Recently the Player Behavior team has been seeing a lot of questions about matchmaking. In the future, we'd like to do more regular updates on how matchmaking is working and share some of the results behind our matchmaking systems. But for now, we'd like to start with a FAQ that answers some of the more common questions we see about matchmaking. Lyte, Glacest, and I will be happy to answer any questions and will add to the FAQ as needed.
Q1: Does ELO Hell exist?
Based on our data and analysis, the answer is no. However, that does not mean that the pain many of our players feel is invalid. Many players feel that, through no fault of their own, they are stuck in a situation where they can no longer progress. And there are several key factors that can heavily contribute to this feeling. The first of these is that when we take a look across all games, some players actually require a pretty high number of games to reach their true matchmaking rating. For these players, what feels like an unsurpassable wall may end up being just a small road bump.
The two major psychological factors which contribute to the sense of ELO Hell are the Dunning-Kruger Effect and negativity bias. The Dunning-Kruger effect refers to the phenomenon where people tend to rate their own abilities as much higher than average, regardless of reality. For League of Legends, this translates into most players rating their matchmaking rating at about 150 points higher than their actual matchmaking rating. Negativity bias refers to the cognitive trend where people negative experiences tend to be more mentally sticky than positive experiences. In other words, negative experiences, such as playing with a leaver or a troll, will have a larger impact on your memory.
Q2: How does matchmaking work right now?
Currently the matchmaker attempts to match players with similar matchmaking rating (referred to as MMR from here on out) and number of wins for that particular queue. Players have a separate MMR for each queue, each of which evolves independently from the others. This means that your wins and losses in Dominion wont impact your MMR for Summoners Rift. The one exception to this rule is that in some special cases, well look at your performance in other queues to help kick off your MMR. For example, when starting a ranked team, the teams initial MMR is impacted by the past performances of the members in ranked solo queue. After the team is made, the individual solo queue experience of the members will stop impacting the teams MMR.
Q3: If thats the case, how do I keep ending up in games with players whose skill level differs greatly from my own?
There are a few factors that can make a match feel more imbalanced than it actually is. A teammate may have just had a bad game and performed worse than they usually do. Alternatively, as weve seen in LCS, snowballing can still occur despite the two teams being very evenly matched. Snowballing can then lead to the very reasonable perception that a mismatch occurred. Finally, both Dunning-Kruger and negativity bias also play a role here in shaping players impressions of their matchmaking experiences overall.
For the matchmaker itself, the conditions often cited for mismatch are low concurrent users (CCU) or having a very high MMR. But when we take a deeper look at this pool of games, the impact on the quality of matches ends up being very very low. Instead, the matchmaker is much more likely to increase the queue time as can be seen in the longer queue times seen in the Challenger Tier.
That is not to say that our matchmaker is absolutely perfect. But we are also constantly monitoring our matchmaker and making any necessary tweaks to help reduce the chance that players of vastly different skill levels are matched together.
Q3.5: But according to their ranked borders and third party website X, their MMR is way higher or lower than mine! Why does this happen?
There are two different issues to consider here. For borders, the main consideration is that they only represent the skill level of the player at the end of last season. Since the last season ended in November of 2012, this represents a skill evaluation that is ten months old and very likely inaccurate. In terms of third party websites, their estimation of players ELO is based off of incomplete information and thus can be fairly inaccurate.
Q4. Is the matchmaker punishing me for my winning streaks? Ive heard that it tries to force everyone into having a 50% win chance. How is this fair!?
This is false. While the matchmaker tries to create games in which the teams are balanced and have an even chance of winning, this is not always possible. As a result, the ultimate goal of the matchmaker is to be as accurate as possible in predicting the winner of the game, thus allowing their matchmaking rating to be adjusted appropriately. In some cases (particularly for players new to that queue) win streaks can lead to a rapid increase in your MMR, which will cause you to face tougher and tougher opponents. Naturally, this will eventually end your win streak, but youll find yourself landing at a new level of more challenging competition.
Q5. Why dont we just put toxic players into a different queue?
This concept, also known as Prisoners Island, has been proposed multiple times and considered by the Player Behavior team. However, we ultimately determined that this model was not a good fit for League of Legends. Not only does it go against our goal of reforming toxic players, but it also feels extremely negative in the rare case someone is incorrectly sent there. Finally, since League of Legends is a free to play game, prisoners may opt to just abandon their account and transfer their toxicity to our newest players.
