|
On September 10 2015 03:24 Frolossus wrote:Show nested quote +On September 10 2015 02:58 Goumindong wrote:On September 09 2015 23:27 Ansibled wrote:On September 09 2015 09:39 Sufficiency wrote:On September 09 2015 09:10 Ketara wrote:
And then if you keep losing games during the promo series and winning games outside of them you stay the same division while your mmr goes up and its super frustrating, and then when you finally win a series you skip a division or start getting 30 LP on a win or whatever which feels hollow and makes you disillusioned with the system.
I have played thousands of league games sufficiency. I know how the system makes me feel. Don't be that guy. It's just your anecdote. The fact of the matter is that it works. A Gold I player from Season 4 takes significantly fewer games to return to Gold than a Gold V player, for example. ![[image loading]](http://i.imgur.com/4vKLCyt.png) To me this graph is a pretty good indicator of frustration, that's a lot of games for some people. It's more than I play in a season, but I never really have issues getting gold for free skins in the 10-20 games I play either so. That is true. But the vast majority of those gold 5 players probably are not actually gold 5 by "centered" MMR but Silver 1 by "centered" MMR. I mean, after 5 to 6 promotions its highly likely that someone who isn't actually gold 5 but rather the middle of Silver 1 will promote. If they play enough games they will get there by random chance. that's because of elo fluctuation right? it's not like the player is silver 1 or gold 5, it'd be more like they are roughly ~1500 elo when the cutoff was 1520 or whatever it used to be. this means that if they pass 1520 then fall back down to 1500 they get placed into g5 because of the cutoff but they are still being matched with and against the same group of players.
Short answer: Yes, its because of ELO fluctuation and the strict cutoff point.
Long Answer: Assume that a player has a static skill level for a minute. If that is the case then MMR is a recurrent centered random walk. It has an expectation of your true skill and will tend to return to that value. Another way to call it is a stationary stochastic process. *
In order for a player to go up in level there must exist a segment of the process which is above the cutoff value for a certain length of time. Given an infinite amount of time, any person centered below the cutoff will be above the cutoff for the requisite time.
What it means is that they might really be 1400 players, but they only have to get 1500 for 3 or 4 games in a row in order to make Gold. So they played a lot and made gold last season. And either didn't play enough to breach the demotion cutoff. They didn't get better or worse over the season they just stayed the same in relative position.
We can see that easily by looking at tier stats
http://na.op.gg/statistics/tier/
We notice that the whole distribution is centered at Silver 3. The difference between Silver 3 and the top of Bronze 5 is about the same difference between Silver 3 a Platinum II. To give a scale of what we're talking about.
That the differences between ranks is most smooth between 4 and 3. And then there is a larger jump to 2 and 1 and then a massive break at 5. (the exception to this being Gold 1-Plat 5 where it seems a lot of people have given up on making plat?)
Anyway what we "should" see if it was just an MMR thing is smooth differences everywhere. Starting small then increasing then decreasing again. While an individual persons MMR is pretty random, the whole of the MMR distribution is very stable. But instead we see dips at 1 and 2 and a glut at 5. According to their current MMR and/or what their stationary MMR ought to be (these might not be the same people but they have the same expected value if that makes sense) there are about 40,000-50,000 people in Gold 5 who ought to be in gold 1 or gold 2.
At the low end this means that roughly 30% of the people in Gold 5 really aren't and so now we can look at that "getting back to gold" chart and see some more context. Most of the people who are at the bottom end of that spectrum and are taking a long time to get gold, probably shouldn't be there.
*its important to note that skill probably doesn't work like that. However, that does not change the overall distribution and so the numbers and theory still hold. That is, since rank is a relative placement of people and not an absolute value of skill we can still come to the conclusions we came to below, even though we know that our model of an individuals MMR is "incorrect". We can model it more accurately but doing so isn't particularly valuable for a number of reasons.
|
|
|
Being at 0LP is a demotion series in itself. Unless you hit it after a decent loss streak, you can usually get by on a few more games before it boots you.
|
On September 10 2015 03:39 VayneAuthority wrote:
this isnt even factoring in things like the fact that riot constantly patches. I've seen single broken champions that are under the radar boost gold players into diamond, sometimes skill level is thrown out in the window due to champions.
