|
On March 25 2016 10:36 Frolossus wrote:
there is no case of how you play solo vs how you play in a team in regards to how mmr works. it simply measures how well you perform in the game. playing in a group does have certain advantages in a game. but it has so many disadvantages for climbing. mainly that if you don't belong in the elo you are playing at, you are a burden to the rest of your team regardless of whether or not you are playing solo or in a group. your mmr reflects this fact either way. if you are in a team then your whole team's mmr will be affected by it.
not every player has the same reasons they are stuck at their mmr. 5 solo players at 1700 vs a team of players who collectively got to 1700 should have a roughly equal chance to win. if it didn't think so then one side or the other would be rated above or below 1700.
This is nonsense. MMR does not measure how well you perform in a game. MMR measures whether or not you win. Playing in a group increases your chance of winning for a number of reasons. Because it increases your chance of winning it must necessarily have advantages for climbing the ranks UNLESS its the case that group queues only and always only play other group queues*. They do not.
The question isn't "does a group that is collectively 1700 have a >50% chance to win against individuals that are 1700?" The answer to that question is no. But its immaterial.
The question is "does a group that is individually 1700 have a >50% chance to win against individuals that are 1700?" and the answer to that question is yes. Similar is the question "does an individual who is 1700 in a group but not in a group have a <50% win rate against individuals who are individually 1700?" And the answer to that question is also yes.
Such that "if you are individually 1700, you can go up in MMR by getting into a group". And "if you are in a group at 1700 when playing soloqueue you will more likely be a detriment to your team"
*technically the percentage of games has to approach zero but its close enough.
|
On March 25 2016 10:44 Goumindong wrote:Show nested quote +On March 25 2016 10:36 Frolossus wrote:
there is no case of how you play solo vs how you play in a team in regards to how mmr works. it simply measures how well you perform in the game. playing in a group does have certain advantages in a game. but it has so many disadvantages for climbing. mainly that if you don't belong in the elo you are playing at, you are a burden to the rest of your team regardless of whether or not you are playing solo or in a group. your mmr reflects this fact either way. if you are in a team then your whole team's mmr will be affected by it.
not every player has the same reasons they are stuck at their mmr. 5 solo players at 1700 vs a team of players who collectively got to 1700 should have a roughly equal chance to win. if it didn't think so then one side or the other would be rated above or below 1700. This is nonsense. MMR does not measure how well you perform in a game. MMR measures whether or not you win. Playing in a group increases your chance of winning for a number of reasons. Because it increases your chance of winning it must necessarily have advantages for climbing the ranks UNLESS its the case that group queues only and always only play other group queues*. They do not. The question isn't "does a group that is collectively 1700 have a >50% chance to win against individuals that are 1700?" The answer to that question is no. But its immaterial. The question is "does a group that is individually 1700 have a >50% chance to win against individuals that are 1700?" and the answer to that question is yes. Similar is the question "does an individual who is 1700 in a group but not in a group have a <50% win rate against individuals who are individually 1700?" And the answer to that question is also yes. Such that "if you are individually 1700, you can go up in MMR by getting into a group". And "if you are in a group at 1700 when playing soloqueue you will more likely be a detriment to your team" *technically the percentage of games has to approach zero but its close enough.
mmr is calculated by how often you win. but the thing that it is trying to measure is your ability to play the game relative to every other individual who plays. playing together as 5 does not somehow make you better at the game. it just means that your mmr is now accounting for more factors that contribute to game outcomes associated with playing on a team.
if you are individually 1700, and everyone in your party would play at 1900 aside from you had they been queuing solo for a long time.
the result is that everyone in your party might be like 1850 or something. this means that you are slightly above where you belong but the other 4 are all below where you belong. however this does not affect the other team at all. the game will go through the player base and try to make whole teams of similar mmr and put them against each other. do you see how it favors solo players yet?
the system works fine.
|
So you think that a group of players who are 1700 individually after playing together for many games would not have an advantage playing against 5 players who do not know each other and have never played with each other before who are 1700 individually?
