League of Legends ELO System

Discussion in 'Game Design' started by TheGodEmperor, Mar 14, 2012.

  1. TheGodEmperor

    TheGodEmperor Member

    In LoL at most you'll get to maybe 1600 with a support being your primary role (at least so my experience). It's not bad, but it's not going to get you to the "high ratings" people will really "respect."

    And yes, even pros will go down 300+ ratings and then slowly work their way back up. If you watch Guardsman Bob (not sure if I'd call him a pro gamer, even he won't, but he's certainly a good gamer and he is high rating consistently) you can see him getting trolled or getting just poor players (or that is players who make poor decision in his games) and has dropped several times back down to 1800 or even 1700 only to have to carry himself back up to 2100. The thing is, he's playing almost every day for a good 8-12 hours straight (it varies) and he'll do this and it takes him a few weeks to get back up.

    The way ELO works is that as long as you win more than you lose you'll go up, or, hell, even if it's even you'll still go up. You can grind your way to high ELO just by playing a lot if you have the time tbh far more than showing an accurate representation of your own personal skill. It ONLY accounts for wins/losses and I think that's a bit shortsighted. I think that's shortsighted for any game in all honesty.

    As for the surrender bit, there's only ever been once or twice I wanted to surrender and that's because it was not likely we were gonna win (still a very low chance), but more I had to go somewhere anyways and wanted to just leave to do something else. If I don't have plans or don't feel like doing something else, even if there's a slim chance I like to keep fighting and trying.

    And maybe it's because I never play a game unless I know I don't have something to do for awhile after. Probably because I like long games. Hell, as I said, I loved ToB in War 3. I think it was the best custom map to date (and it was better balanced more often than DoTA and was out before DoTA :| it just had a community of players who did NOT ever accept any new players and it wasn't as newb friendly). But those games were, on average, about an hour and half long :p but man, the epic back and forth battles. Helped I was a teenager at the time so less responsibility and more free time.
  2. Leartes

    Leartes Well-Known Member

    It is a commun misconception that players lose if the best player does not pick carry. I played so many games of dota where the players that were ranked highest picked supports or ganker and totally won me the game. Sometimes it shows in their K/D but sometimes they just win the game with their leadership skills.

    If you your win/lose turns worse on supports than on carries the only reason is that you are a worse support than a carry. I'd be interesting to compute different elo for different main heroes but sadly that is not possible.
    On the other hand your win/loss shows pretty well how good you are on a hero if you play a large amount of games on a similar level. I for example played over a hundred games as lion on the dotalicious platform and it turned out that my winrate with lion was about 45%. In the same time I played a similar amount of games as tidehunter and got a winrate of nearly 60%. My rating bounced around in the same area during the whole time. I can see from this that my lion play just sucks. I like the hero but I suck with him. It is not my teammates that are the problem it is MY play that is the problem.

    EDIT:
    ELO does not work that way! Maybe LoL elo is tuned badly but the concept of elo is mathematically proven to be the only feasible option.
    Also considering everything else apart win/loss is not possible in a truthful system.
  3. Fenrir

    Fenrir Well-Known Member

    These games are wildly unlikely. They just don't happen often enough to be significant. Mostly if you start to lose early you'll get crushed and that's it. Comebacks are very hard in this game. If you don't think a game can be carried by Taric support then I don't see how you think a game can be won if top lane feeds Trynd eleven kills by the 18th minute.

    If Riot want to remove surrender then, actually that's fine. But I also think that they should increase the shut-down bonus for ending someone's killing spree by alot. And increase the diminishing returns on killing the same idiot over and over. And hell, just make games shorter. And increase the passive gold gain like in dominion. (So you're never that far behind).
  4. Claytus

    Claytus Well-Known Member

    Have you even played Dota? This is so incredibly false. You are *more* at the mercy of your team when playing a support in that game. The extremes work in both directions. Carries are stronger in Dota, but supports are actually weaker in Dota as well. Dota supports tend to be more focused around abilities that primarily effect teammates, unlike, say, Sona's Q, or Taric's shield, where you can get some measure of survivability or damage if necessary.

