Announcement

Collapse

The Infinity Blade Forums Have Moved

We've launched brand new Infinity Blade forums with improved features and revamped layout. We've also included a complete archive of the previous posts. Come check out the new Infinity Blade forums.
See more
See less

How does the ELO ranking work?

Collapse
X
  • Filter
  • Time
  • Show
Clear All
new posts

    #76
    Btw, note in the Elo wiki it says somewhat down below something about "Game activity versus protecting one's rating" - this is exactly about players not wanting to play specific players in order to not damage their ranking score. Like a high ranked player playing a lesser skilled and loosing ranking points to him eventhough he wins.

    quote from wiki

    "Elo rating for a World championship model.[10] Nunn highlights in the section on "Selection of players", that players not only be selected by high Elo ratings, but also their rated game activity. Nunn clearly separates the "activity bonus" from the Elo rating, and only implies using it as a tie-breaking mechanism."

    /quote from wiki

    Comment


      #77
      Originally posted by btcomm
      2. What are you talking about? You may just as well lose? What does that even mean? How is this statement pertaining to anything? Please explain why I may as well just lose?
      Youre gaining higher scores from better ranked players, but also a higher probability to lose the match. statistically you will not only win matches with a very good score, but also lose matches, with a very bad score. (not so vs lower skilled players)

      I have already said that there are better players ranked lower then me, because they don't go out of their way to play with top ranked players, if they did they would go up in rank, meaning the only reason they aren't top ranked is because they choose not to be NOT because the rankings are broken.
      Not true. There is no need to play a specific player to outscore him. This can be achieved by beating other ppl. Also there are plenty of high skilled players, who play public and default servers. In instagib. Top players arent on top because they do not specifically adjust their playstyle to get a good score, which defeats the purpose in such a system, which is supposed to measure skill, and not playing time.

      You ignored my provided link with reality example and scores. Well here is another, now from ut2k4 stats. From the Nr1 Deathmatch player of the week(month/alltimes are highly suspective) Here. Notice lots of short matches, lots of noob players, lots of ownage, lots of scores. Now it would be good to see pictures with stats of games full of high ranked players, since youre saying its the way to go. (im not disagreeing it isnt one of the ways) But still, where is it.

      Comment


        #78
        Originally posted by btcomm View Post
        Where is the love? I play with those guys too and I'm in the top 20 right now hehe. I think I played with you tonight, your name looks familiar.
        I don't think I'd played against you before last night. And yes, you play very very well, props man.

        Anyway, played with a lot of the players named above, including btcomm last night and I was finishing in top 4 (usually 3 or 4 with a k/d ratio of 1.5:1) and my ranking went from 70ish to 48. Then I played this morning agianst a server of about 6 much lower rated players and won 41 to 7 and my ranking went from 43 to 78.

        For people who are flaming the ratings, I agree it can be distracting. But I'm more of a goal oriented person and I like there to be some kind of measure of how I'm progressing. Sure 'just having fun' is a goal, but that's easily achieveable. The ranking system, as flawed as it is, gives me a little more to shoot for (pun intended again), especially when I regularly devote an hour of my time to this game.

        Comment


          #79
          Originally posted by Pser View Post
          Youre gaining higher scores from better ranked players, but also a higher probability to lose the match. statistically you will not only win matches with a very good score, but also lose matches, with a very bad score. (not so vs lower skilled players)
          Ok, so you're saying that vs good players I'm more likely to lose the round, well that makes sense, also makes sense that they would still rank me higher if I lose to good players then if I win with newbs, (meaning I'm not newb bashing to get a high score)

          Originally posted by Pser View Post
          Not true. There is no need to play a specific player to outscore him. This can be achieved by beating other ppl. Also there are plenty of high skilled players, who play public and default servers. In instagib. Top players arent on top because they do not specifically adjust their playstyle to get a good score, which defeats the purpose in such a system, which is supposed to measure skill, and not playing time.
          Yes, there is no need to play a player to outscore him, which is actually good because if you NEEDED to play a player to outscore him he could get to the top and retire his name and never play again with that name and no one could beat him him out on the list ever again. At the same time you need to play very high ranked players and do better then them to go above them in ELO. If you are not playing high ranked players you will go up very very slowly and unless you can go without being killed at all, most likely you are going to go down. Instagib is not default, it is a mutator therefore you do not get ranked for normal servers when you play it for good reason, it's 1 shot 1 kill and completely different then normal weapons, what I think they should do is have a separate ranking system for insta but because it's more work instead they just have a non standard server ranking which by no means can be accurate at all because of all the different game types that are available.


          Originally posted by Pser View Post
          You ignored my provided link with reality example and scores. Well here is another, now from ut2k4 stats. From the Nr1 Deathmatch player of the week(month/alltimes are highly suspective) Here. Notice lots of short matches, lots of noob players, lots of ownage, lots of scores. Now it would be good to see pictures with stats of games full of high ranked players, since youre saying its the way to go. (im not disagreeing it isnt one of the ways) But still, where is it.
          Ok, that is 2k4, let me just say that in 2k4 I noticed that it was much more forgiving in stats about playing with newbs and getting good scores. I used to do that all the time and I would continue to go up and up and up. In ut3 this is not so.