Thats why a lot of the times you will be like wtf how is that guy in this tier?
then you check his lolking or whatever and he has 80% winrate with soraka and then under 50% with everything else.
Gold players in Diamond just bc of broken champs? Isn't that a bit exaggerated lol, I rarely see any players in Plat who were Gold last season. And the Soraka guy might just be really good at his champion, besides she's a champion that I can see low elo players struggling to play against (focus raka, not the carry if you can)
That the differences between ranks is most smooth between 4 and 3. And then there is a larger jump to 2 and 1 and then a massive break at 5. (the exception to this being Gold 1-Plat 5 where it seems a lot of people have given up on making plat?)
Don't quite understand what you're saying, there are more Plat 5 players than Gold 1. The small difference can be attributed to many things, such as the possibility that the MMR differences between G1-P5 doesn't have to be the same as S1-G5, B1-S5, or in other words, the MMR range of G1 is very large. Also possible is it's simply an MMR range that many people 'peak' at.
there are about 40,000-50,000 people in Gold 5 who ought to be in gold 1 or gold 2.
You mean G5 vs S1/2, right? Again, you're making the assumption that the rank system has the same MMR range for every division, and we don't know that. I can totally imagine that Riot fine tuned the system so that every tier has a gradual decline in player population as you go up divisions to give people a sense of accomplishment.
Personally I think what you say holds some truth to it (I think we've all checked op.gg during a game and realized that we're playing only with ppl in divisions below us bc of our big losing streak), but this is all a game of guessing Riot's MMR/LP model.
|
On September 10 2015 05:58 DarkCore wrote:Show nested quote +On September 10 2015 03:39 VayneAuthority wrote:
this isnt even factoring in things like the fact that riot constantly patches. I've seen single broken champions that are under the radar boost gold players into diamond, sometimes skill level is thrown out in the window due to champions.
Thats why a lot of the times you will be like wtf how is that guy in this tier?
then you check his lolking or whatever and he has 80% winrate with soraka and then under 50% with everything else. Gold players in Diamond just bc of broken champs? Isn't that a bit exaggerated lol, I rarely see any players in Plat who were Gold last season. And the Soraka guy might just be really good at his champion, besides she's a champion that I can see low elo players struggling to play against (focus raka, not the carry if you can) Show nested quote +That the differences between ranks is most smooth between 4 and 3. And then there is a larger jump to 2 and 1 and then a massive break at 5. (the exception to this being Gold 1-Plat 5 where it seems a lot of people have given up on making plat?) Don't quite understand what you're saying, there are more Plat 5 players than Gold 1. The small difference can be attributed to many things, such as the possibility that the MMR differences between G1-P5 doesn't have to be the same as S1-G5, B1-S5, or in other words, the MMR range of G1 is very large. Also possible is it's simply an MMR range that many people 'peak' at. Show nested quote +there are about 40,000-50,000 people in Gold 5 who ought to be in gold 1 or gold 2. You mean G5 vs S1/2, right? Again, you're making the assumption that the rank system has the same MMR range for every division, and we don't know that. I can totally imagine that Riot fine tuned the system so that every tier has a gradual decline in player population as you go up divisions to give people a sense of accomplishment. Personally I think what you say holds some truth to it (I think we've all checked op.gg during a game and realized that we're playing only with ppl in divisions below us bc of our big losing streak), but this is all a game of guessing Riot's MMR/LP model.
Back when they changed over the ranked system to hide elo, each tier was 350 points, each division being 70pts.
|
On September 10 2015 05:58 DarkCore wrote:Show nested quote +On September 10 2015 03:39 VayneAuthority wrote:
this isnt even factoring in things like the fact that riot constantly patches. I've seen single broken champions that are under the radar boost gold players into diamond, sometimes skill level is thrown out in the window due to champions.