    But, you know what... you go carry a bunch of games as Omniknight to prove your point, and I'll take you up, and try to do better with any particular LoL support you care to name...
  5. TheGodEmperor

    TheGodEmperor Member

    I'm not quite sure why Riot lowered the shut down bonus tbh from 1000 to 500. I kinda liked that it gave a bit more urgency to protect that player (encouraging a bit more team work to defend the fed carry) and it made killing them oh so more rewarding and help lessen snowballing effect.

    And ya, maybe Riot's Elo system isn't tuned well enough, but basically you can grind your way to a higher rating. I can't say for sure, but it sure seems it with how large a fluctuation I have seen people go.

    And mind, my gripe with their Elo system isn't me saying I'm a pro, far from it. I make really derpy mistakes and I acknowledge that at times. That is, my decision making sometimes is silly (like engaging when I shouldn't or staying further back than I should be), things I have been improving over time and are less and less frequent, but still happen. My problem is that I see so many claim that it is perfect and that's bs. There is no perfect system. And Elo (in LoL at least) doesn't seem to always represent how well you are, but more your luck in what people you get in a match up combined with how much time you can dedicate to playing consistently. It certainly can help show the better players from the worse, but it's not an end all be all and I hate that so many people seem to think of it as an end all be all system.
  6. Waterd103

    Waterd103 Well-Known Member

    I did most of my games in low elo with Nunu (tank support) , Shen (tank support) and Taric (Tank support).
    There has been people out there that already did it with champs like Soraka.

    I won't do it again, I dont even play SR anymore. But it has been done, players going to the top of the elo AGAIN and AGAIN, has been done.

    Now there is some truth in that you must be good at SOLO QUEUE (which requires some different skills than team queue). Is true that Reginald plays mostly carries, but he got all of his accounts to the top of hte ladder agian and again.

    There are certain champs that are better for solo Q? yes. So what? there is also champs that are better for team Q. So what's the point?.
    You can carry in LOL? yes, and in fact the LOWER the easier.

    You will get the occasional unwinnable game, but that is not that common. I hate when people talking about elo hell clings to their games where someone went 0-20 in 15 minutes. But overlook that game that just slipped by close (where a good player would push that smaller difference into their team favor)

    finally I don't know why you think is differnet in lol than HON the support thing. In Hon, you can be master support, but if your carry doesnt know how to farm, you will fail even more than in LOL.

    The concept that is easier to carry in HON than in LOL while I think it's true, it's only slightly so, but , I dont think it has to do with the Carries at all. It has to do with the ability in HON to have high mobility early game and have greater map impact. (which is closer to jungler rule in LOL, but the difference is that in LOL is harder for your ganks to count)
  7. Froh

    Froh New Member

    I did it in the first season. But yeah, it's pretty difficult at the beginning.

    And yeah, I don't like how the hypercarries can actually 1v5 in HoN/Dota. I don't like the fact that you become shitty as a mage at the end of the match.
    But I kinda prefer the ultra-powerful spells, the jungle tricks, the fog of war and the itemisation of Dota/HoN.
    But at first, if I chose to switch on LoL (and stay on it) it's because it evolves A LOT. I mean, they are changing a lot of things (the jungle, supports etc...), and they don't care about "stay with the same old way and never change a thing" like Dota 2 did. At least, HoN is pretty decent on this point.
  8. Fenrir

    Fenrir Well-Known Member

    TrueSkill is pretty close to perfect. Works for teams and single players. You gain and lose points based on the skill of your opponent and it tracks your skill in a neat way that results in very accurate match making.

    I pushed myself up to a TrueSkill rank of 48 (out of 50) in DoW2 and never felt like the match making was screwing me over.
  9. Waterd103

    Waterd103 Well-Known Member

    Mmm Xpecial plays mostly support and has 2258 rating. I would had more examples in the past but i don't play SR anymore.
    Also its not true that you can just grind elo, if you look at the ladder you will find a lot of players with hundred of games and not really high elo. Sadly we can't find whats below 1200 elo
  10. TheGodEmperor

    TheGodEmperor Member

    I should really learn more about TrueSkill hmm.