          There have been multiple people on this board that have said the same thing as me which is when they play with noobies they go down in rank. They end up with much better k/d ratio's like say 30 and 3 or 30 and 6 and in the end when they check their stats they went down 30 to 40 places. Sometimes less but just like the number 1 ranked person said, he played some games with newbs, did really well in k/d ratio and points but went down in ranking. Just like (rypt0 said, he played with noobies he did much better in k/d ratio but when he checked his stats he went down 30 to 40 spots. The truth is if you play with lower ranked players you will go down.

          Comment


            #80
            Hasn't this been discussed enough already? o.O

            Comment


              #81
              in ut2k4 Ig/Nw ranks are separated. But hey, Giggity is right.

              Comment


                #82
                Yup like I said, case closed.

                I win !

                Thread locked !



                lol

                Comment


                  #83
                  Not closed. One thing that you all got wrong is that your rank will always go down if you play with lower ranked players. This is only true if you allow them to kill you. If you go 30-0 in a deathmatch with sub-1000 rated players, you will still gain a few ratings points. Probably not enough to have your ranking affected, but won't have it go down, because you will never lose points. The only reason some players think they lost points by playing against newbies and not dying is because someone right under them in rankings played better players in that same time period and thus gained enough points to surpass them.

                  Comment


                    #84
                    Hah found this thread searching my name on google.

                    Anyways, ELO != skill, it is however a measurement of performance. It might not be perfect but its not luck/random/meaningless. It makes sense, look up how it works. People arguing statistics in sports never gets old. But no matter what is said against them they continue to prove themselves. Quarterback ratings... not perfect but not totally random. If you search enough you can read more on this **** on how statistics are being accepted more in professional sports. Not saying they're perfect but they are used to produce results... thats enough proof for me that they're not worthless :P.

                    Anyways with UT3 few things to take into consideration:

                    1. The time arguement is ignoring the fact this game is still relatively new. If the same people kept playing the game with no new players, the players didn't get better or anything and remained the same 'skill' level, and players played the gave forever (time approaching infinity).. everyone's ELO should settle down on a specific number. At the beginning everyone's ELO will be exactly the same and obviously not the case. ELO is performance based. The more you play the more data there is to create the rating. So yes the more you play the easier it is to raise your rank as there is perforamnce data on you. HOWEVER that doesn't mean if you just keep playing more than anyone else you can keep raising you rating. You will hit a limit doing this. This can already be seen in UT3. Make a new account and play a few games, your rank will jump a ton. Every game you play you rating might to up but won't go up as much. You WILL approach a limit. People who play more might approach a limit faster but over time it won't matter.

                    2. Warfare/CTF screw up ELO rating. DM is hard to rank much less gameplays that arn't kill/death oriented. CTF players get screwed and will never be high on ELO ranking. Warfare players used to control all the top ranks and couple weeks ago the rating system changed and now they've all fallen in rank. Anyways ELO in other game types is almost 100% bs. In DM it makes the most sense.

                    3. ELO is a performance benchmark on the game being played. The game being played is not 1v1 or clan's playing TDM... its pubs. The game is who can spam the most and get cheap kills. That is what ELO is ranking. ELO is ranking a different 'skill' than what its usually thought of. They are somewhat related but not exactly the game. A good player might spam well on pubs or they might not.... or a person who spams well on pubs might be good at dm, or they might not. Keep in mind what ELO is actually ranking.

                    4. You can't kill newbs to raise your rank. It will for awhile but eventually you'll hit a limit where it will no longer work. If you are killing newbs that means you're better than them and yeah your rank should go up. But once you get to some point you'll notice its no longer effective... thats cause you're ranked high and just to stay even you need to rage when playing newbs. If you get ranked high you pretty much ar required to spree vs newbs or your rank WILL fall.... it really does need to be an insane ratio to stay even much less gain.

                    Comment


                      #85
                      geeze i hate to open up a sorely battered thread, but this thread's got some inaccuracies which just may be due to it's age, but when googling this thread comes up and confused me so feel i must post here.

                      i am a ut3 modder i have looked at the current ut3 script code. ut3 stats are mostly written at end of game. you own private stats like number of kills, etc are written. then the group stats are written which rank is based on. to most of your surprise, not a lot is written. the player id of each player and each player's score is sent to the stats server, that all. i looked in the ut3 code and when you kill someone the kill count for that damage type (plasma, flak, etc) is incremented, so that your kills per weapon can be reported on stats page, but system doesn't seem to care you killed Mr Big Guy.

                      So it would seem the rankings are not based on who you kill, but based on your game score verses other human player game scores. From experience I can say if you play lower ranked players your rank tends to go down, and against better ranked players (if you score well), it goes up. problem is some really good players have bad ranks cause they don't even try to raise it, they just play random pubs like me, so it's easy for a lower ranked player who is also a good player to outscore you and hurt your rank, especually in team based games (yup team based games put the system in a bad way).

