Boards›Gameplay

The Real Problem with Ranked: Matchmaking.

Too Practical·9/12/2016, 11:32:38 PM·5 votes·522 views

I've been documenting my ranked games to discover what's wrong with it. Was it toxic players? Overpowered Champs? Overpowered Items? Red side? Blue Side? Trolls? A lot of people point to either of these things as the sole cause of the reason ranked is a terrible place.

League is a team game, and no matter how good you do as an individual, you will always be judged on how well the worst player on your team does. "A team is only as strong as their weakest link" if you will, and so many people are frustrated that no matter how good they do if their teammate plays poorly then it affects the team as a whole. So you get the beginning of a toxic player that then can turn into a troll because "the last game they went 13/0, but mid went 0/11, so I'm going to troll to vent my frustration and lose your LP so you can feel my pain". This then creates a chain of toxic players, where the guy that did good in that game turns toxic against the troll player, and then goes on to troll himself. (Insert Naruto Pain ARC here)

What creates the 0/11 paradox? Sometimes its general lack of skill, maybe just a bad game, but I believe that most of the times its just a difference in skill with your enemy laner. This means that your general skill at the game can be adequate, its just that your laner is simply better than you at it. From an outsider's perspective (or toxic teammate's) if you lose your lane, you're just bad. What's totally out of people's minds is the possibility that you're really not bad, its just that laner is simply better. By adopting the view point that the losing laner is bad at the game (and telling him that), totally discourages the potential outcome of that person of making a comeback with the true skill he actually has. (insert toxic players lose x% games tooltip here).

So what's the real problem? We'll from my analysis, its the matchmaking system. The majority of games are created with improper player skill balance. Ask yourself this: when's the last time I've had a close game? are most of my wins AND losses usually a blow out? Looking at the statistics, the answer for the latter is yes. Close games are rare to come by. Mathematically, players of similar skill level should not be blowing each other out of games. The games should be close, not 40v20 blow outs. Gold should be near identical. Right now, this is not the case. The matchmaking system has a flawed balancing system. Certain teams are given players that out class their enemy's, and it is those teams that are usually winning. Again, look at your matchmaking history, most of your wins and losses should be blow outs. This is not matchmaking balance.

This then leads me to the 50/50 paradox. If you look at the stats, the average player wins 50% of their games. From a superficial mathematical perspective, this is called "balanced". However, if you look deeper @ matchmaking you can see that blow out games are inherently imbalanced. True balance would be 50/50 winrate average, with won/loss games being close.

So what's wrong with matchmaking balance? It's not precise enough. I ASSUME it uses a point system to "balance" the teams. For example: S5 =1pt, S4 =1.2pts, S3 =1.4pts, S2 =1.6pts, S1 =1.8pts, G5 = 2pts, G4= 2.2pts etc etc. As an assumption, teams would be created from a pool of players to = each other in terms of points. Blue side would get a +1pt for having a camera advantage. However, in the context of skill level in this game, this system would be flawed.

Lets use this a recent game of mine as an example:

Red Team = 11.6

ADC Silver 1 = 1.8 Top Plat 5 = 3 Mid Silver 1 = 1.8 Jungle Gold 5 = 2 Support Plat 5 = 3

Blue Team = 10.8 + 1 blue advantage = 11.8

ADC Gold 3 = 2.4 Top Gold 4= 2.2 Mid Gold 4 =2.2 Jungle Gold 5= 2 Support Gold 5= 2

Here we have the team skill point totals nearly equal. However in this specific game blue team destroyed red 39 kill to 18, with an 18k gold lead. This is what I mean by precision. The point total system doesn't take into account the context of in game skill difference:

ADC: Blue Favorite by .8 (nearly a whole elo) Top: Red Favorite by .6 (over half a elo) Mid: Blue Favorite by .6 (over half a elo) Jungle: Even Support: Red favorite by .8 (nearly a whole elo)

*note: blue side advantage included

If we were to delete the laner that lost (because they do lose hard), team fights would look like this:

Red: Support, Top, Jungle VERSUS Blue: ADC, Mid, Jungle

Put quite simply, red team's 2 carries were outclassed. Red team had no damage output compared to blue team, and red lost by a blow out. Despite having 2 Plat players, red team also had 2 silver players against a homogeneous blue team of Golds. And since the nature of this game is teamwork, a team is really only as good as their weakest player, and that floor was much lower for red team than blue team. Hence you have a flawed matchmaking system despite, a theoretical equality in total elo.

10 Comments

Karunamon9/12/2016, 11:41:34 PM1 votes

Agreed completely (and this is an awesome writeup!) - nobody likes getting roflstomped, but that's how most LoL games end. I've had 2, maybe 3 lost games where I wasn't the least bit salty at the end because everyone was damn near even on skill, the game was very close, and by the end, we just got outmaneuvered ever so much.

In other words, it felt like an honorable defeat rather than a steamroll.

LoL would be a totally different game, both community wise and in gameplay, if this were the rule rather than the exception.

I think Riot's in a no-win scenario here, though. The matchmaking system can only evaluate so many different variables, and the primary one has to be skill by way of ELO. If they tighten down the skill matchups, queue times are necessarily going to shoot up, and that just leads to even more endless complaining. And that's before we bring premades into the mix - how do you matchmake a D1 and his three B5 buddies?

That said, I'd never (swear on me mum!) complain about 20 minute+ queue times if Riot could guarantee a close skill match. if the game is close, it's infinitely more fun to play, with much less room for toxicity, trolling, and all the other stuff that winds up complained about on PB&M.