    And yes, Xpecial plays mostly support and is high rating, but (now correct me if I'm wrong because I don't watch his stream or know too much about him) he tends to duo queue and often the other is an AD carry ya? That kinda gives an advantage.

    Hell, I really think Duo Queue is bad for Solo Queue because it does give a pretty important advantage what with communication over vent or skype or what have you.

    Still, back to Xpecial, even then, he has about 600 games played on ranked since the second season started. In Season 1 he had something like 1.5k I think. That's a lot of time invested. Now, of course that investment made him a better player, but it also shows how Riot's Elo system even if you have a 1.0 w/l ratio you'll keep going up in rating. It won't be significant, but you will.
  11. FentonBailey

    FentonBailey Member

  12. Waterd103

    Waterd103 Well-Known Member

    look at Yarnbroder just to give an example 619/603 record and only 1204 ranking
    KiLL2YoU 464/457 1205 rating.
    Excelion Drive 390/362 games 1206 rating
  13. TheGodEmperor

    TheGodEmperor Member

    That's interesting. Atm Excelion's w/l ratio is similar to Xpecial's

    1.0773480662983425414364640883978 vs 1.1758957654723127035830618892508

    not that huge a difference. And yet the rating is so considerably different (1000). Wonder if Xpecial just had better first 10 games in the season (as those are worth like 40 rating each and then it slowly goes down) and Excelion had worse or what. hmm
  14. dereferenced

    dereferenced Active Member

    You're missing the point of ELO completely. The difference is Xpecial's opponents are much more skilled players than Excelion's opponents. If you put Xpecial against Excelion's typical opponents, Xpecial would crush them, while Excelion can barely stay even with them.
  15. Leartes

    Leartes Well-Known Member

    The difference is not the W/L the difference is against who you win and who you lose to. If you get 95% win until you are close to 2000 elo and then you get around 50% winrate over 600 games your w/l will only be slighty over 50%.
    On the other hand when you lose every game where you have elo > 1250 and win every single one with elo < 1150 you'll stay around 1200 forever and your win/loss will stay around 50% forever as well.
  16. TheGodEmperor

    TheGodEmperor Member

    You're assuming it's working as intended and there weren't people intentionally throwing the games being played.

    Again, the problem with Elo (imo) with LoL is that it really has a lot to do in your luck with solo queue partners and the team you're against. I don't think it's as accurate (at least in LoL) in reflecting your actual skill as others. Not that it's necessarily horrible, but I definitely think their solo queue could use some work. Again, for a 1v1 game, I'm sure it works great, but if you ever have checked the stats after the game you'll see at times the system will match people up poorly (as in I've seen 1800s with 1500s or lower even).

    Now, maybe it's not actually Elo that's the problem, but their leaver system, matchmaking system, and other systems around it that can effect Elo math itself. I dunno, only Riot employees really do.
  17. Claytus

    Claytus Well-Known Member

    ^^That's not how ELO works. The amount of points you earn for a win is high if the opponent had a higher rating than you, it's low if the opponents had a lower rating than you. It doesn't just arbitrarily make early games worth a lot, and later games worth less.

    What more likely happened, is that if you're right that the guy was duo-queueing, then the system tends to matchmake with opponents near the highest team members rating. He probably just got massive ELO gains early because his opponents were really high as a result of that, and they levelled off to normal as his ELO centered out. (Even if this isn't true, I think it also tends to match you with opponents farther from your current rating in early games, to help you quickly move towards the proper position)

    Keep in mind that the *entire point* of ELO is to enforce a 50/50 win-lose rating for all players. So, it's actually totally unsurprising and normal that they have about the same ratio, everyone has that same ratio. It just had to give them different ratings to be able to keep them each matched with opponents that allowed for that ratio to properly occur.