Because i think they do. And i think that the suggestion that they do not is ridiculous.
I understand the implication that a group of players who are 1700 as a group does not have an advantage over a group of players who are 1700 individually. But that isn't the issue and your insistence that it is makes me think you do not understand how MMR works.
E.G. lets say you take a 1700 player and 4 1900 players and you put them in a group. The first game they play together they will look like a 1850 MMR. The second game they play together will depend on whether or not they won. The 20th game they play together will be the result of the last 20 games. After playing together for 20 times do you think that they're just as good as 1850 rando's? Or do you think they might be better? Suppose they are actually the same despite that being ridiculous.
Additionally what happens is everyone individual MMR's converges to the groups MMR. So eventually everyone in the group has 1850 MMR if the group really never gets better. But they're not any better individually either and they decided to queue solo.
Now you have 4 1900 players playing in 1850 games and one 1700 player playing in 1850 games. Variance of games at 1850 have increased and playing in a group has negatively effected the rest of the population.
edit: But its pretty clear that playing in a group DOES make you better at the game. It makes it easier to communicate and coordinate. And the more you play the more that happens. You will have better strategy as you are able to easily discuss things in advance, you can construct game plans during champion select knowing what everyone else knows how to play. You can trade champions easier, etc etc etc.
|
On March 25 2016 11:05 Goumindong wrote:
So you think that a group of players who are 1700 individually after playing together for many games would not have an advantage playing against 5 players who do not know each other and have never played with each other before who are 1700 individually?
Because i think they do. And i think that the suggestion that they do not is ridiculous.
I understand the implication that a group of players who are 1700 as a group does not have an advantage over a group of players who are 1700 individually. But that isn't the issue and your insistence that it is makes me think you do not understand how MMR works.
E.G. lets say you take a 1700 player and 4 1900 players and you put them in a group. The first game they play together they will look like a 1850 MMR. The second game they play together will depend on whether or not they won. The 20th game they play together will be the result of the last 20 games. After playing together for 20 times do you think that they're just as good as 1850 rando's? Or do you think they might be better? Suppose they are actually the same despite that being ridiculous.
Additionally what happens is everyone individual MMR's converges to the groups MMR. So eventually everyone in the group has 1850 MMR if the group really never gets better. But they're not any better individually either and they decided to queue solo.
Now you have 4 1900 players playing in 1850 games and one 1700 player playing in 1850 games. Variance of games at 1850 have increased and playing in a group has negatively effected the rest of the population.
ah i see what is missing here
a group of 1700 players playing together and improving as a group would no longer be 1700.
for a group to be 1700 it means that you have a bunch of players that together average out to be 1700. therefore they would be matched against other teams that average out to 1700.
at 1850 when they play solo, what they do does not change other peoples' mmr.
the only thing that changes your rank when you play solo is how well you perform over time. if you are playing above your mmr you go up. if you are below then you drop. more people playing at an mmr doesn't affect your mmr over time at all. the only constant factor when you play solo is yourself. that is the only thing that matters when determining your mmr all other factors like smurfs or leavers or afkers or whatever level out over time.
it does not make you better at the game. all of the mistakes your team makes over time are now affecting your ranking just as much as all the perks of playing with them do. for the 4 guys who are better than the one guy they are carrying, their mmr is collectively suffering as a result of their team. therefore playing on a team makes them all worse in the eyes of matchmaking.
all that playing in a group does is add more variables that influence your individual mmr. when you play solo there are fewer things that can go wrong and cause you to be stuck at a particular mmr. thus for climbing it is better to play solo.
|
But, if this system makes it so the collective winrate of 5s teams is 50.01%, then that only inflates the MMR of people who play in 5 stacks by 0.01% right? So the team that should be 1700 MMR has their MMR inflate by 1.7 for every hundred games that they play?
