General discussions: New Scoring system for Challenges

Rodion (admin) 2014-08-29 17:33:06

As it was shortly discussed here new scoring system for challenges under testing now.

In brief:

nothing is changed in gaining points for problems themselves - those shown in column Score of challenge stats;
bonus enlightenment is awarded now simply according to the place in challenge results table - the person at the first place gets exactly same additional enlightenment according to the exponential law.

Exact formula is:

bonusEnlightenment = taskBlessing * exp((1 - position) / 16)

So that if blessing for task is 12 the first place gives 12 points for solving and 12 more points, the second about 12 + 11.27, the tenth about 12 + 6.42 etc.

Please do not be afraid of slight changes of total enlightenment.

It looks that some features of interface should be slightly changed according to new system and helping in understanding it - I hope to make them soon!

nicolas_patrois 2014-08-29 19:56:25

I think that the challenges should not give so many points, for example max 1 then decreasing ie exp((1-position)/16).

I see them as a bonus for the top two not as a way to count a task twice.

Guy Gervais 2014-08-29 20:22:32

I think the fact that challenges now give a lot of points might encourage more people to participate.

I'm wondering about having some challenges be valued higher than others. For example, I'm ahead on Max-It (I get 10.769 points) while Sergei is ahead on Color Cubes but gets 10.451 for it. Of course, having all challenges worth the same number of points will bring back the original problem of top players with identical scores in some situations.

Rodion (admin) 2014-08-30 08:45:54

Thanks for your hints and suggestions - it seems both are right so I simplified the formula to:

7 * exp((1 - position) / 16)

So the first place gives fixed amount of 7 bonus points. This also simplifies recalculation code a bit... So thank you!

> I think that the challenges should not give so many points

Guy's suggestion is correct - it was just out of my with to attract more attention to these problems. But your objection is somewhat just also - so I currently choose coefficient between 1 as you proposed and max blessing of about 13 which was achievable with previous scheme... :)

> I'm wondering about having some challenges be valued higher than others.

After thinking more about your words I became not quite sure there is a good reason for this. Moreover dependence on blessing can lead to minor cheating with creating additional accounts.

> original problem of top players with identical scores in some situations.

It seems to me that having different tasks scored differently does not truly solve this problem since difference of points is somewhat "artificial".