    Also note this is false, not just Riot employees "get it". The ELO system is a publically-available algorithm, and riot has been pretty open about explaining how they modified it to work for teams (although they haven't published their exact numbers), as well as explaining how their other systems work. Most of us have read all that information, and understand it. It really is working as intended, and that's further backed up by the tests described above done by WaterD and others, that pretty much proved it's doing what we thought it would.
  18. Leartes

    Leartes Well-Known Member

    How much does your elo vary over time? How long are your winning and losing streaks? If streaks are usually short and your elo does not bounce often beyond a confidence intervall of +/- 100 points then elo is working 100% as intended.

    Also this! If most players on all levels of elo have close to 50% winrate then elo is working correctly. Remember you lose as many games due to crappy teammates as you win games due to crappy opponents.
  19. Logo

    Logo Well-Known Member

    Does Lol determine the ELO adjustment based on your personal ELO score or the team's average or how does it work?

    HoN uses a fairly good system, though I couldn't find anything on it. Somehow for certain players (the ones on the extreme I think) it will adjust their stake downwards. So the highest rated player may be playing for +0/-0 if the teams are balanced, but he's way higher rated than the other teammates. I saw one post that said win rate is a factor of how much of your MMR is at stake, but I couldn't verify that.
  20. pkt-zer0

    pkt-zer0 Well-Known Member

    Addendum: it's named after Árpád Élő, the guy who came up with the system.

    Also, even if you wanted to match people (partially or wholly) based on playtime, you could do that without having playtime translate into in-game advantages (i.e. leveling up).
  21. TheGodEmperor

    TheGodEmperor Member

    I dunno, I seem to have bounced quite often a good 300 points and seen many others do the same on various streams. And the forums would lead you to believe this happens.

    And actually, now this is personal experience so take it with a grain of salt, having been matched with an 1800 friend against higher rating players and won, my rating went up no more than when I solo queue and it matches me with people of the same Elo. That is, both went up only maybe 12-13 points each. That's not really working as intended from what I understand about Elo. And this is more than a one time occasion.

    Hell, I don't think I'm the only one who has had problems with the matchmaking. Being matched with someone who has an average leave ratio of 1:5 when you yourself have never left a game or someone who might have been reported considerably for trolling/feeding/afking/etc when you haven't been reported once wouldn't show a very good matchmaking system to me. And then there's the whole nice level difference brought up early where the matchmaking will occasionally put 30s with 20s in normals even when those 20s aren't queued with anyone who's 30 so it shouldn't place them with them.

    There are most certainly still some problems :\

    I didn't know that was the guy's name :p

    And truth, you could use playtime which can be monitored. I never was defending the leveling system in here. I just eventually got to asking if, in the instance of where you're being queued with other people for a team game, people you don't know and likely never played with before, is it so bad that in the competitive mode to have a barrier of entry. I personally still think it can be done and would be a good thing for competition (aka more knowledgeable players), but could go horribly wrong too. Would be fun to look into more of.

    It's not like the game as a whole has a barrier of entry, just the portion that is "serious." Hell, if we were to use real sports, there are barriers of entry into professional sports.
  22. Claytus

    Claytus Well-Known Member

    LoL works it out sort of based on team average. Supposedly they weight towards the highest-rated player, though. (So, like if a 900 rated guy duo-queus with an 1800 rated queue, it will tend towards 1700-rated opponents, not the actual average of 1300) And there's some other rules where for each summoner level below 30, they just lop off X amount of points for the purposes of matchmaking, but I believe they've never publically said what X actually is.

    I believe LoL might also adjust for each player's curent rating when deciding +/- at the end, I have trouble remembering because I read up on it so long ago. I don't remember whether they keep it to actually be entirely zero-sum, if they allow for possible rating rating creep because of this. I think there's a bit of creep, and they never actually create a +0/-0 situation?

    This is actually super easy to explain, too. The matchmaking system prioritizes creating a game as quickly as possible over getting the perfect opponents. You probably even secretly appreciate that... the majority of players would rather wait less while they grind IP, then stare a loading screen longer. *Especially* when there's no guarantee that someone with the same rating as you, or a lower leaver score is even going to log in anytime soon.