Q6: Why is duo queuing allowed?
While duo queuing can sometimes create less than ideal situations in matchmaking, the vast majority of our players actually prefer to play with their friends. While we have to be cognizant of this fact, we are also monitoring the solo queuer experience.
Q7: How does Riot feel about smurfs?
Riot does not officially condemn or condone smurfing. We understand the cases where players make new accounts in order to help teach new players the game, but are working on newer systems to provide a similar experience in a more positive way. Creating smurfs simply to beat lower level players is heavily frowned upon. Permanently banned players who make smurfs are quickly banned again. A permanent ban means that unfortunately, we feel that there is no place for you in our game.
The part where their are players doomed to not be able to play again kinda feelt a bit harsh :/
I know they are toxic people but what if they have changed? Just a question not saying that there shouldn`t be a punishment for toxic players but still, i fell the empathy.
Long ago, toxic players who genuinely wanted to reform and could demonstrate their honesty were offered a second chance. I don't remember the specific details, but it had an extremely high failure rate (~95%) and we discontinued it.
Still think its odd, someone with an unimpressive w/l ratio like myself (who is also level 25),
Yet during PST prime time I would say every 4-5 matches or so I will get matched with several silver/gold/plat players in ARAMS or Normals.
Just a weird system.
Quote:
Originally Posted by EveAsian
Is normal game MMR = ranked game MMR? If I won 20 games in a row in normals does that affect my MMR in ranked?
To answer both questions at once, you have a separate MMR for each queue. 5ackLunch, you might be seeing Silver/Gold/Plat players because you've done better than they have in that particular queue.
...and my reason for doing so was the stupid border around my summoner on the loading page...
We actually ran a few studies on how having a border impacted toxicity. To quote Lyte: "Players with lower border colors did not receive any additional harassment or reports compared to the players with higher border colors in the same game."
Expanding on this question: How do premades work in ranked?
Do individual players have an MMR for, say ranked 3's, and the overall team MMR for a given game is some kind of average or weighted average based on the ranked 3's MMR for each player actually participating? Or is there one MMR assigned to a given ranked team, regardless of who the participants are?
I ask because I suspect it is the former, based on previous experience, while the latter makes more sense to me.
The answer is actually somewhere between. While the MMRs exist independently, ranked teams are one example of where there is a little bit of crossover. When a team is first made, the MMRs of the members impact the starting MMR of the team. Once that's established, the team's MMR exists independently from each of the member's solo queue MMR.
Q: In queues with very few people playing (e.g., Normal Draft, Dominion, TT, etc.), does the matchmaker attempt to even out extremely high rated players by giving them very low rated players?
I would understand if it did, since long queue times are never the best, but I would like to know.
It actually prefers to extend queue times over sacrificing accuracy in those cases. That's why challenger players can face much longer queue times than the average player.
Matchmaking FAQ
Recently the Player Behavior team has been seeing a lot of questions about matchmaking. In the future, we'd like to do more regular updates on how matchmaking is working and share some of the results behind our matchmaking systems. But for now, we'd like to start with a FAQ that answers some of the more common questions we see about matchmaking. Lyte, Glacest, and I will be happy to answer any questions and will add to the FAQ as needed.
Q1: Does ELO Hell exist?
Based on our data and analysis, the answer is no. However, that does not mean that the pain many of our players feel is invalid. Many players feel that, through no fault of their own, they are stuck in a situation where they can no longer progress. And there are several key factors that can heavily contribute to this feeling. The first of these is that when we take a look across all games, some players actually require a pretty high number of games to reach their true matchmaking rating. For these players, what feels like an unsurpassable wall may end up being just a small road bump.
The two major psychological factors which contribute to the sense of ELO Hell are the Dunning-Kruger Effect and negativity bias. The Dunning-Kruger effect refers to the phenomenon where people tend to rate their own abilities as much higher than average, regardless of reality. For League of Legends, this translates into most players rating their matchmaking rating at about 150 points higher than their actual matchmaking rating. Negativity bias refers to the cognitive trend where people negative experiences tend to be more mentally sticky than positive experiences. In other words, negative experiences, such as playing with a leaver or a troll, will have a larger impact on your memory.