Thats why a lot of the times you will be like wtf how is that guy in this tier?
then you check his lolking or whatever and he has 80% winrate with soraka and then under 50% with everything else. Gold players in Diamond just bc of broken champs? Isn't that a bit exaggerated lol, I rarely see any players in Plat who were Gold last season. And the Soraka guy might just be really good at his champion, besides she's a champion that I can see low elo players struggling to play against (focus raka, not the carry if you can) Show nested quote +That the differences between ranks is most smooth between 4 and 3. And then there is a larger jump to 2 and 1 and then a massive break at 5. (the exception to this being Gold 1-Plat 5 where it seems a lot of people have given up on making plat?) Don't quite understand what you're saying, there are more Plat 5 players than Gold 1. The small difference can be attributed to many things, such as the possibility that the MMR differences between G1-P5 doesn't have to be the same as S1-G5, B1-S5, or in other words, the MMR range of G1 is very large. Also possible is it's simply an MMR range that many people 'peak' at. Show nested quote +there are about 40,000-50,000 people in Gold 5 who ought to be in gold 1 or gold 2. You mean G5 vs S1/2, right? Again, you're making the assumption that the rank system has the same MMR range for every division, and we don't know that. I can totally imagine that Riot fine tuned the system so that every tier has a gradual decline in player population as you go up divisions to give people a sense of accomplishment. Personally I think what you say holds some truth to it (I think we've all checked op.gg during a game and realized that we're playing only with ppl in divisions below us bc of our big losing streak), but this is all a game of guessing Riot's MMR/LP model.
i think the tiers are just bottom heavy because it's easier to get promoted than demoted
|
On September 10 2015 05:58 DarkCore wrote:Show nested quote +That the differences between ranks is most smooth between 4 and 3. And then there is a larger jump to 2 and 1 and then a massive break at 5. (the exception to this being Gold 1-Plat 5 where it seems a lot of people have given up on making plat?) Don't quite understand what you're saying, there are more Plat 5 players than Gold 1. The small difference can be attributed to many things, such as the possibility that the MMR differences between G1-P5 doesn't have to be the same as S1-G5, B1-S5, or in other words, the MMR range of G1 is very large. Also possible is it's simply an MMR range that many people 'peak' at.
The difference between Plat 1 and Diamond 5 and Silver 1 and Gold 5 is much larger than the difference between Gold 1 and Plat 5. There are more plat 5 players than Gold 1, but since its much easier to be promoted than demoted(and since people are likely to delay if they think they're on the cusp of being demoted) we expect that there are going to be many more plat 5 players than gold 1 players, like there are many more gold 5 than silver 1 and diamond 5 than plat 1. Its strange that we don't see it, but doesn't really change the conclusions we should come to about Sufficiency's graph.
MMR is a continuum. So the difference between plat 5 and gold 1 and diamond 5 and plat 1 are pretty similar.
There isn't any reason that the MMR range of Gold 1 would be very high but none of the others would be. (the clamping could be tighter though)
Show nested quote +there are about 40,000-50,000 people in Gold 5 who ought to be in gold 1 or gold 2. You mean G5 vs S1/2, right? Again, you're making the assumption that the rank system has the same MMR range for every division, and we don't know that. I can totally imagine that Riot fine tuned the system so that every tier has a gradual decline in player population as you go up divisions to give people a sense of accomplishment. Personally I think what you say holds some truth to it (I think we've all checked op.gg during a game and realized that we're playing only with ppl in divisions below us bc of our big losing streak), but this is all a game of guessing Riot's MMR/LP model.
Yes I do mean G5 and S1/2. I am not making the assumption that the rank system as the same MMR range for every division, simply that the MMR range is relatively consistent. (after all, we can see that the MMR range from Silver 3 to the top of bronze 5 is about the same MMR range from Silver 3 to the top of Plat 2. I expect then that, if anything, the MMR range gets tighter as you get higher simply because we know there are more ranks between the center and the top(17 not including challenger/master) and the center and bottom(6)
The system naturally will have less population in each division unless the range of MMR increased massively as you went up the ladder. This is because the ELO system forces a normal distribution(iirc). Its unlikely that MMR ranges expand like that.
|
|
|
YO
So how shitty is Azir this patch?
|
|
|
Still op if you can play him
|
Azir and Kalista: epic failures of this season.
|
On September 10 2015 11:25 Sufficiency wrote:
Azir and Kalista: epic failures of this season.
Don't be too quick to judge, there is still time for Morde and friends.
|
On September 10 2015 10:58 ticklishmusic wrote:
YO
So how shitty is Azir this patch?
still broken if the guy actually knows how to play not feeling any differences
|
On September 10 2015 11:28 cLutZ wrote:Show nested quote +On September 10 2015 11:25 Sufficiency wrote:
Azir and Kalista: epic failures of this season. Don't be too quick to judge, there is still time for Morde and friends.