    It just limits the amount of rating gained/loss in these situations to make up for the known poor match, as we described above. This is still all not a problem. The math works, and the system is functioning as intended.
  23. Aesa

    Aesa Well-Known Member

    What is more likely:

    1) Some guy overestimates himself and thinks he's always better than his teammate
    2) System that many games including Chess have been using for years is actually broken
    Remy77077, Waterd103 and dereferenced like this.
  24. TheGodEmperor

    TheGodEmperor Member


    I'm not really sure with the +/- bit.

    As for the matchmaking, when it comes to "serious" play (aka ranked) I think I'd prefer more the perfect opponents/team and a longer wait. Where as normals/bots I'd prefer as instantaneous as possible. That's my opinion on such at least.
  25. Logo

    Logo Well-Known Member

    Well following the idea of elo the highest rated guy should be something like +1/-23, but I think that rarely/never happens for HoN (I'll keep a better eye on it, I may be wrong) because they realize that it puts an unfair burden on that top rated player and would cause his elo to fluctuate wildly.

    I think overall it's 0-sum, it's just that the players closer to the average rating are the ones on the hook points wise. It makes sense, you rate the team using everyone's rating, but it's the average players in the mix that should be the ones on the hook performance wise rather than the 1 outlier guy. If the mmr is accurate and the teams are balanced, it's much more likely that the 4/5 of the team blew the game than the 1/5 guy that's 200 points higher so why would that 1 guy be the one losing 20 points.

    Again I'm not sure, but I want to say it's basically that points to gain is weighed towards the lower rated players while points to lose is weighed towards the players closest to the team's average. If everyone is close in rating on your team then you all more or less equally share the reward/risk. Even if that's not the case, maybe it should be! It makes a lot of sense, if your team wins despite your low rating then you probably aren't that bad. If a team loses despite your high rating it's probably your teammates fault.

    Overall I know my HoN mmr is pretty damn stable and only has a swing of about +/- 30 right now.

    (HoN also has a slider for longer wait/better match that you can adjust while sitting in queue).

    The variance HoN DOES have comes more from the lack of the 1-30 grind and ability to reset mmr. Bad players who return to the average ELO or new players who start at 1500 ELO, but are bad, cause volatility around 1500.
  26. Aesa

    Aesa Well-Known Member

    It happens in HoN if you play with friends who are not your rating that your point swing might be +1/-1 basically saying that this game doesn't mean anything. +1/-23 is absurd and should never happen in solo matchmaking in any 5v5 game, for obvious reasons.
  27. Aesa

    Aesa Well-Known Member

    This thread makes me /facepalm but I will address this. That is not a 300+ point swing. That is a 150 point swing. Furthermore, that is super standard variance for someone that plays as much as he does.

    The reason everyone thinks it's 300 is because if you are a guy who's top rating is 1600 but have ~300 rating downswings, what do you think your true skill is? 95% of LoL players will say their true skill is 1600. No. That's wrong. You are a 1450 player who fluctuates between 1300 and 1600.

    Sure, there are swings. No system can put you at the perfect ELO 100%. Even in a 1v1 game, you can have 150 rating swings, so stop blaming your teammates. Basically, this is how ELO works:

    Me playing at 800 ELO: 98% winrate
    Me playing at 1200 ELO: 90% winrate
    Me playing at 1600 ELO: 75% winrate
    Me playing at 2000 ELO: 50% winrate
    Me playing at 2400 ELO: 40% winrate

    And just in case that isn't clear, yes, that means someone with a trueskill of 1600 can drop to 1450 during a bad run. They can also rise to 1750 but no one complains about that. And sure, during EXTREMELY RARE CASES, they can drop to 1300, fine.
    Remy77077 likes this.
  28. TheGodEmperor

    TheGodEmperor Member

    When you see 300+ swings from several people on maybe a monthly basis (and in some cases the same people) you think there isn't a problem? And 2100->1800/1700->2100 again isn't a 300 point swing? I wasn't saying he went from 1700 to 1800 frequently. It's not good to have THAT amount of swing THAT frequently.

    That is, if I can clarify, he usually will stick around 2050-2100. But then he'll have (about once a month) a horrible losing spree, so bad he'll go down to 1800 or even 1700 range. And then he spends a week or so getting back up to 2000-2100 and stays there and then swings back down after a few weeks. How is that a good system? He more frequently stays up to the 2000-2100 range, but on a fairly consistent basis he'll fall down due to just bad games and have to play a lot to get it back up.

