Errichto's blog

Do we like the dynamic scoring?
By Errichto, 9 years ago, In English

If you don't know what the dynamic scoring is, you can read about it here. In my opinion it's a useful alternative. With smoother steps (250 instead of 500) it doesn't work so bad. Or does it? The main advantage is that organizers don't have to correctly estimate the difficulty. What are drawbacks? And should it be changed and improved in some way?

  • Vote: I like it
  • +87
  • Vote: I do not like it

Comments Comments (14)
| Write comment?
»
9 years ago, # |
  Vote: I like it +501 Vote: I do not like it

You just want the first place in contribution list, don't you?

As we discovered at VK Cup Final Round, it works awful on rounds with small number of participants. It becomes too discrete and unsmooth.

→ Reply
  • »
    »
    9 years ago, # ^ |
    ← Rev. 2   Vote: I like it +199 Vote: I do not like it

    I was on the first place before writing this blog and now we have a tie. Damn. A few more upvotes for you and I will consider removing my blog to also remove your comment.

    There were only 20 people in the finals, right? So it shouldn't a problem when someone decides about a standard CF round? Though P_Nyagolov and Um_nik gave strong arguments against it too.

    → Reply
    • »
      »
      »
      9 years ago, # ^ |
        Vote: I like it +100 Vote: I do not like it

      My work here is done

      → Reply
    • »
      »
      »
      9 years ago, # ^ |
        Vote: I like it +44 Vote: I do not like it

      As far as I know, removing a blog will not affect your contribution (and contribution of those ho commented on it). BTW, you can prepare another CF round as a simple way to secure first place — at least for you it should be simple, taking into account how productive you are:)

      It is also going to work bad in div1+div2 contests — but in opposite way. Take a look at standings of recent Manthan contest :) It had two issues at the same time: unbalanced time penalties (submitting at 2.30 gives not a lot of points) and dynamic scoring. Results look weird — like people getting more points for C than for H... Really? Even implementing bruteforce for H should give more points :) For contest with several thousands of participang current dynamic scoring formulas don't see huge difference betwen problem solved by 3 people and problem solved by 60 people — both numbers are considered as very small. That's similar to an issue described by Um_nik, but the more contestants you'll have — the more notable it is going to be at the upper end of a scale.

      → Reply
  • »
    »
    9 years ago, # ^ |
    ← Rev. 12   Vote: I like it +15 Vote: I do not like it

    What a Clear hard match with Contribs.

    So Edvard will be added soon with help of Educational Rounds. :)

    → Reply
»
9 years ago, # |
  Vote: I like it +71 Vote: I do not like it

I think one of the disadvatages of dynamic scoring with step 250 is that the penalties for unsuccessful submisions on the easiest 1-2 problems (those with score 250) are way too painful :)

→ Reply
  • »
    »
    9 years ago, # ^ |
      Vote: I like it +28 Vote: I do not like it

    Maybe the penalty could depend on problem difficulty somehow? I don't know if it is a good decision but anyway.

    → Reply
»
9 years ago, # |
  Vote: I like it +124 Vote: I do not like it

Only the number of participants who solve the problem counts, time and amount of unsuccessful attempts don't. Usually the scoring looks like 250-250-1000-3000-3000. Score for problem B is too low, for problem D — too high.

→ Reply
»
9 years ago, # |
← Rev. 2   Vote: I like it -29 Vote: I do not like it

Deleted.

→ Reply
»
9 years ago, # |
  Vote: I like it +13 Vote: I do not like it

Maybe you can modify the function that determines maximum score of problem. Something like this — If the final maximum score looks like 250-250-1000-3000-3000 , then you allocate A=500, B=750, C=1750, D=2250, E=2500. And then calculate score based of these allocated scores. Basically, use dynamic scores to get a hint of the difficulty, but use a more even distribution to do the real scoring. Just a thought

→ Reply
»
9 years ago, # |
  Vote: I like it +13 Vote: I do not like it

No.

→ Reply
»
9 years ago, # |
  Vote: I like it +11 Vote: I do not like it

Dynamic score is a great thing for problemsetters — you need not think about appropriate scoring, just use dynamic and let it go:)

→ Reply
»
9 years ago, # |
  Vote: I like it +45 Vote: I do not like it

Dynamic scoring adds another dimension to participant's strategies (which would be a good thing by itself), but does that in an inconvenient way.

Imagine a participant who has two problems left to solve, D and E. This participant knows he would probably solve any one of the problems but not both. What should he solve to expect a better rank? When the points are known in advance, the answer is clear. When the scoring is dynamic, too many things can change: pretests of unknown strength, hacks, system tests, other participants' choice all alter the score. It is difficult to make a good decision in such situation, and the essence of this difficulty is a bit too far from algorithmic problem solving.

The same goes for rounds with cutoffs, say, top 300 advancers or top 50 T-shirts. For participants who end up around the cutoff, there is no clear strategy even in retrospect to maximize their chances. It just seems random at this point.

That said, I'm fine with dynamic scoring when it's used in regular rounds (many participants, no cutoffs). It encourages experimenting with strategies.

→ Reply