Fair point. Overall I strongly dislike all the decisions Riot has made this year. The worst thing you can do to piss off existing players is to create grossly problematic champions that breaks the meta of t he game, especially when the game is likely on its decline.
|
I don't mind Azir. There's a huge payoff but his skillcap is ridiculously high and champs like that rarely become consistently frustrating for me. I hate Kallista but at least she's been dialled back from the obscenely broken pos she was in the beginning to just being regular broken.
I hate the Morde rework though. What I hate most is that facing him feels just like facing old Morde when he was strong, with the caveat that he's sometimes in your botlane. He either gets fed and blows you away while being unkillable or doesn't get ahead and does nothing. Stuff like being overlevelled in botlane and "lol free dragon without using ult" are just extra kicks to the groin.
|
On September 10 2015 12:58 Amarok wrote:
I don't mind Azir. There's a huge payoff but his skillcap is ridiculously high and champs like that rarely become consistently frustrating for me. I hate Kallista but at least she's been dialled back from the obscenely broken pos she was in the beginning to just being regular broken.
I hate the Morde rework though. What I hate most is that facing him feels just like facing old Morde when he was strong, with the caveat that he's sometimes in your botlane. He either gets fed and blows you away while being unkillable or doesn't get ahead and does nothing. Stuff like being overlevelled in botlane and "lol free dragon without using ult" are just extra kicks to the groin.
I don't mind Azir or Kalista for my own games, but find them deplorable for pro play, particularly Kalista. These are champions that simply make me want to turn off a game unless it is particularly meaningful. Although Azir requires "high skill" its not actually the kind of skill that any pro lacks, he requires good mechanics, which all relevant mids have. For a pro he is basically Orianna with higher DPS and less decisionmaking. Kalista is just a lane bully like S2 Darius without the immobility and with a much smaller lategame dropoff.
Also, juggernauts, yes, just stupid for pros and humans.
|
Despite how strong Kalista and Azir were, they were never at a pick/ban level of frustration. Sometimes people would pick them and destroy the game. Most of the time they didn't. Their winrates weren't astronomic either, so it's never really felt like they were too OP in solo q.
Overall, I just don't like the idea of Mord being a substitute ADC, I can live with it though.
|
On September 10 2015 14:44 cLutZ wrote:Show nested quote +On September 10 2015 12:58 Amarok wrote:
I don't mind Azir. There's a huge payoff but his skillcap is ridiculously high and champs like that rarely become consistently frustrating for me. I hate Kallista but at least she's been dialled back from the obscenely broken pos she was in the beginning to just being regular broken.
I hate the Morde rework though. What I hate most is that facing him feels just like facing old Morde when he was strong, with the caveat that he's sometimes in your botlane. He either gets fed and blows you away while being unkillable or doesn't get ahead and does nothing. Stuff like being overlevelled in botlane and "lol free dragon without using ult" are just extra kicks to the groin. I don't mind Azir or Kalista for my own games, but find them deplorable for pro play, particularly Kalista. These are champions that simply make me want to turn off a game unless it is particularly meaningful. Although Azir requires "high skill" its not actually the kind of skill that any pro lacks, he requires good mechanics, which all relevant mids have. For a pro he is basically Orianna with higher DPS and less decisionmaking. Kalista is just a lane bully like S2 Darius without the immobility and with a much smaller lategame dropoff. Also, juggernauts, yes, just stupid for pros and humans.
You mean s3 draven  . Remember picking him up and just legit walking behind the creep wave and trading 1v2 lol.
Honestly azir really isn't all that hard I have no fucking idea how his winrate is that low. Guess I overestimate soloq players by a lot.
|
S3 Draven (For Pros) at least could be dodged or poked.
On September 10 2015 14:59 Sonnington wrote:
Despite how strong Kalista and Azir were, they were never at a pick/ban level of frustration. Sometimes people would pick them and destroy the game. Most of the time they didn't. Their winrates weren't astronomic either, so it's never really felt like they were too OP in solo q.
Overall, I just don't like the idea of Mord being a substitute ADC, I can live with it though.
Plebs who care about soloQ...
|
|
|
|
|
|
EPT
