then :)
> By the way, would it be possible to get rid of the "Confirm Navigation" dialog when leaving a code listing?
You are not the first to complain :) Well, let's try to disable this "feature". Probably instead it would be good to make voting button more clear / attractive...
Well there could be ways to manage the "no code" button; people who abuse it could simply have that privilege removed... (although it's one more thing to manage; I'm guessing this site is not your full-time job :) ) it might only apply for some of the harder problems (those with a "blessing" of 6 or 7 and above for example; no one cares how you solved "A+B"). It could require multiple "votes" from different participants to apply. The problem of what to do when something is pasted in later is a bit more complicated...
It's probably more work and management than it's worth.
> The annoying "feature" (confirm navigation) is still there
Please, excuse me for being so slow! I've pushed it just now along with few insignificant changes to other pages. Hope it will not annoy us any more (it affected even admin's account) :)
> I'm guessing this site is not your full-time job
Ha-ha, you are quite right - just a hobby (though it is more interesting to me compared to my full-time job).
It is a luck I found significant feedback of ideas from participants to help me in many troubles!
> The problem of what to do when something is pasted in later is a bit more complicated
Yes, to be honest I always was in doubts about whether we should disable resubmitting of successful solutions. Another problem is that, for example, user can erase the solution after receiving vote :-(
Personally, I like being able to resubmit a successful solutions as my first submission if often to confirm that my implementation is correct (with that dreaded moment: the code works on the example, but fails on the real test data...) Then I'll clean-up the code and optimize a few things like removing intermediary steps that are useful for debugging, but unnecessary in the algo itself.
I like to resubmit too, for the same reason as Guy.
I clean up the forgotten comments, sometimes the code itself.
I also prefer making code refactoring and clean-up so I completely agree with you :)
That is why I never tried to restrict resubmission, though as we discussed it makes some things work out of logic sometimes.
Most correct way was to save all versions ever submitted, but this will require really much surgery of existing code (improving logic of counting results, showing solutions, votes etc) and will yet leave some puzzles in interface to users so probably it is not worth spending time :)