    Also, you don't have to double post, edit button :)

    LoL needs that slider imo. Sounds like a very nice thing to give your players.
  29. Aesa

    Aesa Well-Known Member

    Yea yea semantics. If you want to call it a 300 point swing, call it that, whatever, fine. The fact is it's a 150 point swing in both directions.

    Seriously, think about it. If GB's trueskill was 2100, do you really think that even though he plays all the time, he somehow ONLY gets unlucky enough to go down to 1800 but NEVER gets lucky enough to go up to 2400? That anyone could possibly believe that blows my mind. Downswings exist but upswings don't? Seriously where does all this magical disappearing ELO that people are losing go?

    Anyway I'm in a bad mood sorry for being mean.
    Remy77077, Ehrgeix and Waterd103 like this.
  30. Logo

    Logo Well-Known Member

    Yeah exactly what I mean. It's just not clear to me if Lol adopts this approach.

    (it has ramifications even in more reasonable settings, if you're even like 40-50 points up on the next highest guy you'll usually be risking something small like 7 points or so).
  31. Aesa

    Aesa Well-Known Member

    Yes, LoL does. Their systems may differ slightly but they are pretty much the same.
  32. TheGodEmperor

    TheGodEmperor Member

    If he's usually hovering around 2100 more often than he ever goes up or down, but he does have more downswings than up swings and they are that consistent, do you think it's a good system? Serious. I see him more frequently at 2000-2100 range (as in he'll be there for weeks) and then he'll drop down to 1800. Are you saying his real skill is 1950 even though he spends the majority of his time closer to 2100? You don't seem to factor time at all then. He plays a lot, he is more frequently closer to 2100 than he is 1900 and yet there'll be times (once a month or a month and half) he'll dip down to below 1900. I don't think it's fair to say his average is 1950 just cause 2100-1800 = 300/2 = 150 and ya... at least that's what it seems to be you're saying. That his swings are really 150 up and down and really his rating is 1950 (as it's in between the two extremes) yet the majority of his time is more toward the upper limit than the lower.

    And as I said, Riot's system is set up where if you lose, it's essentially on average a 10-11 point loss. When you win it's a 12-13 gain. What's this talk of "magical elo lost" when Elo in their system isn't a finite resource? :confused: Elo can be lost/gained. There's no maximum amount of Elo being distributed between people. Or maybe I mistook what you mean with that "Seriously where does all this magical disappearing ELO that people are losing go" line? hmm.

    And, as I've said, the system still doesn't account for leavers or people reported consistently for trolling/feeding/afking (and more importantly actually found guilty of it) and other aspects that DO matter in matchmaking.
  33. Aesa

    Aesa Well-Known Member

    I literally cannot comprehend how you think someone's who's trueskill is 2100 NEVER gets a lucky streak to 2250, yet somehow frequently gets unlucky streaks to 1800.

    By disappearing ELO I mean the fact that somehow hundreds of players have unlucky streaks yet 0 players ever have lucky streaks.
  34. TheGodEmperor

    TheGodEmperor Member

    I never said no one else has lucky streaks or he doesn't. As I've said, he's on average around 2000/2050 - 2100 (atm I think he's 2150 and gaining a bit, so maybe he's going on his upward streak now, but who knows) and he'll stay around that range frequently but every month or month and half drops pretty far down only to have to build his way back up. Far more than any "well honed Elo" will allow in fluctuation since, as many here have said, +/- 100 is a good balanced system. Yet he and others will fluctuate considerably more. And yes, I mean they'll go from say 1700 down to 1100. I've seen that too from other streamers. It's not like Bob's the only example with some consistent massive fluctuations.

    Even so, I'm not sure if it's really a problem with Elo specifically or matchmaking or maybe just enough of a calculation error on both that it's causing such problems for people. And no, not myself cause, as I've said, I don't bother with ranked anymore. I just don't feel like it.
  35. Aesa

    Aesa Well-Known Member

    OK, let me ask you this.