So after 1000 games, which is a fairly appreciable amount for an entire season, their MMR has inflated by 17, about one games worth?
That sounds horrible!
Also didn't this same "issue" occur previously with duos?
This is dumb. This makes me feel dumb. Why do I read this thread, I wanted to talk about Lux. I'm going back to offtopic where at least the conversations are supposed to be dumb.
|
|
|
On March 25 2016 11:21 Ketara wrote:
But, if this system makes it so the collective winrate of 5s teams is 50.01%, then that only inflates the MMR of people who play in 5 stacks by 0.01% right? So the team that should be 1700 MMR has their MMR inflate by 1.7 for every hundred games that they play?
No. Suppose you had 90/10 games. Such that your win rate was 50.1% instead of 50.01%. The final MMR would be the same. Suppose that you had 0/100 games such that your win rate was 51%. The final MMR would be the same. Changing the ratio only changes how fast you converge to the MMR which gives you 50% win rate in both queues.
The time it take to converges depends on all of the 5's play and not simply your group. Because all of the 5's play is interconnected. When you lose a game against a team which has had its MMR increased due to playing solo's you will lose less MMR. And when you beat them you will gain more.
The difference in MMR would be whatever the difference in MMR would be to cause a 1% win rate difference(between 5's and a solo queue). Now, a 51% win rate really isn't that much MMR. By the original ELO system about 6.9.[no clue what that translates to in league] So no big deal. A 60% win rate would be about 70 MMR, and a 70% about 150 MMR.
If you think the advantage is pretty large(and i do, frankly) then the effect is pretty large.
Also didn't this same "issue" occur previously with duos?
Yes, it did. And riot previously but no longer, gave players who played together a positive penalty on their MMR to counteract the advantage they had playing together. That is, your MMR would be calculated as if it were higher, but it would not increase.
On March 25 2016 11:37 JimmiC wrote:
What is annoying is how some people feel the need to turn this thread into a stats argument thread only post about that. Someone open an "argue about stats loosely related to league thread" pls.
Anyone played the new champ? how does he/she feel?
Banned in my game thus far. One report of hilarious OP but it was on treeline so no guarantees
|
On March 25 2016 12:17 Goumindong wrote:Show nested quote +On March 25 2016 11:21 Ketara wrote:
But, if this system makes it so the collective winrate of 5s teams is 50.01%, then that only inflates the MMR of people who play in 5 stacks by 0.01% right? So the team that should be 1700 MMR has their MMR inflate by 1.7 for every hundred games that they play? No. Suppose you had 90/10 games. Such that your win rate was 50.1% instead of 50.01%. The final MMR would be the same. Suppose that you had 0/100 games such that your win rate was 51%. The final MMR would be the same. Changing the ratio only changes how fast you converge to the MMR which gives you 50% win rate in both queues. The time it take to converges depends on all of the 5's play and not simply your group. Because all of the 5's play is interconnected. When you lose a game against a team which has had its MMR increased due to playing solo's you will lose less MMR. And when you beat them you will gain more. The difference in MMR would be whatever the difference in MMR would be to cause a 1% win rate difference(between 5's and a solo queue). Now, a 51% win rate really isn't that much MMR. By the original ELO system about 6.9.[no clue what that translates to in league] So no big deal. A 60% win rate would be about 70 MMR, and a 70% about 150 MMR. If you think the advantage is pretty large(and i do, frankly) then the effect is pretty large. Yes, it did. And riot previously but no longer, gave players who played together a positive penalty on their MMR to counteract the advantage they had playing together. That is, your MMR would be calculated as if it were higher, but it would not increase. Show nested quote +On March 25 2016 11:37 JimmiC wrote:
What is annoying is how some people feel the need to turn this thread into a stats argument thread only post about that. Someone open an "argue about stats loosely related to league thread" pls.
Anyone played the new champ? how does he/she feel? Banned in my game thus far. One report of hilarious OP but it was on treeline so no guarantees
You're focusing to much on the theoretical end state where everyone has played enough games to even out their MMRs. 5's are not matched against smaller groups enough for this to matter.
Assume Riot is accurate that 5's are only matched against smaller groups 1.4% of the time. Let's say you play 1000 games in a season. You get around 20 LP for winning a game. Assuming the 5 stack wins 100% of the time in this 1.4% (which they won't), then they gain 10 free LP each time. (You would expect to win 50% of the time anyways). Over 1000 games, this is 140LP aka about 1 and a half divisions. This is within the variance I see when playing normally. Since you aren't going to win 100% of the mismatched games, the real bonus is going to be less than a division which is not a noticeable difference.
|
It's also assuming that all 1000 of your games are as a 5 stack.
|
Yes. My main point is that even if this does boost people to some degree, I don't think they amount they gain will be enough to really matter to anyone. It will be less than normal fluctuations that happen by chance.
|
On March 25 2016 12:59 droserin wrote:Show nested quote +On March 25 2016 12:17 Goumindong wrote:On March 25 2016 11:21 Ketara wrote:
But, if this system makes it so the collective winrate of 5s teams is 50.01%, then that only inflates the MMR of people who play in 5 stacks by 0.01% right? So the team that should be 1700 MMR has their MMR inflate by 1.7 for every hundred games that they play? No. Suppose you had 90/10 games. Such that your win rate was 50.1% instead of 50.01%. The final MMR would be the same. Suppose that you had 0/100 games such that your win rate was 51%. The final MMR would be the same. Changing the ratio only changes how fast you converge to the MMR which gives you 50% win rate in both queues. The time it take to converges depends on all of the 5's play and not simply your group. Because all of the 5's play is interconnected. When you lose a game against a team which has had its MMR increased due to playing solo's you will lose less MMR. And when you beat them you will gain more. The difference in MMR would be whatever the difference in MMR would be to cause a 1% win rate difference(between 5's and a solo queue). Now, a 51% win rate really isn't that much MMR. By the original ELO system about 6.9.[no clue what that translates to in league] So no big deal. A 60% win rate would be about 70 MMR, and a 70% about 150 MMR. If you think the advantage is pretty large(and i do, frankly) then the effect is pretty large. Also didn't this same "issue" occur previously with duos? Yes, it did. And riot previously but no longer, gave players who played together a positive penalty on their MMR to counteract the advantage they had playing together. That is, your MMR would be calculated as if it were higher, but it would not increase. On March 25 2016 11:37 JimmiC wrote:
What is annoying is how some people feel the need to turn this thread into a stats argument thread only post about that. Someone open an "argue about stats loosely related to league thread" pls.
Anyone played the new champ? how does he/she feel? Banned in my game thus far. One report of hilarious OP but it was on treeline so no guarantees You're focusing to much on the theoretical end state where everyone has played enough games to even out their MMRs. 5's are not matched against smaller groups enough for this to matter. Assume Riot is accurate that 5's are only matched against smaller groups 1.4% of the time. Let's say you play 1000 games in a season. You get around 20 LP for winning a game. Assuming the 5 stack wins 100% of the time in this 1.4% (which they won't), then they gain 10 free LP each time. (You would expect to win 50% of the time anyways). Over 1000 games, this is 140LP aka about 1 and a half divisions. This is within the variance I see when playing normally. Since you aren't going to win 100% of the mismatched games, the real bonus is going to be less than a division which is not a noticeable difference.
Actually, it makes more sense if you just think of it like this: All 5stacks are boosted by the 1.4%, because every time any 5stack wins it brings extra ELO into the 5v5 bracket. Thus the 5v5 pool is always inflated. This means you are actually always better off queuing as 5 because you get greater ELO gains just by the virtue that other 5s teams win, even if your 5s sucks.
Also, it matters because epeen is actually the goal, and people should stop saying silly things like "play to improve". Its all about next season's sweet Kindred skin, or a border.
|
holy shit aurelion sol's engage potential is insane
im pulling off the most retarded engages from like 2 towers away
the fact that damage cancels his W is pretty huge though...
|
On March 25 2016 13:19 droserin wrote:
Yes. My main point is that even if this does boost people to some degree, I don't think they amount they gain will be enough to really matter to anyone. It will be less than normal fluctuations that happen by chance.
A clutz notes:
Consider a set of two games involving the same five stack. This can be two five v fives of one five v small and one five v five. The net mmr in the first two games is necessarily zero for the five stacks. In the second the net mmr for the five stacks is positive if the five v small was won and negative if the five v small was lost. This mmr transfer between queues effects the second fives group that plays after the five v small regardless of who wins.
The mmr transfers continue until the probability of winning the five v small is 50% and that initial win rate difference determines how far the correction happens and nothing else
How fast it happens depends on how many five vs small games take place in total not how many you play in particular. Because of their above reason where the other team benefits from your win
So if there were only 1000 games played by five stacks in the entirety of a server in a year you might have an argument. But that number seems widely low to me.
|
On March x 2016 y:z EVERYONE wrote:
ELO
Please boys it's Elo not ELO, Elo-rating is named after it's inventor Arpad Elo. As a statician it hurts my eyes and very soul to see it so often it's almost come to be a standard error (hue).
On March 25 2016 11:37 JimmiC wrote:
Anyone played the new champ? how does he/she feel?
Hard to say, he's pretty unique and so far the only games I've seen with him he has been solving world hunger or being on the receiving end of a Flame horizon. I think once people learn to play him he's fairly strong. Currently I'm using the secret special operations tactic of banning him if someone on my team wants to play him and leave him open if not granted of course someone else doesn't ban him. Kinda dickish but it have worked flawlessly since his release.
The only space dragon I've met who was relevant played him top and went Rylai's into tank with Frozen Heart as the mana item.
|
I think he's one of the better new champs at release.
theres lots of angles you can take the with the long initiation and as long as you have an initial cc/engage to keep people in place for a bit, you can get sick initiates or picks.
just make sure you sweep your jungle and take a route through it to mid or w.e
pretty sure he'll be picked up by pros within a month or so.
kinda feel similarly about him as when i first evaluated bard, but might be a hit or miss depending on his role. at least bard was clearly sup
|
Sadly Lee Sin cant travel with him, just tried in a custom and his dash is too fast. :'(
Wonder what happens if Sol is inside a tree when Lee arrives on him?
|
He's very interesting build wise too.
He wants mana / regen so he can keep his W on for the duration of a fight, but doesn't want CDR since really only his Q gets returns from it, and his Q cooldown is long enough that it's not particularly significant. He also doesn't really want Seraphs (does anybody these days?) because he will have some issues stacking it.
He also really wants spell effects since he'll be AoE proccing them the whole fight long.
He also likes Movespeed as a stat, since the faster he moves the longer he'll be able to stay with his Q and the bigger it will get.
So Rylais + Liandry seems really obvious, either RoA or Frozen Heart seems really obvious. His other 2 items could very easily be tank items for a more AP bruisery sort of build, or they could be like Ludens + Void for a more damage heavy build.
How does he work with Zhonyas? When he's stasis'd do his orbs still orbit?
He seems neat. Neater than I thought he would be. I like him.
|
He can actually stack Tear super fast since W spam triggers it, like Singed's Q. W cost per second ramps up really fast as you level it tho, but it's easily toggled on/off. Interestingly W also triggers Shen items.
|
Sure, stacks really fast on that 6 second cooldown.
|
Eh. Tear can stack twice over 8 seconds. Fits just fine with his W cooldown.
|
|
|
|
|
|
EPT
