North America - English

League of Legends Community

League of Legends Community > League of Legends > Player Behavior
After Hours with Matchmaking and Lyte

View Poll Results: Which matchmaking 'issue' is the most important to you? (See post for more details!)
1) AFKs in Champion Select Lobby 4,690 36.97%
2) Duo-Queue Elo Disparities in Ranked 910 7.17%
3) Skilled Ranked Players in Normal Modes 603 4.75%
4) Premade Matching 633 4.99%
5) Transitioning from Normal to Ranked Mode 1,281 10.10%
6) Free to Play Champions in Ranked Mode 791 6.24%
7) Random Champions in Ranked Mode 627 4.94%
8) Provisional Matches in Ranked 645 5.08%
9) Duo Queue Prevalence in Ranked 384 3.03%
10) Level Disparities 583 4.60%
11) Team Margin of Victory 1,538 12.12%
Voters: 12685. You may not vote on this poll

First Riot Post
 
Comment below rating threshold, click here to show it.
Lyte Lyte's Avatar ?? Lead Designer of Social Systems

Follow RiotLyte on Twitter
This user has referred a friend to League of Legends, click for more information This user has accepted the summoners code, click for more information
291 of 362 Riot Posts
07-18-2012

Quote:
Originally Posted by Metronomotopoeia View Post
Actually, you could use the same logic to say that it's easier to climb at the start of a season, because half the population is fighting teams they'd never otherwise be against, which results in super quick stratification.

For example, a 2200 team can get matched against several sub-1600 teams, which would be practically free wins. By the same token, 900-tier players would lose almost every single match, as opposed to if they were just fighting straight 1100-1200 teams, which would have some opportunities for back-and-forth given the lower relative skill difference.

It goes back to the earlier point about it being a committed, stable team. It'd be different if the 2200 team in question were 3 players with wildly different individual/"real" ratings that just averaged out, but no one ever really made those kinds of teams. In other words, you'd never find two 2400 players who would be willing to hang with a 1800 player on their team because he was a super cool dude.

Not disagreeing with your line of thinking or philosophy about matchmaking, but I think this point in particular worked the opposite way in reality. To the system, things definitely seem noisy, because it's trying to put together even matches. But to the goal of getting people to their ratings quickly, it worked pretty well.
This wasn't an opinion :P It was just data. You have to consider that statistically the outcomes are almost the same, except the matches are more lopsided due to more noise. This is because at any point in a season about half the players starting in Ranked should be above 1200 and half should be below, so at any point in the season the vast majority of players are playing against players outside their brackets until they are placed appropriately. At the start of a season, however, almost all of the top players are bunched at 1200 and climbing out. At the middle or end of a season there's far fewer Ranked players starting who should be at the very top of the ladder (because they all jumped in early), so there's slightly less noise at the 1200 range.

 
Comment below rating threshold, click here to show it.
Brodhi ?? Senior Member
This user has accepted the summoners code, click for more information
07-18-2012

Quote:
Originally Posted by Lyte View Post
I think you might have a misunderstanding of how matchmaking systems work. Almost every game with a matchmaking system tries to predict which team might be the winner statistically... in fact, I'm not even considering just eSports but many real sport ratings do the exact same thing. Not having set amounts for a Win or Loss is also one of the reasons all the matchmaker systems that do use granular Elo adjustments are more accurate too. For example, if a 0 Rating player beat a 2200 Rating player, should he still get 40? What if a 0 Rating player beats a 1100 Rating player, still gets 40? Or if you want to specifically look at games within a 'bracket,' isn't a 0 Rating player beating a 1199 Rating player more meaningful than a 0 Rating player beating a 400 Rating player?

Also, for a 1500-1800 player, it only takes a dozen games or so to climb out of 900-1100 Elo so I'm not sure where you are coming from about "Elo Hell." I also never said it takes 300 games to get to true Elo because you got players with 1-3 wins during seeding... or that players are placed too high because of players with 300 wins during seeding...

Confused.
Clearly, you have a ton more metrics than we the Community have, so all our ideas, comments, and thoughts are based off our own individual experiences and not what the numbers say. Go check out a Streamer that is in the 600-900 ELO. Tell me how well he does in "climbing" out after a dozen games. In fact, I don't even know where you think a dozen games gets you from 1100 to 1500? The LoL Matchmaker has it so that, in general, you will more than likely accumulate a 50% W/L ratio. Using this idea, you will only gain about 36 ELO from 6 wis / 6 losses (assuming 13 ELO for a win, 10 for a loss because the Matchmaker put you up against "better" players for getting on a win streak or whatever). Even if you won all 12 games, that would still only be 156 ELO (if all games gave 13 ELO), which only puts you at 1256. Can you explain what numbers you are using to figure out how a 1100 player can get to 1500 from a dozen games?

I 100% think that a 0 rated player beating a 200 rated player or a 1199 rated player should give the exact same ELO amount, because for all intents and purposes, they are NOT RATED. You don't get rated until you break 1200 ELO, and as such it is unimpressive for a non-ranked player to beat another non-ranked player. Now, if a 0 rated team beats a 1250 rated team ( how this happened, I do not know), then sure, award the 0 rated team like 140 ELO, because a non-rated team beating a rated team is a big deal--its an upset.

The same is so true when you gamble on the spread in a Football game, for example. You don't bet money on who is going to win, you bet that a team either will win by X amount, less than X amount, more than X amount, or will flat out lose. The predictions they use come directly from HOW THOSE TEAMS HAVE PLAYED AGAINST EACH OTHER IN THE PAST. It is very rare for the same 5 players to play against the same 5 players over and over in LoL, as such it is highly inaccurate to determine if one Mid-Lane opponent is going to beat another, or even if a team is going to beat another, without any data about these exact same teams playing against each other before.

Confused.

Edit: What I was saying is that you said it takes 300 wins to get to your true ELO. Therefore, using my understanding, a player with 300+ wins at 900 ELO belongs there. When a player first plays ranked and they are on a team with a bunch of people with 300+ wins against a bunch of players with 1-3 wins, the team with 300+ wins is more than likely going to win, because that team is more proficiency in the "900 ELO mindset" as well as having a lot of experience in the game.

3) Provisional Matches in Ranked
When a player joins Ranked for the first time, the system starts them at 1200 Elo and begins their "placement matches." Unlucky strings of losses or lucky streaks of wins can propel a player into an Elo tier they do not belong in. Alternatively, players generally do not like playing with or against players in their placement matches and seeing they have only 1-9 wins.

^ That is you saying the seeding process can place someone in a bracket they do not belong in. This has a LOT to do with during seeding being placed on a team with a bunch of people who have 1-3 wins against a team with a bunch of 300+ wins. Data can show you this.

 
Comment below rating threshold, click here to show it.
Metronomotopoeia ?? Senior Member
This user has accepted the summoners code, click for more information
07-18-2012

Quote:
Originally Posted by Lyte View Post
This wasn't an opinion :P It was just data. You have to consider that statistically the outcomes are almost the same, except the matches are more lopsided due to more noise. This is because at any point in a season about half the players starting in Ranked should be above 1200 and half should be below, so at any point in the season the vast majority of players are playing against players outside their brackets until they are placed appropriately. At the start of a season, however, almost all of the top players are bunched at 1200 and climbing out. At the middle or end of a season there's far fewer Ranked players starting who should be at the very top of the ladder (because they all jumped in early), so there's slightly less noise at the 1200 range.
Yeah, I get that, and it makes sense. I meant that in the WoW setting (as opposed to the LoL one), stratification to a person's/team's appropriate rating actually happened very quickly at the start of a season, compared to in the middle of the season. In my personal observation/experience (as I had no access to Blizzard's internal spreadsheets), very little rating fluctuation occurred past the midway point of a season. Where people ended up after the first two weeks/80 games of a season tended to be pretty accurate.

I wasn't trying to translate it to today's ranked scenario, just picking something to play Devil's Advocate on - which I guess makes it a somewhat irrelevant point for the League of Legends forums. =P

 
Comment below rating threshold, click here to show it.
Beltroniko ?? Member
This user has accepted the summoners code, click for more information
07-18-2012

Quote:
Originally Posted by Brodhi View Post
Edit: What I was saying is that you said it takes 300 wins to get to your true ELO. Therefore, using my understanding, a player with 300+ wins at 900 ELO belongs there. When a player first plays ranked and they are on a team with a bunch of people with 300+ wins against a bunch of players with 1-3 wins, the team with 300+ wins is more than likely going to win, because that team is more proficiency in the "900 ELO mindset" as well as having a lot of experience in the game.

3) Provisional Matches in Ranked
When a player joins Ranked for the first time, the system starts them at 1200 Elo and begins their "placement matches." Unlucky strings of losses or lucky streaks of wins can propel a player into an Elo tier they do not belong in. Alternatively, players generally do not like playing with or against players in their placement matches and seeing they have only 1-9 wins.

^ That is you saying the seeding process can place someone in a bracket they do not belong in. This has a LOT to do with during seeding being placed on a team with a bunch of people who have 1-3 wins against a team with a bunch of 300+ wins. Data can show you this.
The problem is that some people with 3 wins might be as good (or better) as the guy with 300+ wins if his true ELO is the same.

What affects that is that when ur under 1200 ELO, u get people who are on (example) 800 ELO because they belong there and people who are on 800 ELO cause they are on their way to (example) 300 ELO. If you have 300+ wins and play with a 3 win player who has a skill level of 800 ELO that shouldn't be a problem or difference while if u play a 3 win player that is on his way to 300 ELO, you'll have a huge disadvantage, but if this didn't happen, those players level wouldn't be assessed (cause if they only play against people like them, some will get higher when they should get lower).

 
Comment below rating threshold, click here to show it.
Brodhi ?? Senior Member
This user has accepted the summoners code, click for more information
07-18-2012

Quote:
Originally Posted by Beltroniko View Post
The problem is that some people with 3 wins might be as good (or better) as the guy with 300+ wins if his true ELO is the same.

What affects that is that when ur under 1200 ELO, u get people who are on (example) 800 ELO because they belong there and people who are on 800 ELO cause they are on their way to (example) 300 ELO. If you have 300+ wins and play with a 3 win player who has a skill level of 800 ELO that shouldn't be a problem or difference while if u play a 3 win player that is on his way to 300 ELO, you'll have a huge disadvantage, but if this didn't happen, those players level wouldn't be assessed (cause if they only play against people like them, some will get higher when they should get lower).
Which is exactly why starting everyone at 0 ELO and working your way UP to your ELO rather than UP or DOWN is far, far superior than the current system and is exactly why Blizzard changed it coming into Wrath of the Lich King.

 
Comment below rating threshold, click here to show it.
Lyte Lyte's Avatar ?? Lead Designer of Social Systems

Follow RiotLyte on Twitter
This user has referred a friend to League of Legends, click for more information This user has accepted the summoners code, click for more information
292 of 362 Riot Posts
07-18-2012

Quote:
Originally Posted by Brodhi View Post
Clearly, you have a ton more metrics than we the Community have, so all our ideas, comments, and thoughts are based off our own individual experiences and not what the numbers say. Go check out a Streamer that is in the 600-900 ELO. Tell me how well he does in "climbing" out after a dozen games. In fact, I don't even know where you think a dozen games gets you from 1100 to 1500? The LoL Matchmaker has it so that, in general, you will more than likely accumulate a 50% W/L ratio. Using this idea, you will only gain about 36 ELO from 6 wis / 6 losses (assuming 13 ELO for a win, 10 for a loss because the Matchmaker put you up against "better" players for getting on a win streak or whatever). Even if you won all 12 games, that would still only be 156 ELO (if all games gave 13 ELO), which only puts you at 1256. Can you explain what numbers you are using to figure out how a 1100 player can get to 1500 from a dozen games?

I 100% think that a 0 rated player beating a 200 rated player or a 1199 rated player should give the exact same ELO amount, because for all intents and purposes, they are NOT RATED. You don't get rated until you break 1200 ELO, and as such it is unimpressive for a non-ranked player to beat another non-ranked player. Now, if a 0 rated team beats a 1250 rated team ( how this happened, I do not know), then sure, award the 0 rated team like 140 ELO, because a non-rated team beating a rated team is a big deal--its an upset.

The same is so true when you gamble on the spread in a Football game, for example. You don't bet money on who is going to win, you bet that a team either will win by X amount, less than X amount, more than X amount, or will flat out lose. The predictions they use come directly from HOW THOSE TEAMS HAVE PLAYED AGAINST EACH OTHER IN THE PAST. It is very rare for the same 5 players to play against the same 5 players over and over in LoL, as such it is highly inaccurate to determine if one Mid-Lane opponent is going to beat another, or even if a team is going to beat another, without any data about these exact same teams playing against each other before.

Confused.

Edit: What I was saying is that you said it takes 300 wins to get to your true ELO. Therefore, using my understanding, a player with 300+ wins at 900 ELO belongs there. When a player first plays ranked and they are on a team with a bunch of people with 300+ wins against a bunch of players with 1-3 wins, the team with 300+ wins is more than likely going to win, because that team is more proficiency in the "900 ELO mindset" as well as having a lot of experience in the game.

3) Provisional Matches in Ranked
When a player joins Ranked for the first time, the system starts them at 1200 Elo and begins their "placement matches." Unlucky strings of losses or lucky streaks of wins can propel a player into an Elo tier they do not belong in. Alternatively, players generally do not like playing with or against players in their placement matches and seeing they have only 1-9 wins.

^ That is you saying the seeding process can place someone in a bracket they do not belong in. This has a LOT to do with during seeding being placed on a team with a bunch of people who have 1-3 wins against a team with a bunch of 300+ wins. Data can show you this.
Ah, there are definite misunderstandings about how matchmaking works.

1) For the 600-900 Elo Streamer, he should only climb out if he actually is a top tier player.

2) The matchmaker does not try to put players at a 50/50 win ratio. It just tries to place players as quickly as possible into their tiers. For players who are not 1200, they most definitely will not be going 6 Wins/6 Losses at the 1200 bracket.

3) If you actually won all 12 games at the beginning of Ranked, you would actually get about ~500 Elo. This is because we have advanced seeding algorithms that will give a lot more than +14 Elo when a player is climbing out of 1200 range making it actually very easy to get within 100-200 of your true Elo in a small number of games.

4) Finally, no data so far has shown that # of Wins has a huge effect on skill. As far as statistics go, a player with 300 Wins and 1500 Elo is about the same as a player with 200 Wins and 1500 Elo. However, if you keep randomly selecting 300 Win, 1200 Elo players and put them against 0 Win (and therefore automatically 1200 Elo) players... you actually will see them split their matches 50/50.

5) Unlucky strings can put you at brackets you don't belong in but it has nothing to do with players with 1-3 Wins or 300 Wins. It has to do with things like even if the match is an expected 50/50 match, you could dice roll and lose every game. Or, it has to do with connection issues. More importantly, even if you are placed in the wrong bracket, you are placed in the right direction within 25 games because of the advanced Elo adjustments as you play the ladder that give you more than +/- 14.

 
Comment below rating threshold, click here to show it.
Brodhi ?? Senior Member
This user has accepted the summoners code, click for more information
07-18-2012

Quote:
Originally Posted by Lyte View Post
Ah, there are definite misunderstandings about how matchmaking works.

1) For the 600-900 Elo Streamer, he should only climb out if he actually is a top tier player.

2) The matchmaker does not try to put players at a 50/50 win ratio. It just tries to place players as quickly as possible into their tiers. For players who are not 1200, they most definitely will not be going 6 Wins/6 Losses at the 1200 bracket.

3) If you actually won all 12 games at the beginning of Ranked, you would actually get about ~500 Elo. This is because we have advanced seeding algorithms that will give a lot more than +14 Elo when a player is climbing out of 1200 range making it actually very easy to get within 100-200 of your true Elo in a small number of games.

4) Finally, no data so far has shown that # of Wins has a huge effect on skill. As far as statistics go, a player with 300 Wins and 1500 Elo is about the same as a player with 200 Wins and 1500 Elo. However, if you put keep randomly selecting 300 Win, 1200 Elo players and put them against 0 Win (and therefor automatically 1200 Elo) players... you actually will see them split their matches 50/50.
1) I am not trying to tell you the skill level of that particular Streamer, what I am trying to do is to get Riot, and specifically the department within Riot, to sit down and actually watch how these games unfold. Read the chat (because you can't in Spectate), learn the champions the 600-900 ELO players play, etc. I want you guys to legitemtly BECOME a 600-900 ELO player and use that to understand why people think there is an ELO Hell, and why so many people think it is incredibly hard to climb out of 600-900 once you get placed there. Riot has been denying ELO Hell saying, "if you are good enough, you will climb", but the problem there is that how long does that take? If it takes 500 games played is that fair? Shouldn't we have a system that gets them there faster?

2) I truly do not see how this is the case. If someone who is 1200 rated (Ranked or Normal) wins 3 games in a row, they will immediately be placed against players who are 1300+ because your system assumes they belong there, yes? This happens if you are level 20-29 and get on a winning spree, you'll soon be placed against a bunch of level 30s. There are a ton of differences between a level 20 and a level 30 player and a 1200 rated player and a 1300 rated player. In both instances, generally, the latter have more Runes / Champions / Knowledge of the game. So placing them in a tier you just "think" they belong in because they are on a winning spree to me seems lacklust.

3) We are not talking about the beginning of Ranked, Ltye. You said someone who is 900 ELO, NOT 1200, could get to 1500 ELO in a dozen games. I still want your math on that. You can't tell me "if they win all 12 games" because that is extremely, extremely rare. Please show me some kind of metric or math that proves to me that after only 12 games someone who is 900 ELO is now 1500.

4) See my post above.

 
Comment below rating threshold, click here to show it.
Metronomotopoeia ?? Senior Member
This user has accepted the summoners code, click for more information
07-18-2012

Also, out of curiosity, how come you never pass through The Tribunal forum, Lyte?

Not that we're in dire need of attention over there, but I would have thought that'd be your home base. =D

 
Comment below rating threshold, click here to show it.
Clerigon ?? Senior Member
This user has accepted the summoners code, click for more information
07-18-2012

Quote:
Secondly, it only takes about 12 games to hit 1500 Elo since there is a seeding system in play just like in other games.
One of the main problem of ranked is this, This is like a double edge sword, Since all the team is punished for an afk, leaver or just a guy with bad luck and had to leave or bad internet connection, thats like -90 elo in the first games.

Why i dont play ranked? because i dont find fun how low elo games are, stomp or be stomped, i dont feel good stomping someone in 1100 elo because that guy deserve to be there and i need to act like: *****imabus get of my way. I think is one of the reason i dont have smurfs..

I dont have a problem if i deserve to be in low elo, only when the match are fair.


Quote:
3) If you actually won all 12 games at the beginning of Ranked, you would actually get about ~500 Elo.
so i get sky rocket to 1700 elo, 1200-1400 elo is not the same as 1700+ elo, is just other world. Now im in 1700 elo with 15-0 in ranked games, first game in 1700 i get stomped hard because im only there playing pubstomp champions, you think my team mates will find fair they are playing with a new ranked player? XD

Quote:
Finally, no data so far has shown that # of Wins has a huge effect on skill.
lets make an experiment, do you have any experience in dota, w3 rts or AoC? lets play a match, you said wins mean nothing, because experience worth nothing in lol, by your logic, you will have 50% chance to win in you first w3 game vs me or other experienced player.

But lol is not 1v1!! ok lets make a 3v3 team in age of empire, the game will be 3v2 because you are new in the game..nor fair for you team, not a good way to recognize the enemy team skills.


Why ranked games dont count normal elo/stats for the seeds? it sound logical, for lvl 30 you need like 300 games with at least 50% win ratio. Is a good way to know where that player belong in ranked.

 
Comment below rating threshold, click here to show it.
Apollinarius ?? Senior Member
This user has accepted the summoners code, click for more information
07-18-2012

Quote:
Originally Posted by Lyte View Post
Ah, there are definite misunderstandings about how matchmaking works.

1) For the 600-900 Elo Streamer, he should only climb out if he actually is a top tier player.

2) The matchmaker does not try to put players at a 50/50 win ratio. It just tries to place players as quickly as possible into their tiers. For players who are not 1200, they most definitely will not be going 6 Wins/6 Losses at the 1200 bracket.

3) If you actually won all 12 games at the beginning of Ranked, you would actually get about ~500 Elo. This is because we have advanced seeding algorithms that will give a lot more than +14 Elo when a player is climbing out of 1200 range making it actually very easy to get within 100-200 of your true Elo in a small number of games.

4) Finally, no data so far has shown that # of Wins has a huge effect on skill. As far as statistics go, a player with 300 Wins and 1500 Elo is about the same as a player with 200 Wins and 1500 Elo. However, if you keep randomly selecting 300 Win, 1200 Elo players and put them against 0 Win (and therefore automatically 1200 Elo) players... you actually will see them split their matches 50/50.

5) Unlucky strings can put you at brackets you don't belong in but it has nothing to do with players with 1-3 Wins or 300 Wins. It has to do with things like even if the match is an expected 50/50 match, you could dice roll and lose every game. Or, it has to do with connection issues. More importantly, even if you are placed in the wrong bracket, you are placed in the right direction within 25 games because of the advanced Elo adjustments as you play the ladder that give you more than +/- 14.
Why would it take 300 wins to reach your elo? My elo is horrible (around 850). At some point I was down to 700 when I first started, managed to crawl my way back up to 1150, but then an unfortunate string of losses (some were partially my fault, some were really not) I've dropped down to 850. Now when I play, there is about a 70% chance that someone on either team will afk or troll, meaning that only 30% of the games I play are in any way affected by player skill on either team.

I'm not saying I'm great player. Playing my best role (support or ad carry) I expect I would be comfortable in the 1200 elo range. The problem is that it is impossible for me to actually play with people in that skill bracket. In the bracket I'm in, teams are so severely affected by new players who are only getting used to ranked and seasoned trolls who managed to still be 850 elo after 500 wins that games are simply not enjoyable.

If I want to play with and against people of skill comparable to me, I'm best playing normal games. There is something wrong with that. Normal games matching is supposed to be without rating and just thrown together haphazardly for people to have fun. Ranked game matchmaking is supposed to be based on skill. From where I stand, I feel that Ranked matchmaking is only based on skill in the 1500 and above range. Once you get over 2k, then it is really only people who belong there. Everything below is filled with people who are experiencing 10 game win or losing streaks for one reason or another and they spend only fleeting moments in the same bracket as people of comparable skill to them.