    Theoretically, let's say he really is 2100. Why do you think he has unlucky streaks to 1800 DRASTICALLY more often than he has lucky streaks to 2400?
  36. Logo

    Logo Well-Known Member

    To be fair to TGE ELO is a bell curve as we know. Someone who is near the end of that curve should have swings downwards of greater severity than upwards because you end up playing more games vs people of a lower rating. So if you win 5 in a row it's likely going to be games that offer you little points (some high % of your matches are vs lower ELOs). If you lose 5 in a row you're going to dip a lot since most of your games are vs people lower rated than you.

    So as an example it might be possible to have a 'real' rating of 2300 with a swing of 2150-2350 because your win streaks offer less points than your loss streaks lose.

    In this example he may not be high enough to hit this limitation, but it is something that can/will happen at the edges of the bell curve.

    Yet at the same time against what TGE is saying, this isn't a problem really. It's just a nature of playing in an ELO system at the far ends. It also has nothing to do with anyone who's near the center of the bell curve.
    Fenrir and Aesa like this.
  37. dereferenced

    dereferenced Active Member

    Re: Swings:
    Do you really expect a player's performance to be consistent day to day, especially given the vast amount of problem space in LoL?

    Maybe I play with a skill level of N baseline, but I dip to N-100 when I get an hour less sleep a day, had an argument with the gf, stressful day at work, etc, or I peak to N+100 when the enemy team plays a champ that I'm really good at exploiting. There's a million other factors that affect your day to day performance. The entire idea that 'my skill/elo is fixed and shouldn't be swingy' is wrong.
  38. TheGodEmperor

    TheGodEmperor Member

    No one's saying it shouldn't swing, but it shouldn't swing so vastly so frequently is the point. At least if it's a well honed Elo system as I'm lead to believe (from this thread and my own readings on Elo) the swings shouldn't be as great as many have. There's a problem that people do keep forgetting. Elo was designed for Chess. Chess and LoL are vastly different games. There are more variables in LoL than Chess.
  39. Aesa

    Aesa Well-Known Member

    @Logo

    GB isn't there, even if he is, there are people in the middle who complain about this all the time happening to them. TGE even mentioned some guy at 1700 in his most recent post for example.
    Logo likes this.
  40. TheGodEmperor

    TheGodEmperor Member

    The 1100-1700 person who is more frequently 1550-1600 from what I can tell (more often 1650 since they've been duo queueing with another good player who's about 1800-1900 I think). Are you saying their real Elo is 1400? Since that's in between 1100 and 1700? Or are you saying that the 1550-1600 they were at on average is right? But then why is a system where they can have a downward swing as great as 500 and an upward swing only so great as 150 (or 200 now) a good balanced system? I dunno. Just seems like there's SOMETHING skewing things (again, be it their own tweaks to Elo or their matchmaking or something else or a combination of various things, etc) a bit in the game and when it comes to rating. I really don't think it's "perfect" like you seem to be getting at.
  41. Aesa

    Aesa Well-Known Member

    I'm saying, what's more likely:

    1) Somehow people in the middle of the bell curve have larger downswings than upswings (???)
    2) It is human nature for people to complain and think more highly of themselves than they really are
  42. TheGodEmperor

    TheGodEmperor Member

    Well humans created the system, humans err, what makes you think the system is perfect? Especially when this system was designed for an entirely different style of gameplay originally.
  43. Claytus

    Claytus Well-Known Member

    Also realize, there is no such thing as someone being a "1600 player". The ELO number is a rating, it measures your current rank against other players.

    If memory serves, being exactly at 1600 roughly means "better than 90% of the playerbase". If the playerbase itself changes, you can go up or down, and still be ranked completely accurately. These numbers are not some universal truth about a person, as we said above... they are an arbitrary number that happens to create a roughly 50/50 win-lose record.

    Note that if 1600 is the 10% mark... imagine what the 1900-2100 rating fluctuation means. These are insanely high ratings that are already sticking someone in like the top 0.1% of players or something. For someone to move seemingly move up and down a lot because the system doesn't know exactly who's the best among that small a group at the extreme end of the scale means basically nothing at all.

