Page 3 of 5 FirstFirst 1 2 3 4 5 LastLast
Results 21 to 30 of 41

Thread: Matchmaking is fine. There are other things that need to be fixed.

  1. #21
    Beta Player Marquess Genavelle's Avatar
    Join Date
    Jun 2017
    Posts
    788
    Level
    50
    Points: 6,101, Level: 50
    Level completed: 76%, Points required for next Level: 49
    Achievements:
    Closed Beta Tester Forum Tier 3 5000 Experience Points 1 year registered
    Saying Matchmaking isn't a problem is just ignorant. You can go to the Matchmaking thread and see screenshots of Matches from Paladins.Guru, which SHOW plenty of examples where Ranked matches had a crazy range of Elos and rankings. I mean sure, someone could have gotten unlucky in their qualifiers and been given a low rank...but their Elo is probably a little more accurate. And when you're getting like a 500+ difference between teammates' Elo, or worse- one team has a significantly higher average Elo...then that's an issue. When one team is stacked with parties, and the other team is 5 solo players, that's a problem.
    The Stone Oracle of Blades

  2. #22
    Member Recruit
    Join Date
    May 2018
    Posts
    6
    Level
    1
    Points: 57, Level: 1
    Level completed: 14%, Points required for next Level: 43
    From a quick search I did, unless there is something I missed, most of the pics I saw were either from casual siege, or very old (nearly a year) or both, so I don't really see how they apply here. Comparing casual siege and ranked is not possible since one is highly competitive and the other is as the name suggests casual. The algorithm runs a lot looser to give as lower queue times as possible in the later.

    I believe that making remarks like x/y/z is broken without giving any explanation as to what you believe is broken in it is not constructive at all. So again I will say that from my experiences in the ladder, the problem lies only on the fact that MMR/ELO offer an incomplete picture of why someone won or lost a match and as a result MM gets a lot of false data. This means that MM works as intended but due to the bad data it gives bad outcomes. Saying that MM is broken is like blaming someone for failing a test when he is taught the wrong answers.

  3. #23
    Beta Player Baron TrueVingador's Avatar
    Join Date
    Mar 2018
    Location
    In a house
    Posts
    243
    Level
    15
    Points: 895, Level: 15
    Level completed: 95%, Points required for next Level: 5
    Achievements:
    Closed Beta Tester 500 Experience Points 3 months registered Forum Tier 2
    Casual is not that casual as it is suposed to be. And guess for what? Bad MM...

    Well.. anyway, I supose that we're facing a dillema here of cause and effect:

    1. Does the MM influence Ranking System?

    2. Does the Ranking System influence MM?

    3. Does both influence each other?


    The OP is suggesting the 2nd option. I strongly believe the 3rd option. If both are not fixed at the same time, we'll be runing in circles.

  4. #24
    Beta Player Emperor Valkure's Avatar
    Join Date
    Feb 2017
    Posts
    11,441
    Level
    100
    Points: 74,676, Level: 100
    Level completed: 0%, Points required for next Level: 0
    Achievements:
    Closed Beta Tester Overdrive Tagger First Class 50000 Experience Points Forum Tier 4 1 year registered Forum Tier 3
    Quote Originally Posted by Orion532 View Post
    I don't really feel matchmaking is the problem because it is not to blame here. Matchmaking does what its supposed to do based on the data that are provided to it. It's the data which are put into it that need to be cleaned and polished in order to fix everything.
    Actually, no. We KNOW MM considers ONLY W/L rate to create games. Now, if I am on a 5 winning streak at level 30 of my account, I could easily finish matched against Mutu, on another 5 winning streak, but at level 500 and what else, with an elo of 4000 points.
    Is it fair? I doubt.

    W/L should not be the only one thing considered when weighing players in order to create teams.

    Quote Originally Posted by Orion532 View Post
    From a quick search I did, unless there is something I missed, most of the pics I saw were either from casual siege, or very old (nearly a year) or both, so I don't really see how they apply here.
    1) there are tons of new pic, but people are tired of posting the same problem over and over again. In the meantime, MM has not been reworked, so also older pics "apply".
    2) casual and ranked share the same MM. Of course, it works with W/L of casual and ranked game, respectively, but the algorithm is the same. So, they all "apply".
    I've finally found it!

