Boards›[ARCHIVED] Miscellaneous

Some Thought About the Current Status of Matchmaking

GoodOlAlex·10/21/2014, 2:08:45 AM·8 votes·1,816 views

Before I begin I'd first like to point out the fact that there are many, many, many forums where the creator simply provides empty blame towards matchmaking because they lost to a high elo player. While this post may seem similar, and it sort of is, I intend for this forum to actually provoke greater thought towards the future of matchmaking. With that being said, let's begin.

The first thing I'd like to explain is how the system is designed to compensate for a high elo queueing with a low elo player. Typically the game will seek the average elo between the two players, and try to build a match around this factor. While in theory this sounds ok, there are some key flaws which cannot be detected by simple algorithm. However I'll touch on this later. There are also some cases where players spend an excessive amount of time waiting for players, and the system is pressured to create a lobby as soon as possible. When it does this it expands the pool of elo that it grabs players from, usually rising and decreasing by little amounts until a compromise is found. This is typically where you end up with a plat elo and a bronze vs high silver / low gold. Like the first example, there are some flaws to this.

Now that systematic algorithms have been clarified, it's time to begin analyzing the actual theory of why averaging elo should work. in theory, as you move up the tiers, you begin to correct previous mistakes that you used to make in the lower elo. For example, farming / last hitting. The higher in the elo ladder you climb, the closer to perfection you become (although it's nearly impossible to reach a perfect status). One huge factor people notice when viewing high elo vs low elo is the ability to capitalize on people's mistakes. Now this is the key factor that one should realize: high elo players are very good at exploiting mistakes and turning them into advantages. Well, most of the time at least. That being said, it is theorized that the platinum will exploit the golds, and the golds will exploit the bronze.

This brings me to the main point of this article, the ability to exploit mistakes is increased in higher elos than it is for the middle-end of the tier. So if you place a high platinum in a low gold game, in addition to adding in a level 20 player, the platinum is going to have an easy time exploiting the low golds. In theory, the golds should be able to gain an equal advantage by killing the level 20. This is where the mistake in theory is made. Platinum's are way better at explotiing mistakes made by gold's, than gold's are exploiting mistakes made by level 20 players.

When you have a conflicting advantage through this, it obviously causes some issues. Not only does the platinum player gain an advantage, but it causes the mood of the other team to drop. Nobody likes losing, and nobody likes admitting they're the reason they're losing. You'll end up with players throwing insults left and right, quitting the game out of rage, and people just unhappy in general. Now I don't know about you, but it frustrates me to see players insulting each other when in reality they both contribute to the failures and successes of the team.

In conclusion, I suggest that the matchmaking system provide a sort of curve in the matching of elos. Rather than averaging the two players elos, curve it higher by a small amount. Chances are, players will be able to compensate more rather than rage and insult. The matchmaking system has been complained about for a while now, and I try to formally submit my thoughts and provide an alternate solution rather than complain about it. If you see any sort of loophole or simply want to provide feedback, by all means do so. I wish to have a healthy discussion about the topic rather than submit a one and done.

TL;DR The matchmaking system averages a high elo and a low elo to create a "healthy" game. In reality, it is much easier to not feed than it is to kill lower elos, so this theory provides some obvious errors. For an answer to such a problem, I propose a curve. Take the average and add some additional elo proportionate to the gap in elos. Any feedback is welcome, so long as you wish to have a healthy discussion about the topic

23 Comments

deathTouch2k10/21/2014, 2:30:14 AM4 votes

Relatively one-dimensional look at matchmaking. I feel like it could have been more briefly summarized as non-linear progression skill vs. MMR.

This is one of many factors that can influence a game, and a relatively uncommon one at that in my experience with ranked. A sampling of more significant inefficiencies include:

  1. MMR is champion and role dependent (i.e. a player might be able to perform, on average, at a 1900 MMR level with hecarim jg, but only at a 1700 MMR level with vi jungle, and a 1300 MMR level with any support champ). If this player is shoehorned into a non-main role and matched against a team who are all playing in their main role/champion, the former team will be at a significant disadvantage. This is particularly a problem with choosing "1st time champs" or reworked/new champions. This is probably the single biggest inefficiency in matchmaking as it currently stands. A good example of this occurs at high elo, where relative differences between "main roles" and other roles for individual players may be incredibly significant. A team of 3 jg mains vs a team who has a more diverse main distribution is close to a guaranteed loss for the former team.

  2. MMR assumes that equal player effort is given for an individual game (i.e a plat 5 0 LP 1500 MMR trolling with jungle teemo will still be matched against 50 LP G2's trying hard to get to Plat 5 for the season).

  3. Failure to account for psychological impact of winning/losing games (i.e. I hypothesize a player on "tilt" losing the last 5 games in a row will likely have a greater chance of performing below their average MMR on their 6th game than a player who has alternated win/losses or one who has won 5 games in a row).

These are just the ones I can think off the top of my head.

SuperiorWalrus10/21/2014, 2:12:14 AM3 votes

Wow this was thoroughly thought out and very well stated.

