If we're gonna revise the ladder, this is where its going to happen.

Discussion in 'Planetary Annihilation General Discussion' started by bengeocth, December 14, 2014.

  1. bengeocth

    bengeocth Post Master General

    Messages:
    1,285
    Likes Received:
    657
    p=current rank
    p2 = other player's rank
    p -= 1/2p2

    Look at that simplicity. Look at it STARE AT IT. ROLL IT OVER IN YOUR BRAIN. DO IT. NOW.

    What's wrong? well, it goes negative. If you are the 8th player, and you beat someone who is 17 places below you, that somehow puts you in 1st place? So clearly, further revision is needed. So... um... why don't we just use that for positive? Let's come up with a negative now. If you beat someone who is below you:

    0 1 2 3 4 5 6 7 8 9 10
    ^ ^
    Your Rank Their Rank

    p += -p2 + p

    So what does this do? Well, in example above, you are player 3 and they are 8. So, replacing all the variables in the equation...

    3 += -8 + 3

    Thus, you would be in first place. That really doesn't seem legit, however. Look at it on a bigger scale...

    238 += -300 + 238

    That would take you down 62 places. Granted, that's not too bad... so that's when we start using fractions! Yes, Fractions! (fractions speak louder than words)

    bad at writing with mouse. p(mod) eq.png = Pmod = -Popp/4 + Pold/4
    Let's try this:
    Your rank: 238
    Their Rank: 300

    238 += 1/4(-300) + 1/4(238)

    so that will make our new place:

    222.

    the quarter system looks good- however, it might be necessary to replace the 1/4 with 1/p or something. tell me what you guys think. But just to make sure this crap works, lets do another example.


    Your Rank = 600. Their Rank = 200. That's pretty damn good for you.

    600 -= 1/4(-200) + 1/4(600)

    Pmod = 100, Pold-Pmod = 100

    New Rank: 500.

    What's the problem with this, you ask?

    It doesn't account for if you are in 10th place, and you beat someone in 600th, you shouldn't be in first.

    So, two equations are necessary- one if you beat someone below you, one if you beat someone higher than you.

    I have created a small console program that you can find here that will allow you to input two values and return them.

    @squishypon3 @igncom1 @cptconundrum @KillerKiwiJuice I think you guys will have some fun with this
    Last edited: December 14, 2014
    theseeker2 likes this.
  2. squishypon3

    squishypon3 Post Master General

    Messages:
    7,971
    Likes Received:
    4,356
    What's wrong with the current system? =o
    xankar, pieman2906 and cptconundrum like this.
  3. bengeocth

    bengeocth Post Master General

    Messages:
    1,285
    Likes Received:
    657
    it makes me sad
  4. squishypon3

    squishypon3 Post Master General

    Messages:
    7,971
    Likes Received:
    4,356
    Well why's that? It is a formal ranking system after all... Glicko was it called? Bah I can't remember, I have a terrible memory. x3
  5. killerkiwijuice

    killerkiwijuice Post Master General

    Messages:
    3,879
    Likes Received:
    3,597
    You made a mistake, logging a larger/lower skilled number for the opponent's rank will increase your rank tremendously in the program.

    Err wait I guess that's what you're trying to show. That can't be correct.
  6. bengeocth

    bengeocth Post Master General

    Messages:
    1,285
    Likes Received:
    657
    I was thinking that if the rank is lower than you, it would get how far the other person is from lat place and do something with that.
  7. bengeocth

    bengeocth Post Master General

    Messages:
    1,285
    Likes Received:
    657
    Something else- in the program, if you are in second place, no matter how many times you face the king, you will never become the king. I will have to change that.
  8. crizmess

    crizmess Well-Known Member

    Messages:
    434
    Likes Received:
    317
    Well, Glick doesn't work on ranks. It uses a number to depict the estimated strength of the player. For example 1560,3 or 1488,1 (the initial value is 1500 in Glicko 2). Ranks a simply derived by sorting according to the strength estimate.

    So if you look at the first and second player, the rank itself does not tell you how far they are away from each other in terms of estimated strength. And this is part of the reason for the situation, where the second player beats the first, but the rank does not change.

    http://www.glicko.net/glicko/glicko2.pdf
  9. bengeocth

    bengeocth Post Master General

    Messages:
    1,285
    Likes Received:
    657
    Well, honestly, if you beat someone 1 rank above you, their place should be assumed... that's just an example of humans forgetting to give a thoughtless machine common sense.
  10. sebovzeoueb

    sebovzeoueb Active Member

    Messages:
    110
    Likes Received:
    71
    I don't think so, it's not because you beat someone that you are necessarily better than that person. It could be that your way of playing counters theirs quite well, or that you just got lucky in some way. I think the current system (or at least it should) takes into account all the people you have beaten, and those who have beaten you. Would it be fair to take someone's place because you beat them this one time, when they have overall defeated players of a higher skill level than the ones you have defeated?
    xankar and cptconundrum like this.
  11. Dementiurge

    Dementiurge Post Master General

    Messages:
    1,094
    Likes Received:
    693
    You can theoretically have a situation where, between three players, there's never a situation where one player is better than the other two.

    http://en.wikipedia.org/wiki/Voting_paradox

    Throw in a few thousand players and it's guaranteed to happen. I don't know why we have to get so scientific about coming up with math heavy ranking systems, when we're really just trying to feed someone's ego.
    xankar, cptconundrum and sebovzeoueb like this.
  12. bengeocth

    bengeocth Post Master General

    Messages:
    1,285
    Likes Received:
    657
    Well, in the current ranking system math is WAAAAAAAAAAAAY more complex than this.
  13. knub23

    knub23 Active Member

    Messages:
    181
    Likes Received:
    152
    Complex math is needed because statistics takes into account your chance of winning and your skill level to calculate your loss/winning points. It further allows intelligent matchmaking. You have to add some statistical values to your calculation, like rating, confidence of rating and winning chance otherwise it is inferior to the current system.
  14. bengeocth

    bengeocth Post Master General

    Messages:
    1,285
    Likes Received:
    657
    I have a question. If I lose to someone who is above me (closer to 1st place) does that negatively affect my rank?
  15. slocke

    slocke Active Member

    Messages:
    121
    Likes Received:
    244
    Look into how the glicko rating system works. Once you have done that you will have a better understanding on why the current rating system is good.

    To answer your question you will always lose points if you lose. You will not lose many points if they are much higher on the ladder than you. They will also not gain many points by beating someone much lower than them.
  16. Dementiurge

    Dementiurge Post Master General

    Messages:
    1,094
    Likes Received:
    693
    Err... umm... what's with these curly braces?
  17. crizmess

    crizmess Well-Known Member

    Messages:
    434
    Likes Received:
    317
    Ugh. I would guess that's normal braces. He just used curly ones to distinguish the term of losing from the others. Well, or he isn't used to LaTeX ;)
    Later on in the example (page 5) the braces are normal again.
  18. squishypon3

    squishypon3 Post Master General

    Messages:
    7,971
    Likes Received:
    4,356
    Braces? I've never heard them called such a thing, normally brackets. Admittedly I'm from the US, though I have lived in Britain... Is this the norm there? Haven't heard it before. :p
  19. cptconundrum

    cptconundrum Post Master General

    Messages:
    4,186
    Likes Received:
    4,900
    Brackets are [] and curly braces are {} to me.
  20. reptarking

    reptarking Post Master General

    Messages:
    1,321
    Likes Received:
    1,577
    I thought {} were only used to make weird faces......

Share This Page