                      Good Rank can just mean you know how to manipulate the ratings system to your advantage, but you can still be just a normal skilled player. I play warfare a lot on pubs so it's easy for even a noob to out score me if he's on a good team and my team is poor. then his rank goes up and mine goes down. if i outscore him, my rank stays put cause system knows outscoring a lower rank means little.

                      so the key seems to be, play against better ranked players and outscore them, and you should raise the rank. easy to do in warfare. just join the strong team if you can, grab every vehicle can so you can own people on foot, heal and destroy nodes like crazy, and especually, and i mean especually, camp the power core for big points. and lol and behold the camping noob who did nothing but camp the core is at the top of the scoreboard wtih big points and skyrockets in rank. core damage points are really too high and puts things off balance in warfare, as any noob can camp it and let his team do all the work but he gets all the big points. lol.

                      if you really way outscore a much better ranked player, your rank will skyrocket in one match. obviously the more time you play better ranked players and outscore them, the more your rank goes up, so time played is a factor as long as you keep playing better ranked players and score well.

                      it also doesn't seem to matter if your team wins or not. in the stats array sent to server it just sets everyone's team id to 255 (that means no team) to ignore what team you are on and just sends your player id and score. so your team could lose but you score at top of scoreboard anyway and still get a rank boost if lower scored players had better rankings.

                      make sense?

                      this general scheme is also in accord with this thread:
                      http://forums.epicgames.com/showthread.php?t=685097

                      Comment


                        #86
                        All I know is that if you play against good players you will rank up a lot comfortably (assuming you do well against them). If you do great against newer or average players, you'll still barely increase at higher levels.

                        My guess is that most of the top players play each other regularly and boost their own stats by simply being high in stats.

                        There are definite limitations as I really don't think players with a 1:1 kill/death ratio should be anywhere near the top.

                        I think the main thing is that... if you're playing regularly... a decent player will exceed 2700 ELO and 60% efficiency+. Apart from that... you have to play them.

                        I get a good efficiency (70%ish) but I'm sure I could still be more effective in terms of winning games. I have been working on that (e.g. focusing on nodes over killing the defender or getting more flag caps), yet the efficiency goes up when I try to do this anyway! Crazy

                        Comment


                          #87
                          Originally posted by warhead2 View Post
                          make sense?
                          Of course not! I understand what you are saying, and I'm not arguing, but applying ELO to a team based gametype doesn't make any sense anyways, so why should anything else about it? Waaaaahhhhhhh!!!!!
                          (elo cannot account for your teammates)

                          Correct: R'a = Ra + K(Sa-Ea)

                          R'A represents the new rating
                          RA represents the initial rating before the most recent competition
                          K is a factor used to change the significance of the result based on the setting
                          of the competition.
                          SA is the number of points scored by the player
                          EA is the number of points the player was expected to score

                          Correct: Ea = 1/1+10^(Rb-Ra)/400
                          RB is the rating of the opponent and RA is the player's rating

                          Now here's where they go all wacky: R'a = Ra + KG(Sa-Ea)

                          The "G" has been introduced to alter the equation. The part that's wacky is... G is a freaking variable!!!!!! There is a floating "weighted" system that alters what value G holds, and it has it's own crappy altered equations. On top of the already complex K which is usually discarded because it weights the forum (Grand Championship of Chess carries more weight than winning a regional final). I'TS NOT EVEN REALLY ELO anymore.

                          Been dying to say that. Wah!

                          Comment


                            #88
                            i like elo ranking

                            http://www.dfbgaming.com/ut3_stats/i...ame_name=ut3pc

                            Comment


                              #89
                              yay i opened up a sore thread. from now on, you can call me "Dr Evil"

                              yea i totally agree., slick, team based types, especially warfare with it's unbalance core damage points, really whacks the system. yup i agree too it is not even Elo anymore.

                              thanks a heck of a lot for those equations. not sure where you got them at but good to see them. they further verify that game score is one of the key factors and that it helps to play against better ranked players, whereas just playing against poorer ranked players tends to pull you down. isn't it nice to rip the hood off this thing and demystify it?

                              your teammates do matter. all players are sent to server as individuals and team is ignored, so you are competing against your teammates too.
                              however playing on a weak team pulls your score down, playing on a strong team boost your score, so in that way (if that was what your were saying), no, elo really can't perform well for team based types.

                              please clarify what "Se" means, and the ea equation can use some more () to clarify how it works.

                              Thanks!

                              Comment


                                #90
                                Se is a mistake, should be Sa, fixeded it. The equations are from Womens Soccer, the only official athletic governing body to use elo to rank a team based sport. The only reason it isn't controversial is...... well, it's womens soccer. Epic would need to introduce thier own weighting equations to the sliding variable based on the players in the match.

                                Comment

                                Working...
                                X