Ranking Systems: Elo, TrueSkill and Your Own

แชร์
ฝัง
  • เผยแพร่เมื่อ 2 พ.ย. 2024

ความคิดเห็น • 71

  • @looksintolasers
    @looksintolasers 4 ปีที่แล้ว +61

    It's extremely interesting how you compare players to maps and plop them on the same ranking scale. I wonder if this could be applied to students vs. problems in a math class, for example...

    • @Unelith
      @Unelith 3 ปีที่แล้ว +10

      And then include various tiers. A diamond calculus task, lol

    • @nelsonthekinger
      @nelsonthekinger 9 หลายเดือนก่อน

      Of course it can

  • @manuelcarpio-fl2pz
    @manuelcarpio-fl2pz ปีที่แล้ว +4

    😊for a moment I thought it'd be another boring lecture done wrong. But I'm actually very satisfied with the synthesis of the information.
    I'm currently designing a ranking system for the sport of BJJ, based on the ELO system. I had already started making modifications to account for the complexities present in grappling sports by adding a level of uncertainty to a particular grappling skill. Little did I know, this has been done and it's called Trueskill and Glicko. Your presentation made me feel that my intuition was on the right track.

    • @joshpck
      @joshpck ปีที่แล้ว

      interesting, I'm here researching on how I might do something like this too

    • @abcxyz-ld4fu
      @abcxyz-ld4fu 5 หลายเดือนก่อน

      How is your progress bro?. Im thinking same thing here.

    • @manuelcarpio-fl2pz
      @manuelcarpio-fl2pz 5 หลายเดือนก่อน

      @@abcxyz-ld4fu I finished it, did some sensitivity analysis and adjusted accordingly. Are you a math major too?

    • @GaurangPendharkar
      @GaurangPendharkar หลายเดือนก่อน +1

      @@manuelcarpio-fl2pz Hey bro, I would love to talk more about the Elo system you've implemented. I'm a math and cs major looking to implement such systems.

  • @도파민-m1w
    @도파민-m1w 5 ปีที่แล้ว +11

    Very nice explanation on rating system, it was good to listen

  • @reckoscort3044
    @reckoscort3044 4 ปีที่แล้ว +8

    Speaker has great experience and very confident.

  • @larfeeheelarfeehee9600
    @larfeeheelarfeehee9600 4 ปีที่แล้ว +24

    Gotta love his "millennials always need to win" and "gamers just don't understand the algorithms and that's why they think it's unfair" attitude.

    • @revimfadli4666
      @revimfadli4666 3 ปีที่แล้ว +6

      Even as someone who understands the point of the rating system(to keep matches interesting by having opponents at similar skill levels), it still hurts to see my effort being "invalidated" by the rating going up and down. Let alone those who don't understand. Perhaps a separate progressions reward would help to counter this effect

    • @RobFera
      @RobFera 3 ปีที่แล้ว +2

      @@revimfadli4666 Yep that's an excellent point!
      I agree that showing the player that their rank went down after a match, is like punishing them for playing. I guess there needs to be a reward for playing, no matter the outcome.

    • @tsaoh5572
      @tsaoh5572 ปีที่แล้ว +1

      @@revimfadli4666 That is exactly why league of legends let go of their ‘absolute’ elo rating, and introduced the different ‘tiers’ to reflect your rating. However, you still have a ‘hidden elo/mmr’ that actually determines how fast you’re progressing through these ranks.

    • @tsaoh5572
      @tsaoh5572 ปีที่แล้ว +4

      I think the frustration as a gamer doesn’t come from the ‘unfair rating system’ at all, but from how companies then use this elo score to trick you into playing more. For example, EA has patented an algorithm that basically adjusts your opponent’s ‘game strength’ based on your real-time elo/trueskill/glicko rating.
      Basically, the weaker player will always get a boost. Their ‘players’ in sports games will run faster, shoot more accurately, etc. Not extremely, but just enough to not make people notice.
      Whether you get a boost or not depends on many factors, but the core idea is that EA needs you to keep playing. Does a losing streak make you keep playing because you’re a ‘stubborn’ type of player? Then your opponents keep getting boosts when you’re on losing streaks. Does a winning streak make you keep playing? Then you are getting boosts when you’re on a winning streak.
      Of course, it is far more complicated than this… But my point is: The ratings are not wrong. The issue is how companies abuse it to make you addicted.

    • @3Torts
      @3Torts 3 หลายเดือนก่อน

      @@RobFera ...you're not "punished for playing" ...you just lost and losing is real.