    Like, seriously, who frickin' cares??? If you're that good, you go show up tournaments and try to earn some money. The numbers are still accurate, and they're still creating a good matchmaking experience overall, and even people that high are getting good practice in against other pretty much equally skilled players. The whole thing is a moot point if you're hovering closer to 1300, because there's waaaaay more people with about the same rating, and it's much harder to fluctuate by 100s of points on a regular basis. I can say that much from personal experience.


    I can also say from personal experience that the playerbase is notably different at different times. Play LoL on a saturday morning, and being 1400 kinda means nothing... there are thousands of terrible people online to crush. Hit 1400 on a Sunday afternoon, and I'm impressed... it tends to be a smaller pool of players with some decent skills. (I'm not even very good, and I easily noticed this, so I'm guessing this should be obvious to everyone else, since you all claim to have a higher rating than me?) Again, it's all relative, these numbers are just a rating of how you compare to other people playing at the moment.
  44. TheGodEmperor

    TheGodEmperor Member

    Actually I think my rating is about 1400 or so similar to yours. Again, I haven't bothered with ranked in a long time. I imagine I think I can get up to 1600ish and probably maintain it if I play more and improve my decision making (aka a little less aggressive, a bit more passive early game). I dunno if I really have the patience/time to hone my skills to be significantly higher. It's not that huge of a deal for me.

    I just wish the matchmaking more than anything took a bit more into consideration. My 0% leaves/bans shouldn't be teamed up with a person who's been banned/suspended 3 times and has such a horrible leave ratio.

    And considering how much I've seen people I watch fluctuate and all the threads on the forum it would seem there could be a problem with the system in some way.
  45. Waterd103

    Waterd103 Well-Known Member

    There may be problem but really all this fluctuations are EXPECTED, and ok. I think you really don't get the Math behind it that AESA pointed out.

    It would be more helpfull if this guy would record his elo after every game, and then analyze a graph after 100 games.
    But i guess that's too much work...

    Also guardsman bob is a high level troll, so no surprised at all he has more variance than most players.
  46. TheGodEmperor

    TheGodEmperor Member

    High level troll? Oh come now. Do you actually watch Bob? He's prolly one of the more well mannered and good players up there. Not the best (he clearly makes mistakes and points them out himself), but ya... don't see why you'd think a troll.

    And, as I've read, fluctuations of THOSE levels (as in well over 100 points) seems a bit... bad for the system.
  47. Xom

    Xom Well-Known Member

    I'm not really familiar with Guardsman Bob, but I happened to open his stream once, and during the game he claimed to be eating oatmeal. Assuming he doesn't eat oatmeal during 100% of his games, phoning it in like that probably increases his variance.
  48. dereferenced

    dereferenced Active Member

    This is a troll thread
    Waterd103, Polari and Aesa like this.
  49. TheGodEmperor

    TheGodEmperor Member

    Honestly, don't come in trolling the thread or posting pointless crap by claiming I'm a troll when I'm quite serious. There's certainly a problem with Riot's system. As I've said, I'm not sure if it's a huge problem, but there definitely is one. No system is perfect though.

    And even Riot has said that as long as you keep playing, if you are at least winning the same amount as losing you'll still be going up in rating even if small at a time.
  50. acidile

    acidile Member

    I'm assuming this is a troll thread, anyone with half a brain knows that the Elo system works, most people just think they're like 500 ELO better than they actually are and get mad and stuff.

    On the playing as support in solo q thing, it's something that I think about a lot. The best support players in the world sit on the first couple pages of ladder with a ~60% winrate on their main heros. In high Elo matchups, this is a really really respectable winrate, comparable to most of the top ap/top/jungle players.

    However, I don't think support attains nearly as high of a winrate when you are way better than people. In low Elo, you can attain sicko (90%+) winrates playing hard carries, but still probably 60-70%ish playing support. Because of this, playing support will see more variance. You will still go up, but it will likely take longer.
    Waterd103 likes this.

Share This Page