AdeBug9/12/2016, 11:57:23 PM1 votes

I agree with the venting in chat part, and complaining doesn't help, me being guilty of this as well. I try to be friendly and positive, MOST of the time, because I saw that helps overall. But if you ask a bad laner to farm under the turret and you still see them 1v1 under enemy turret, 3 levels under, I can't help myself say something because it looks like a feeder. If the players are actually bad, they should play safer and let the team help them out. Let yourself be carried ffs. If I'm camping your lane trying to gank, then back the fudge up for 2s and miss one lousy minion so I don't have to camp forever and miss my farm, and I don't want to dive for you either, cus you can't take 2 steps back. A true team would help each other and let be helped. If one lane is weak, then others can roam to it to pick up the slack. But some top laners are stuck into getting their next tower sooo bad, they won't stop til inhib. What I have against match making, besides the smurfing that totally ruins the mmr, is the premades. They should only be put against other premades. And unranked is NOT silver, that never worked for me, to have someone in their placements against other silvers is just unfair, the only time that pans out is when they're smurfing. And don't get me started on people in Bronze 2 that don't know the jungle buffs, or supports that don't know S can help them stop auto'ing and ks'ing. There should be a basic test to pass before playing ranked.

Too Practical9/14/2016, 1:01:16 AM1 votes

BUMP

Too Practical9/14/2016, 6:32:23 AM1 votes

Bump

UpDownLeftRightAB9/14/2016, 8:13:46 AM1 votes

I just think the dynamic Q has thrown more confounding variables into the matchmaking system than it can take. Mechanically (top cs, positioning, etc), I am normally the top player in my games, but I'm not the one who ends up deciding the win because my main roles are heavily team dependent. In bot lane, normally it seems to be a fair matchup for me, I'll only be even to at most 30-50cs up. What I notice is the main deciding factor is just how premade each team is. If I start getting 4fairly man camped from 3 minutes onward, there isn't to much I can do other then ask my team to step up in their lanes, and that is normally where I can start to see how a game will play out. Such as is my mid/top able to win their lanes with no jungle pressure on them? Does my jungler/mid look to countergank the roam/enemy dive? Do my teammates even communicate? playing the predominantly duo lane as a solo player I think leaves my team at a disadvantage when it comes to typical premade team roles, which are either a 3 man bot/support/jungle or mid/top/jungle. I can have a good premade on my team but largely they aren't focused on protecting me even if I get ahead, they are more about protecting each other, which means it becomes a protected adc vs an unprotected one. I don't think it needs mention when I face 4 and 5 man teams that I get rarely any teamwork started even though I attempt (Ashe and Jhin are my primaries) and set up picks/help teammates pick, but rarely does it convert to what a coordinated enemy team can do more often. It seems like being the solo player in a game of premades gives me more scewed matches then fair matches. It's pretty much always a complete shutout or stomp for one team by 10 minutes, I haven't seen any really epic comebacks, but I've seen a lot more toxicity.

jaymc11309/14/2016, 8:33:41 AM1 votes

The whole reason the ranking systems have been dysfunctional for years is indeed the problem you have identified. I identified it too. Nearly 5 years ago now. I've been bitching about sense, gotten now where with it either.

What lies at the core of the problem is this: We use a ranking system that ranks individuals on a ladder without ever taking into account a single individual skill metric that pertains to any individual's skill set. Naturally this is going to cause very poor matchmaking. What we need, and have needed for years, is a sabermetric style rating system for players that takes into account ever action they take on the rift (including inaction). The system already tracks where each kill happens and at what time, when and where which objectives are taken, some of a players individual skill metrics are measured by Champion Mastery (a system I originally designed 4 some odd years back to help with the exact issue you describe, but Riot never implemented the functionality of it, it has no effect on MMR gains or losses), and what role and champ a player was playing.

What we NEED from the league ranking system is a system that measures how well players perform in every aspect of the game relative to the champion they are playing and the role they are playing with respect to expected performance when compared to a specific skill range. This algorithm would be like Wins Above Replacement player as is used for baseball and we could even modify it based on the aforementioned criteria to create a Wins Above Replacement player style system for League.

So how would this help improve matchmaking? Well, now that we are accounting for each players individual skill set as well the general skill range with MMR matchmaking will be better able to find similar opponents. Instead of having a scenario where you mentioned in the OP where various lanes are almost an entire tier apart in difference we have each lane having its own general skill level (MMR) and specific skill set (the WAR style component) matched against the opponent playing that lane for much closer skill level gaps between opponents with respect to the role they are playing.

This type of improvement to matchmaking would help ranked play significantly as the teams would be much closer to each other in relative skill and the individual matchups between role opponents would be much closer as well, making every phase of the game more competitive. There would be significantly fewer stomps with a system like this, and the displayed skill level of players would be far more accurate.

It's really what league needs to be a fun and competitive game again, a WAR style system for measuring a players worth and value to a team that is taken into account for matchmaking and, possibly even, used to help rank players on the ladder.

As long as we have a game that ranks players individually on the ladder without ever taking into account any of these players individual skills the system will always fail to perform its function as it has for years. Its simply impossible to measure a single individual's worth by W/L record in a vacuum when he doesn't play with the same players every single game. You can only have accurate measurement of skill with W/L in a vacuum when the Unit participating in the competition stays the exact same every game. Since the ranked queue is NOT team 5s and you ALWAYS have different players on a team this is just impossible to accomplish without modifying and fixing the system to account for it.