GoodOlAlex10/21/2014, 2:42:03 AM2 votes

These are all pretty significant, and after reading your post I realized how much I had assumed the average. There are indeed many inefficiencies that can cause a lack or increase in significant gameplay, and I don't think I mentioned a single one. However the strong counter i hold against your first point is that while champion and role MMR may be less than average MMR, there are some solid knowledges that never vanish across roles. For example, not knowing how to play top lane does not deprive one of the knowledge of when and how to pressure dragon. This is usually the common downfall I see in matchmaking gaps. A high elo player will not only win their lane, but they will apply pressure to certain areas, create vision in others, and just overall out rotate the enemy team.

Tocosamuri10/21/2014, 3:17:43 AM2 votes

Very well thought out, hopefully they can improvise something for S5 because this season has just gone way down hill when it comes to matchmaking, IMO.

Just a thought, why would they have say level 20s playing with Plats. IMO this is a huge error, because yes, the Plat will be good and capitalize on the Golds, but it is unfair for the Plat to have to carry the 20s. There is a good chance that they don't even know what cs is, they struggle for cs, they don't ward or don't know proper builds etc. To make a more fun and competitive playing field, it should be 5 plats vs 5 plats. Just even skill. This is what really gets me, id much rather play against people my own skill, with people my own skill. It makes for more entertaining and intense games. I cant handle another game where I have to teach someone to use a ward, or what cs means. I don't want to play against people who shiit on me, and I don't want to play against people who I shiit on. I just want a fair and fun playing field, that's all I ask.

Matthias911910/21/2014, 5:45:30 PM1 votes

In conclusion, I suggest that the matchmaking system provide a sort of curve in the matching of elos. Rather than averaging the two players elos, curve it higher by a small amount.

Riot has stated they already do this. It's not just a straight numerical average.

There's also an MMR bonus applied solely for being in a premade group.

I've been in groups with a wide skill differential and it usually finds a decent matchup. Admittedly if it ends up with a Gold/Plat/Diamond laning against a sub-30, that one player's going to have a miserable time. But it can't always find an equally unbalanced team to put you up against, and even if it does there's no guarantee the two (or more) high-level players will be up against each other. With Team Builder you might have a better shot at that...

Drunk Rummate10/21/2014, 5:58:50 PM1 votes

I tend to get extraordinarily bad teammates when I duo with my higher ranked friends - so bad that even I play better than them despite being several ranks lower. Maybe it's just a coincidence but I think duoing ensures that the MMR of your team will be lower than the MMR of the other team, usually by a pretty significant margin.

Kingbeasty2410/22/2014, 1:27:46 AM1 votes

dont delete my comments homie

Kingbeasty2410/22/2014, 1:28:36 AM1 votes

gey

GoodOlAlex10/22/2014, 9:46:42 PM1 votes

@Aithos I'd disagree with this post. The reason why I still play this game after a year and a half isn't because of the community. It isn't because of the champions, and it isn't because my friends play it. It's because you achieve a sense of accomplishment from mastering a skill, and watching yourself progress through the ladder. I enjoy knowing that everyday I grow stronger in this game, because it gives me a goal to strive for.

When I'm placed in a game full of lower levels, and I simply smash them, I don't have fun. I don't enjoy it. Basically, the only thing that has been proven in that 20 - 30 minutes period is that I'm better than people who are new to the game. I know that, if I didn't then I would still be at their level. The same goes for the opposite side, why should I waste 20 - 30 minutes of my time playing a game against people who smash me? The only thing proven is that a diamond player is better than me, which I would agree with. I have fun in this game because I play it to learn. You don't learn anything from getting beaten 10 - 0 in lane, and then zoned out of every objective for the next 20 minutes. Typically this is the result of one fatal error that snowballed into a game, and while you can still detect this error and hope to correct it, the rest of the game is unavoidable.

Rather than be told I'm worse than a diamond but better than a low level, I'd like to play with players slightly better than me. I'll make mistakes, sure but i can learn from this. I'm not going to learn anything from getting smashed because they know the game better than I do.

GoodOlAlex10/21/2014, 2:28:26 AM1 votes

Any feedback is welcome

Mediocre Joe10/21/2014, 11:55:26 AM1 votes

Maybe Its just a coincidence but occasionally I will get a streak of games where every game included a feeder or troll the other day I had 4 games straight and no I didn't reque right away after a bad game I waited took some time watched a video then requed and the same thing happened. I would get some games where people just kept running into the enemy and saying that they just wanted the game to be over with so that is why they did it and then some games where people afked like 10 min in. 4 games in a row this happened and I don't understand how this can happen and everytime it was a different person it seems there is much more about this matchmaking than just averaging elo or maybe it is just coincidence that I got qued up with trolls consistently.

on a side note plats are bad and is by far the most frustrating league diamond was much more enjoyable than plat

Mikal Pikal10/21/2014, 3:09:49 PM1 votes

I have noticed the same issues...not even so much with high elo's queueing with low elos, but with uneven matches in general. Most of the time it is a stomp in one direction or another, but I am not sure of the cause. Maybe snowballing is too easy or team comp is actually very important but not recognized enough at my skill level...as in, had to picked different champs the outcome would have been different. I also think in normals, like others have said, this discrepancy could come from some players choosing their mains and other choosing new champs for the first time. At equal elo, the "new" champ will get owned 9/10 times. I appreciate teambuilder's attempt to fix this, but I cannot say how well it works.