  • @strykerhalo
    @strykerhalo 22 วันที่ผ่านมา

    “Rule of Play”
    Normal difficulty must be played or selected first before, “Real Play Ranking”is activated. After your 1st death on normal mode, “Hard difficulty” is then used to determine your skill level for, “True Play”.
    1st death on hard difficulty counts as true defeat in game. Continued play through deaths must be accrued, and tallied for total skill once game is completed on, “Hard mode”. Deathtracker determines your level of skill.
    Deathtracker
    0 deaths. Rank Grandmaster
    1 death. Rank Master
    2-4 deaths Rank Expert
    5-8 deaths. Rank Professional
    9-12 deaths. Rank Skilled
    12-18 deaths. Rank Amateur
    19-25 deaths. Rank Average
    26-35 deaths. Rank Beginner
    36-50 deaths. Rank Novice
    51-65 deaths. Rank Noob
    66-80 deaths. Rank Terrible
    81-99 deaths. Rank Trash
    100+ deaths. Rank Just Quit Restart

  • @hyeonggeunyun5694
    @hyeonggeunyun5694 3 ปีที่แล้ว +7

    6:11 can someone please explain further about how to integrate the normal distribution of two players and how this leads to the equation? Thx

    • @patrolin
      @patrolin 3 ปีที่แล้ว +1

      X ~ N(u_X, o_X^2)
      Y ~ N(u_Y, o_Y^2)
      P(X > Y) = P(X-Y > 0)
      X-Y = X + (-Y) = N(u_X-u_Y, o_X^2+o_Y^2) = Z
      en.wikipedia.org/wiki/Sum_of_normally_distributed_random_variables
      P(X-Y > 0) = 1 - P(X-Y < 0) = 1 - Z.cdf(0)
      en.wikipedia.org/wiki/Normal_distribution
      ...then you make weird assumptions about the standard deviations and you get ELO, probably.

    • @swizzamane8775
      @swizzamane8775 3 ปีที่แล้ว

      en.wikipedia.org/wiki/Elo_rating_system
      This will explain it better...

    • @patrolin
      @patrolin 3 ปีที่แล้ว

      @@swizzamane8775 this doesn't really explain how one arrives at the Elo system, moreover the page claims to use Logistic distributions, which i have absolutely no idea how to add :/

    • @RaineriHakkarainen
      @RaineriHakkarainen 3 ปีที่แล้ว

      Arpad Elo tought that The chess rating average is 1400 And standard deviation SD=282,8 ELO was WRONG! Jeff Sonas Stanford honor student was More WRONG! Jeff Sonas was using wrong Standard deviation SD=166 THE RIGHT numbers Are really The chess rating average is 1650 and standard deviation SD=256 These numbers You can easily calculate! The top 2% of The chess players reach 2176 rating! We need 100000 players that one of them Have 2720 rating! If we Have 2 billion players then The Best one gets 3237 rating! How SHOCKING IS THAT The Stanford university is ranked number 3th or 4th mathematics university in The world! Jeff Sonas Stanford honor student was WRONG! SD=256 NOT SD=166 COME ON!! Jeff Sonas!

    • @casperdewith
      @casperdewith 2 ปีที่แล้ว +3

      @@RaineriHakkarainen I don’t quite understand what you want to say, but it feels like you’re making a wrong assumption: as far as I know, Élő and Sonas didn’t make any assumptions on the ‘average’ chess score, but they gave new players a score (1000, I think) and then that score got updated as they play more matches. A score is not an absolute measure of skill, but it depends on other players’ skills.
      Could you elaborate your argument and provide sources? And could you limit the number of exclamation points to 1 per comment? Thank you in advance.

  • @TheGelatinousSnake
    @TheGelatinousSnake 3 ปีที่แล้ว +11

    Even if you aren't competitive... The more accurate a ranking system matches players, the better the experience.
    Casual players drop and rank and get matched against people performing on simular levels. Competive players will also be able to find more competent players that are actually trying to win.

    • @revimfadli4666
      @revimfadli4666 3 ปีที่แล้ว +3

      Although casual players often miss the point of matchmaking rating, and feel frustrated when the number fluctuates up and down(even though that's a sign of the system working as intended; placing them with others of similar skill levels). Even as someone who understands that, there's an innate psychological pain in having my efforts "invalidated" that way. Especially in games where one player's mistake can ruin the entire team

    • @THAC0MANIC
      @THAC0MANIC 2 ปีที่แล้ว +1

      except it legit dosnt matter on SO Many games ... Dota 2 / LoL been proven year in and out that ELO / MMR means jack shit, just take Support role in that game, TOTALY Dependant on other fucken players to raise your MMR coz its based PURELY on Win/Loss.

    • @TheGelatinousSnake
      @TheGelatinousSnake 2 ปีที่แล้ว

      @@THAC0MANIC so.. do you think that situation gets better with no rankings? How is your takeaway not “these games need a better ranking system”???

