Hello, ladies. I'm Xeroxable. We recently received a patch claiming changes to the matchmaking system that would place us with team members that have around the same amount of wins as each other. For the record, I have 1810 normal wins.
Well, today I got matchmade with 4 people (none of which were premade) that had less than 300 wins each. Their combined wins were less than half my wins alone. I would really like an explanation as to why I'm getting matchmade with these people.
Alright, ladies. I'm checking out Lyte's thread regarding the recent changes to matchmaking.
The most I can make of this is that I was a "rare" case of poor matchmaking or I've just been reported for unskilled player way too many times. I have bad games, but not that often.
That being said, how is skill rating determined? Win streaks? Loss streaks? Reports? Honours?
Skill is determined using a proprietary algorithm that takes into account a lot of different parameters; in fact, we're not even calling it Elo anymore because it has deviated a lot from the original matchmaking algorithms we first used when League launched.
Regarding the Number of Wins parameter, this is just one of the parameters the matchmaker uses when creating matches--this feature does not mean it is impossible to be in matches where the teams have a large difference in number of wins. The feature just ensures that in a large majority of games, the matchmaker will try to equate teams on skill and number of wins.
Why does the normal matchmaking algorithm even take wins into consideration? When matchmaking normals, why not just remove wins and take the higher of the normal mmr and the ranked mmr?
Assuming 2 players of equal skill (say, 1500 MMR), there are still differences that are not entirely accounted for by our skill rating systems. If Player A has won 1500 games, and Player B has won 300 games, Player A actually has a larger champion pool and rune pool--in other words, Player A has more flexibility in his ability to counterplay.
So, if we put Player A and Player B together in a match and they had to use the exact same runes and champions, they'd probably split their games 50/50. However, in reality Player A will win slightly more games because of his rune and champion diversity. Accounting for Number of Wins addresses some of this imbalance.
Isn't this then skewed by people who pour lots of money in game and people who hoard their ip?
In some cases, yes. Some players will buy RP to buy champions and artificially boost their champion pool beyond what you would expect from the number of games they have played and won; however, players cannot convert RP to runes so that aspect of it is accounted for.
The Number of Wins parameter isn't perfect--it just accounts for some of the imbalances that skill rating systems do not cover.
Lyte this is false. Player A got his 1500 MMR with a rune and champ advantage while player B got his with a rune and champ handicap. Since they are both 1500 that means they played at 1500 (equal ability to win) with their curret champs and runes. The 1500 is an indicator of their likeliness to win in their current state.
This means both are equally like to win if matched against eachother with their current accounts, but player B would have an advantage if they had the same champ/rune pool - this is the opposite of what you said.
I oversimplified the example for the discussion.
In reality and in the old MMR system, I wouldn't be surprised if Player B actually had the advantage if they had the same champions and rune pool. However, the vast majority of players unlock runes and champions (i.e gain experience) as they play games and improve their MMR (i.e mechanical skill)--these two factors interact throughout a player's career in League. It also gets far more complicated when you consider that maybe Player A got to 1500 by playing 50 champions and Player B got to 1500 by playing 1 champion every game. In this case, even if we gave Player B the same champions and runes, Player A would have more experience with them and would probably know of more potential counterplay.
Anyways, I agree my example was overly simple and it's more complicated; at the end of the day, we noticed in the data that when we added Number of Wins as a matchmaking parameter, we gained a meaningful improvement over just using MMR.
I thought the main reason the number of wins was added to the matchmaking equation was the starting ELO being the middle of the bell curve and thus subject to an extremely wide spectrum of player ability which negatively influenced games. ( 2600 ELO player makes new account and plays his first 2-3 ranked matches in the 1200-1300s ) OR ( Brand new lvl 30 player who belongs in the 500s plays his first 2-3 ranked matches in the 1100-1200s ) The 1200 players who belonged in 1200s didn't like having either of these players in their games. Anti-fun and all.
This is just another edge case that is covered by the Number of Wins parameter in the matchmaker.
@Lyte: Please explain this to me.
Well, today I got matchmade with 4 people (none of which were premade) that had less than 300 wins each. Their combined wins were less than half my wins alone. I would really like an explanation as to why I'm getting matchmade with these people.
EnLyten me.
Edit: http://i45.tinypic.com/257kcc2.png
The most I can make of this is that I was a "rare" case of poor matchmaking or I've just been reported for unskilled player way too many times. I have bad games, but not that often.
That being said, how is skill rating determined? Win streaks? Loss streaks? Reports? Honours?
Regarding the Number of Wins parameter, this is just one of the parameters the matchmaker uses when creating matches--this feature does not mean it is impossible to be in matches where the teams have a large difference in number of wins. The feature just ensures that in a large majority of games, the matchmaker will try to equate teams on skill and number of wins.
So, if we put Player A and Player B together in a match and they had to use the exact same runes and champions, they'd probably split their games 50/50. However, in reality Player A will win slightly more games because of his rune and champion diversity. Accounting for Number of Wins addresses some of this imbalance.
The Number of Wins parameter isn't perfect--it just accounts for some of the imbalances that skill rating systems do not cover.
This means both are equally like to win if matched against eachother with their current accounts, but player B would have an advantage if they had the same champ/rune pool - this is the opposite of what you said.
In reality and in the old MMR system, I wouldn't be surprised if Player B actually had the advantage if they had the same champions and rune pool. However, the vast majority of players unlock runes and champions (i.e gain experience) as they play games and improve their MMR (i.e mechanical skill)--these two factors interact throughout a player's career in League. It also gets far more complicated when you consider that maybe Player A got to 1500 by playing 50 champions and Player B got to 1500 by playing 1 champion every game. In this case, even if we gave Player B the same champions and runes, Player A would have more experience with them and would probably know of more potential counterplay.
Anyways, I agree my example was overly simple and it's more complicated; at the end of the day, we noticed in the data that when we added Number of Wins as a matchmaking parameter, we gained a meaningful improvement over just using MMR.