    New to the forum? Have a look >>>here<<<! Some useful answers.

    #StopPlayingSupports
    Let them see how a game with shit-nerfed supports is.

  5. #25
    Banned Grand Prince
    Join Date
    Oct 2016
    Posts
    2,882
    Level
    98
    Points: 27,358, Level: 98
    Level completed: 1%, Points required for next Level: 992
    Achievements:
    Closed Beta Tester Overdrive Forum Tier 4 Forum Tier 2 1 year registered 25000 Experience Points
    @Orion : Basically what you want is for MMR to be calculated taking into account personal performance, weighted by the Win or Loss and relative to the champion you play and the average result of said champion in a n<MMR Range<n+x .

    Quick exemple : You play Shalin and your MMR is 2000, you Lose a game doing 100k damage and 10/10/5 kda

    --> On average, Shalin in the 1900<MMR<2100 range does 89k damage and go 6/12/5 kda

    --> Based on that, one can tell that even if you lost you performed better with your champion that the average for said champion, on a lost game, in your MMR range, ergo --> You loose Less TP and MMR because your performance was above average

    Using the same method, and assuming you gain a fixed 10 TP for a win and lose 10 TP for a loss, you can either :

    - Win more TP if you win a game and if you performed better than average : +15 TP
    - Win TP if you win but performed worst than the average : +10 TP
    - Lose less TP if you lose but you performed better than average : - 5 TP
    - Lose TP if you lose and played worst than average : -10 TP

    These fixed values will :

    A) Still segregate the players by skill, someone who always carry will mathematicaly have higher MMR and TP that someone who allways perform below average, so the same WL ratio will result in two distinct MMR, TP and thus rating.
    B) Will hit the troll / afk / throwers the hardest, because they'll allways perform below average thus will be at the bottom of the MMR range
    C) Be a motivation for everyone because less punishing than the actual system who shift the TP gain / loss more and more with time and win / loss streak only (the more you lose, the less you gain TP and the more you lose, the game pull you toward the bottom constantly after the initial artificial push)
    D) This will ensure that someone with an above 50% winrate gain TP on a somehow constant basis


    The flaws are that this system doesnt take team performance into account --> Fix : incremental +1 / -1 TP based on an index of average comparison between the team members. Exemple you outperform everyone in your team, play way above average and the rest of your team play all below average, you gain another +5 tp, the second best in the team gain +4, the third +3 etc etc and in the losing team -1, -2, -3 etc.

    This will segregate more the good from the bad AND will force everyone to play at their best because they will compete to win, play good and play their role with said champion AND compete within their own team to be the best (ie ; the carry).

    Last but not least : What is an estimate of a "good" performance?

    The usual thing that keeps coming back is "Yeah but if I'm a flank pip I shouldnt be judged on my heals, because indeed they will be lower than a healer pip", to what I answer, we have statistics, use it! You can create an index based on damage + healing + KDA + Shielding + point cap and weight this by the LC you chose.

    Exemple :
    Performance index for pip with megapotion --> damage + healing*2 + KDA + Shielding*0 + Point cap
    Performance index for pip with catalyst --> damage*2 + healing + KDA*2 + Shielding*0 + Point cap

    This way you'll only compare performance segregating by the LC and everyone can play the role he wants for said champion and still be compared to average for said champion WITH the same LC.

    I think I covered most of the basics of the system I'd like to see implemented and answered all the problems that could rise. Feel free to discuss this because usually I write a Wall'o'text for nothing ^^ This post will now go into my signature to save me some bleeding fingers in the future.
    Last edited by EpplZ; 05-30-2018 at 02:40 AM.

  6. #26
    Beta Player Emperor Valkure's Avatar
    Join Date
    Feb 2017
    Posts
    11,441
    Level
    100
    Points: 74,676, Level: 100
    Level completed: 0%, Points required for next Level: 0
    Achievements:
    Closed Beta Tester Overdrive Tagger First Class 50000 Experience Points Forum Tier 4 1 year registered Forum Tier 3
    I'd like to point out another thing. Carrying capacity is not the same for all champs and no matter how good you can be, a Grohk is carrying less than a Drogoz. If you make that performance is lessening the loses/increasing the victories, but the majority of point are still related to W/L, the outcome are 2:
    1) buff supports in order to make also them able to carry, for they are generally "subjected" to DD/flank performance (ie: you can be the best support in the world, but if your team is shit, you will lose. Ok, now you lose 5TP and not 10, but you are still losing)
    2) give a fixed bonus to support players, in order to compensate for that.
    I find the 2nd option shit, because the fact that you lose, as support, is not ALWAYS related to how bad allies you have, so many bad supports would be able to gain free TP for nothing.