    • @THAC0MANIC
      @THAC0MANIC 2 ปีที่แล้ว

      @@TheGelatinousSnake im sorry your making it seem like I said something I did not, Thus you misunderstood what I said or your CLEARLY a troll one of the two.

    • @TheGelatinousSnake
      @TheGelatinousSnake 2 ปีที่แล้ว

      @@THAC0MANIC the basic jist of the first post “the more accurate the rankings… the better”
      And you open with “except”
      Misunderstanding I hope, because right at the start it sounds like you are trying to say less accurate rankings or no rankings are better than accurate rankings.

  • @Fanaro
    @Fanaro 6 หลายเดือนก่อน

    Awesome talk.

  • @chance2480
    @chance2480 3 ปีที่แล้ว +1

    Amazing video thank you

  • @mattd8725
    @mattd8725 5 ปีที่แล้ว +24

    The problem with skill rating is more complicated than "millennials cannot emotionally deal with losing". People like to think that they have some control when it comes to their own performance in a game. They do not like to think that they are being herded by an algorithm which seemingly decides in advance if you are going to win or lose. The enemy becomes the algorithm, more so than the level or opposing players. How do you convince players that they have some sort of control over their own "fate" in a game? They will take matters into their own hands and fight the algorithm by lobby dropping, smurfing, cheating, throwing, flaming teammates.

    • @natking1u1z99
      @natking1u1z99 4 ปีที่แล้ว

      You must be referring to league of legends 😂

    • @succubusnirriti913
      @succubusnirriti913 4 ปีที่แล้ว +3

      10y ago when I was 19, I already made myself a conspiracy based off game forcing 50/50 win ration. Not hard to come up with it (and this guy here showed just that on his powerpoint.....). At the time, I stopped playing LOL since I knew it was rigged. (they can give passive boost to TEAM B under the table to help them win, and keep playing longer (or buy skins). 50/50 means well why bother if im just GAMBLING at this point. Also the sexy-ness of the skins/art nerfed (to appease china) and other things just made me quit for good.

    • @revimfadli4666
      @revimfadli4666 3 ปีที่แล้ว

      Even as someone who understands the point of the matchmaking system, it still hurts to see my efforts "invalidated" by the rating fluctuating. Let alone those who don't get it.
      And when players get a losing streak after a winning streak(due to being placed a bit too high, naturally), they might think they're "robbed" of the winstreak they "deserve", that the systems doesn't want them to perform well

    • @mattd8725
      @mattd8725 3 ปีที่แล้ว +1

      @@revimfadli4666 The addition of the higher "meta" system may rationally tell you that the game of watching this number go up and down is "fair" and has rules, and so on, as it downgrades you. It does not tell you how your game to game experience is fair. The meta sorting game can be made to appear fair while the game to game experience is a ten game streak of boring easy wins followed by another ten game streak of hopeless unwinnable games. This is not a good thing just because it is rationalized and is in the service of the higher purpose of the meta game of skill rating! This is a philosophical point maybe, a Kantian perspective that the ends do not justify the means, and people are ends unto themselves.

    • @swizzamane8775
      @swizzamane8775 3 ปีที่แล้ว +1

      @@natking1u1z99 He's using LoL terms you only find in LoL with ones who complain SOOOOOO hard... When really, they just don't realize THEY ARE NOT AS GOOD AS THEY THINK. If one wants their "numbers" to increase, they NEED to play SMART, and NOT get greedy, then BLAME teammates for ONES own shortcomings.
      Hint hint: One is queued with others with like ratings. If ones "team" is failing, it is most likely one is failing also (possibly FEEDING into the failure state)
      Tbh, persons such as OP make me cry with laughter. Pathetic attempts to get attention and PERSONAL validation for their failures. (which is CLEAR this person fails, and fails often. and when they do.... boy will they BLAME ANYTHING but themselves xD) I USED to play LoL (till i found how pathetic and toxic the playerbase was [STILL is]), and I found this to be consistent. I'd see a dude DIVE a tower Solo, SPAMMING chat for support (even though is still early, and if ANY lane left, we'd lose it...) Kid dies, then UNLOADS in global bitching how teammates REFUSE to support, and how SELFISH they are for not pushing HIS tower... It was sad, but freakin hilarious. We lost, OFC, cause the kid just camped spawn most them time typing in chat, only displaying how inept his manhood truly is. Afterwards, EVERYONE spammed laugh and roasted the kid (whom STILL went on and on)

  • @LFSPharaoh
    @LFSPharaoh 4 ปีที่แล้ว +1

    13:25 this is why “booster” accounts work so well

    • @luciowins9946
      @luciowins9946 2 ปีที่แล้ว +1

      If a player gets a big win/loss streak it’s common to increase their σ since it’s clear they’re in the wrong place. That system rewards/punishes gameplay outside your expected rank so booster accounts should quickly punish the player.

  • @sykkms
    @sykkms 8 หลายเดือนก่อน

    learning how this algorithm collects and analyzes data is as important as the game. dismissing the human part of a competitive ladder is psychopathic and more than anything the algorithm is OCD and has cultivated these attributes in its players

  • @Drecon84
    @Drecon84 5 ปีที่แล้ว +3

    So this means that TrueSkill doesn't give you a rating, it gives you the chance that your rating is a certain value.

    • @revimfadli4666
      @revimfadli4666 3 ปีที่แล้ว +2

      Aren't ratings themselves such approximations, rather than being exact skill gauge?

    • @Drecon84
      @Drecon84 3 ปีที่แล้ว +2

      @@revimfadli4666 In an abstract sense, sure. The difference might be more semantic than anything else.
      The biggest reason for my comment is that TrueSkill uses statistical methods rather than a raw numbers system.

    • @swizzamane8775
      @swizzamane8775 3 ปีที่แล้ว +6

      The percentage values are actually just a function of a function, used to determine the weight value in the gain/loss factor after a win/lose state. And yes, Revi is correct in a more literal sense than you give them credit. As ALL (current) rating systems STILL use percentage values in factoring how many points to give or take away. It wouldn't be fair if everyone got the same value for change. So the percentage system helps establish success/failure prospects. Certainly more so than you seem to understand at time of your writing original message (which appears to be cursory or just from watching this vid, which both are inexcusable admissions of fault),
      "It is only "abstract" (that being pulled away or detached) because one does not understand it". If one wants more representable, or actual information, then one needs to expand their avenues of research

  • @surelock3221
    @surelock3221 5 ปีที่แล้ว +36

    What's Jesus doing here?

    • @bocatadeclavos1274
      @bocatadeclavos1274 5 ปีที่แล้ว +10

      Improving his english pronunciation

    • @metagen77
      @metagen77 5 ปีที่แล้ว +8

      Saving devs who strayed too far from god

    • @GunsLinger94SRB
      @GunsLinger94SRB 4 ปีที่แล้ว +10

      Explaining how he judges people.

  • @veggiet2009
    @veggiet2009 5 ปีที่แล้ว +2

    I really liked the idea of treating the random levels as a "player" ... But it makes me wonder how Mario Maker 2's systems work, particularly the online multiplayer works where 4 people are matched and a level is randomly chosen, but there's no [easy, normal, expert...] Code for multiplayer, but all of the levels do feel 'fair' to every player... So I wonder how that works, because if you have 4 players and 1 always wins then the level would always 'lose' is they used his system.

    • @Kenionatus
      @Kenionatus 3 ปีที่แล้ว +1

      I don't know Mario Maker, but off the top of my head, I'd probably treat each player as having their own match against the map for ranking purposes.

  • @microcosmos9143
    @microcosmos9143 5 ปีที่แล้ว +4

    It does not make it seem unfair - you're just failing at conveying the rating - you need service modelling consulting.

    • @Ronald-vz6pl
      @Ronald-vz6pl 2 ปีที่แล้ว +1

      Look at current reddit on halo infinite csr system 🤣

  • @Myst165
    @Myst165 5 ปีที่แล้ว +2

    good slime

  • @hassanalzaher3893
    @hassanalzaher3893 5 ปีที่แล้ว +1

    nice taLk

  • @th0rn3gaming
    @th0rn3gaming 9 หลายเดือนก่อน

    Explain why I win 15 in a row and lose 3 and de rank. 😂

  • @Padopoulosman
    @Padopoulosman 5 ปีที่แล้ว +2

    Tenemos un acento precioso UwU

    • @vikingo987
      @vikingo987 5 ปีที่แล้ว

      jajajajajaja, al principio estaba pensando "¿y este tío de dónde será con ese acento?" Cuando dijo que era español pensé "ay madre, si yo hablo igual o peor..."

  • @turkeyphant
    @turkeyphant ปีที่แล้ว +1

    How does this guy misunderstand so many of the questions?

  • @MeatBunFul
    @MeatBunFul 5 ปีที่แล้ว +17

    To many memes in the slides its hard to focus

    • @Lishtenbird
      @Lishtenbird 5 ปีที่แล้ว +11

      Are you not entertained?

  • @texteel
    @texteel 3 ปีที่แล้ว

    I do not know if that is a stage persona, his real persona, or just someone else wrote this for him, but almost every joke felt forced and none landed

    • @smort123
      @smort123 11 หลายเดือนก่อน

      OK