Q2: How does matchmaking work right now?
Currently the matchmaker attempts to match players with similar matchmaking rating (referred to as MMR from here on out) and number of wins for that particular queue. Players have a separate MMR for each queue, each of which evolves independently from the others. This means that your wins and losses in Dominion wont impact your MMR for Summoners Rift. The one exception to this rule is that in some special cases, well look at your performance in other queues to help kick off your MMR. For example, when starting a ranked team, the teams initial MMR is impacted by the past performances of the members in ranked solo queue. After the team is made, the individual solo queue experience of the members will stop impacting the teams MMR.
Q3: If thats the case, how do I keep ending up in games with players whose skill level differs greatly from my own?
There are a few factors that can make a match feel more imbalanced than it actually is. A teammate may have just had a bad game and performed worse than they usually do. Alternatively, as weve seen in LCS, snowballing can still occur despite the two teams being very evenly matched. Snowballing can then lead to the very reasonable perception that a mismatch occurred. Finally, both Dunning-Kruger and negativity bias also play a role here in shaping players impressions of their matchmaking experiences overall.
For the matchmaker itself, the conditions often cited for mismatch are low concurrent users (CCU) or having a very high MMR. But when we take a deeper look at this pool of games, the impact on the quality of matches ends up being very very low. Instead, the matchmaker is much more likely to increase the queue time as can be seen in the longer queue times seen in the Challenger Tier.
That is not to say that our matchmaker is absolutely perfect. But we are also constantly monitoring our matchmaker and making any necessary tweaks to help reduce the chance that players of vastly different skill levels are matched together.
Q3.5: But according to their ranked borders and third party website X, their MMR is way higher or lower than mine! Why does this happen?
There are two different issues to consider here. For borders, the main consideration is that they only represent the skill level of the player at the end of last season. Since the last season ended in November of 2012, this represents a skill evaluation that is ten months old and very likely inaccurate. In terms of third party websites, their estimation of players ELO is based off of incomplete information and thus can be fairly inaccurate.
Q4. Is the matchmaker punishing me for my winning streaks? Ive heard that it tries to force everyone into having a 50% win chance. How is this fair!?
This is false. While the matchmaker tries to create games in which the teams are balanced and have an even chance of winning, this is not always possible. As a result, the ultimate goal of the matchmaker is to be as accurate as possible in predicting the winner of the game, thus allowing their matchmaking rating to be adjusted appropriately. In some cases (particularly for players new to that queue) win streaks can lead to a rapid increase in your MMR, which will cause you to face tougher and tougher opponents. Naturally, this will eventually end your win streak, but youll find yourself landing at a new level of more challenging competition.
Q5. Why dont we just put toxic players into a different queue?
This concept, also known as Prisoners Island, has been proposed multiple times and considered by the Player Behavior team. However, we ultimately determined that this model was not a good fit for League of Legends. Not only does it go against our goal of reforming toxic players, but it also feels extremely negative in the rare case someone is incorrectly sent there. Finally, since League of Legends is a free to play game, prisoners may opt to just abandon their account and transfer their toxicity to our newest players.
Q6: Why is duo queuing allowed?
While duo queuing can sometimes create less than ideal situations in matchmaking, the vast majority of our players actually prefer to play with their friends. While we have to be cognizant of this fact, we are also monitoring the solo queuer experience.
Q7: How does Riot feel about smurfs?
Riot does not officially condemn or condone smurfing. We understand the cases where players make new accounts in order to help teach new players the game, but are working on newer systems to provide a similar experience in a more positive way. Creating smurfs simply to beat lower level players is heavily frowned upon. Permanently banned players who make smurfs are quickly banned again. A permanent ban means that unfortunately, we feel that there is no place for you in our game.
EDIT: Should be up now
Long ago, toxic players who genuinely wanted to reform and could demonstrate their honesty were offered a second chance. I don't remember the specific details, but it had an extremely high failure rate (~95%) and we discontinued it.
The answer is actually somewhere between. While the MMRs exist independently, ranked teams are one example of where there is a little bit of crossover. When a team is first made, the MMRs of the members impact the starting MMR of the team. Once that's established, the team's MMR exists independently from each of the member's solo queue MMR.
It actually prefers to extend queue times over sacrificing accuracy in those cases. That's why challenger players can face much longer queue times than the average player.