    As usual, the only possible solution to give a fair game to everyone (you can't win without supports, but supports are unable to carry) is to buff supports.
    I've finally found it!

    New to the forum? Have a look >>>here<<<! Some useful answers.

    #StopPlayingSupports
    Let them see how a game with shit-nerfed supports is.

  7. #27
    Member Recruit
    Join Date
    May 2018
    Posts
    6
    Level
    1
    Points: 57, Level: 1
    Level completed: 14%, Points required for next Level: 43
    @EpplZ

    Now this is post with a lot of juice. You are actually spot on for the direction that I was trying to push the MMR/ELO calculation.

    Saying that there is a ton of data available for the game already is a understatement, so its a huge wasted opportunity to use them in such a grouped up fashion that you loose all that valuable information you could acquire. The biggest problem/ confusing point is, as you found out too, having a way to measure someones performance and I must admit your idea of splitting performances by talent card is quite good since talents are what make most people play a certain way. In general you make excellent remarks on how to fix the data input of MM and I really hope we see something along those lines implemented. However it is really tricky to find the right combination of fixed values in order to achieve the optimal outcome of MM.

  8. #28
    Banned Grand Prince
    Join Date
    Oct 2016
    Posts
    2,882
    Level
    98
    Points: 27,358, Level: 98
    Level completed: 1%, Points required for next Level: 992
    Achievements:
    Closed Beta Tester Overdrive Forum Tier 4 Forum Tier 2 1 year registered 25000 Experience Points
    Quote Originally Posted by Valkure View Post
    I'd like to point out another thing. Carrying capacity is not the same for all champs and no matter how good you can be, a Grohk is carrying less than a Drogoz. If you make that performance is lessening the loses/increasing the victories, but the majority of point are still related to W/L, the outcome are 2:
    1) buff supports in order to make also them able to carry, for they are generally "subjected" to DD/flank performance (ie: you can be the best support in the world, but if your team is shit, you will lose. Ok, now you lose 5TP and not 10, but you are still losing)
    2) give a fixed bonus to support players, in order to compensate for that.
    I find the 2nd option shit, because the fact that you lose, as support, is not ALWAYS related to how bad allies you have, so many bad supports would be able to gain free TP for nothing.

    As usual, the only possible solution to give a fair game to everyone (you can't win without supports, but supports are unable to carry) is to buff supports.
    The problem will fix itself with my method, if you play grohk you'll be judged based on Average (losing or winning) grohk with X LC in your mmr range, if the index is correct and your heals are weighted accordingly you basically CAN be the carry with grohk.

    PS : rest of my post via mp to stay on topic here.

  9. #29
    Member Recruit
    Join Date
    May 2018
    Posts
    6
    Level
    1
    Points: 57, Level: 1
    Level completed: 14%, Points required for next Level: 43
    What I'd like to add is that your performance should be judged based on the map that you play as well, because some maps are great for some champions while others no so much for example.

  10. #30
    Banned Grand Prince
    Join Date
    Oct 2016
    Posts
    2,882
    Level
    98
    Points: 27,358, Level: 98
    Level completed: 1%, Points required for next Level: 992
    Achievements:
    Closed Beta Tester Overdrive Forum Tier 4 Forum Tier 2 1 year registered 25000 Experience Points
    Quote Originally Posted by Orion532 View Post
    What I'd like to add is that your performance should be judged based on the map that you play as well, because some maps are great for some champions while others no so much for example.
    Yeah this can be implemented aswell, the best method to compare performance is to set the maximum of the parameters so you can really compare 2 players in an equal environment, the only problem with setting too many subgroups (ie : to set as many parameters as possible) is to keep the data set relevant, if the average of X champion with X lc on X map and in winning side is calculated from 5 matches top, your personal performance compared to that average doesnt mean much.

Tags for this Thread

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •