The MMR Issue and my Answer to it.

General Elephant··5 votes·1,494 views

The Issue:

We all know the ranked system has trouble matching evenly skilled players. It is also know for its "flawed rating system" where players feel like their abilities are going unappreciated or undervalued. I agree with this, to an extent. The real issue I feel is that the invisible "MMR" rating is not showing players enough reason to believe that number is right. It may be much more accurate than people believe, but the issue is that people refuse to believe it because of how simple the idea of w/l = MMR. The fact of the matter is w/l is definitive, and all encompassing. If Riot decided to break it apart and say "hey, now your rating is dependent on these qualities as an individual player!" then a lot would be left out in the form of more abstract abilities, ie: teamwork, positive attitudes, and over all decision making to influence the outcome. So, now that I've stated why Riot has good reason to leave things the way they are, let me give them a better reason to change it.

My Analysis of the Issue:

Some things in life are too complicated to be defined by a basic number. Take your credit score for example, there are three agencies that each determine your over all credit reputation. I feel that MMR is the equivalent of saying "you should trust just one of these companies to determine your credit score!". If this were to happen surely people would be suspicious of their own rating and the validity of it, and because it is a very important value, people get very defensive, because it defines them and how good they have been doing in terms of loan payments and credit management, which follows them for the rest of their lives. This is not as true in relation to MMR, but still valid because shorter time spans are more impacting in this video game experience because the total time to accrue a good value is much smaller. Because of this, the diminishing returns from the number of games played leaves players with very little efficacy in their games impacting their rating. These factors are what make up . All of this leads to several points that a crucial for a good rating system for player performance. The first is that MMR is too basic and independent to be credible to the player base. The next is people take it very seriously, so trust in this number is extremely important, and in the current state is very lacking. Finally, with such small time spans, people feel helpless after so many games played in the season, they feel their true skill is not being reflected because of the lack of mobility due to diminishing returns.

--please read the above sections if you skipped them and went right to the solution, it really helps emphasize the reasons behind the changes I think would benefit the current system.

My Solution to the Issue:

We should have a cumulative composite score, that is influenced by 3 categories of results from each match, and that each category is determined by several related factors which are given different levels of importance.

Confused? Let me Elaborate:

Imagine we call this "cumulative composite score" your MMR. After each game, three numbers are combined into a +/- value based on the average distance from a base number, which I will use 100 as an example. So each number is around 100, and they get added, then that total is divided by 3, we will make this number be called "game rating" or GR from here on out. The +/- value is determined by the GR's difference from 100. This is then applied to a starter value of 1000 (or maybe even your current MMR in ranked right now). Ranks can still be applied, ie not ranked <=499, bronze 500-1000, silver 1001-1500, gold 1501-2000, plat, 2001-2500, diamond 2501-whatever holds non-challengers, challenger x% of the highest scores.

Now you may ask yourself, what determines these oh so important categories? the answer is simple. Imagine you're taking 3 tests, where each grade starts at 100, and depending on how you stack up to everyone else, you get points, or lose points.

Let me demonstrate what would happen at the end of a game:

first category: "gold"- you got 35/30 points for cs, you got 20/30 for objectives, and 50/40 because your team had more gold at the end, so your total was 105. your 35/30 is determined by amount of personal gold accrued through individual creeps, your 20/30 is determined by the amount of gold your team accrued through objectives(more or less than enemy team, +/- 10), and your 50/40 is a result of having more team gold.

  • this does not include gold acquired from kills/assists because those are PvP related gold incomes. CS is an issue for supports and junglers, and it has been stated Riot is doing their best to solve this for Season 4. This does mean they will probably get a few points less, but out of 300, 5 or 10 is negligible, and can easily be made up through objective gold, and team gold.

second category: "PvP"- you get either 30/40 for your Kills, or 40/40 for your Assists because your Assist score is higher, you get that for that 40 point section. you get 20/20 for getting the middle amount of deaths on your team, and you get 50/40 for having a higher K+A/D than the enemy team(you get either 50 or 30), so your total is 110.The value for your Kill/Assist, as well as Death score is determined by adding or subtracting 5 points based on how many positions from the 3rd best on your own team, assuming it is 5v5. In the event of leavers, the bottom spots are dropped.

  • this allows supports to gain points for lack of deaths, as well as most assists. It also promotes teamwork because assists now give more value to supports and junglers. It allows the team to say "we are going to get summoners A and B all/most of the kills, and we will get all the assists.

third category: "W/L"- simple enough, if you win, you get 130/100, if you lose, you get 60/100.

  • this seems very steep yes, but overall, this is 10 points down for a loss, and 10 points up for a win, and it can fluctuate based on over all performance in the other categories.

My TL;DR: (you really should read it...)

MMR needs to be based on more than just w/l because of how complex the game is. I think having a tiered point system with the largest impact being w/l, will allow some of the over-looked aspects of each match to show through and have a little impact on over all rating players are given.

I am glad you read this, thank you, I would love some feedback and questions. Please remember to up-vote and post if you agree! I want this to hang around as long as possible.

35 Comments

MackleDoge3 votes

I'll be upfront in saying that I really don't play a lot of ranked personally, because in between work and school I'm just not that into it. However, I personally have experienced many instances of drastically disparate matchmaking and have witnessed it even more on streams or spectating friends that do play ranked. So I certainly agree that there are issues with the current manner of matchmaking (especially with premades because they seem to murder the matchmaking system).

That being said, I'm not certain that your solution would solve the problem entirely. I do think that it has merit enough to be explored (Hey Riot, you guys are throwing Team Builder into PBE, why not try an altered MMR system like this?). I would have to see more solid, actual game results and evidence from statistical spreads over multiple games to really subscribe to this being a full scale implementation though, just for the sake of rather sticking with the evil I know then implementing a system (that is fantastic in theory to me) that doesn't have substantial backing to prove it solves the problem and it just degrades into another failed matchmaking system.

TL;DR I think your concept is worth testing in PBE but I would want significant evidence it improves matchmaking before actually seeing it implemented over the current system.

ploki1222 votes

Basing scores on kills/assists/deaths/cs is very wrong, since supports don't get cs, and junglers get low cs. Also, the same applies for KDA, a split pushing champion will necessarily have less KDA, but overall he will still be helpful. In the same way, deaths are frequent on initiators, which means that people would take a lot less initiators.

General Elephant2 votes

Is it wrong though? Yes, a splitpusher will have less kills/assists, but he will gain the top value for cs and will snag a bonus 10 for most objectives assumig he does his job. Lose 15, gain 25, comes out ahead if he does what he was supposed to effectively. As an initiator, they need to learn the limits, they can't just dive a solid team, that would be stupid, and they should not get as many points for it if they do, and seeing as it is team dependent, the loss/gain cap is only 10, so it isn't like you lose out if one of the categories is the lowest. Dying as anyone hurts the team, and right now this isn't shown well enough. If you do decide as the initiator to go in, you should be much more wary of when you do. Thinking about your death count just reminds you dying hurts your team, and your score. A death essentially gives the enemy more gold, and gives them time where you are not playing, meaning they can do more.

This set up punishes being stupid, and allows recovering for points as each role. The point of this is so you can't get ALL the points, but rather a healthy balance that tells you where you need to improve.

ploki1222 votes

I believe it is wrong, since there should never be an incentive that goes against the best case. Sometimes, dieing won't net your team a real advantage, but it will sill let you win more.

General Elephant2 votes

dying when it is for the good of your team nets a huge advantage.... if 3 of your deaths is the thing that allows you to win... losing 10 points is COMPLETELY worth the 60 point difference of W vs L. Also, if your death allows your team to net dragon, then you could possibly get the 20 point difference from more objective gold... there is more to this system than just the kda portion... much more. If you do all around poorly, you deserve to lose more than your team mates, if you did all around better, you deserve to get less points taken if you lose. That is just how it works. Let the good players maintain, and bad players sink.

ploki1222 votes

Let me phrase it differently. If your actions always net an advantage toward winning, doesn't that make you win until you are just as proficient as the other 9? Meaning that the W/L is actually an excellent way to calculate MMR?

If all the things you calculate truly does end up with you raising to somewhere where your win to lose ratio out of 3 billions games would average to roughly 1 (aka your intended level), then why does win/loses don't?

Or are you telling me that your intended level can be at a level where win and losses aren't, on average, 1 for 1?

Basically, what does your new system truly promotes and what does it truly fix?

General Elephant2 votes

Given infinite games, you can assume everyone will reach this true 1 to 1 ratio, but due to resets, diminishing return, and anomalies in game, this golden ratio is never truely reached. It is also safe to say players skill levels do not remain constant, nor are they equal at every match up, so trying to assign a value based solely on W/L will leave your rank to chance by an insurmountably large number of small variables. The current MMR system makes blanket assumptions that I am not okay with, which are that you will reach an equilibrium in your w/l. This is not true, and anyone who has bounced between ranks knows how frustrating it is. It also assumes every game is a 5v5, which is also not true. Leavers impact the game because they are statistical outliers, and are rarely balanced to the point where it doesn't matter. I feel like the system assumes everyone is going to play an extreme amount of ranked games, and because of this it is hard to move up and down from where they are placed.

ploki1221 votes

So you answered what you system fixes (or tweaks actually), which is simply making the MMR changes happen faster. Now what does it promotes?

Right now, the current system promotes that winning is the only thing that matters. The better you are, the more you will iwn and the more you win, the better you probably are.

General Elephant1 votes

a straight w/l also promotes things like elo boosting... if the system was more fluid.. the companies trying to do this for money would no longer exist... yes it makes the changes happen faster... that is the point. if you're on top, you need to stay on top, if you get placed low, it shouldn't take you 400 games to get up to where you should be with anything less than an 80% win rate. It just seems wrong to force people to be ranked and not allow mobility upward and downward. Also the current system does nothing to allow proper game techniques to be reinforced... It simply says "win however you can, and you will be rewarded". the system I proposed give an actual value to portions of the game that are important in order to be successful.

4thaGe1 votes

Right now all the system promotes is grinding. And more grinding. Yeah, after 1000 games maybe all those teams of feeders starts to balance out. But that is a long journey that doesn't feel fun.

If I felt like I could still improve my score by grabbing some objectives and fine tuning my CS despite my team being down 5 - 30, it would make the game much more rewarding.

BrunikRokbyter1 votes

I main Jungle Sejuani. I never have lower than 3rd place CS, I always have high Assists, and it is EXCEPTIONALLY rare that I don't have fewest deaths on my team, or even both teams. It is also rare that I don't have highest damage TAKEN for my team while keeping fewest deaths (it does happen though).. this is even durring losses.

I am a dive in, jungle initiator, and I don't have this problem. I know my limits because i have 100+ ranked games as Sejuani jungle, and I hyper analyze every death to find out why it happened and if it was necessary. I have also been climbing rank mercilessly since I started actually caring.

Who was, during most of season 3, the highest KDA because they didn't die? Ohh right.. tanky initiator... Meteos. He knew his limits to perfection... even when he wasn't on Zak, it was rare for him to die.

ploki1221 votes

He had the highest KDA, yes... but only because he didn't die... Bump everyone's death counter by 1 per match, and suddently Meteos falls down a few places... so yes, he didn't die. But that's only because his games were stomps. If you're a tanky initiator and you win teamfights, you probably won't have to die.

As or your overall experience, I never said that initiators will 100% get worst scores than other people. I'm only saying that since most initiators have way more initiation tools than disengage ones (see that Sejuani doesn't fit this), it is easier for them to die, comparatively to their allies. If an initiation go worng, ranged champs bail out and you, end up stuck behind. If you push forwar ot get your carries that double, you may end up trading 1-2, if you have to go smite steal, you may end up trading baron for 1 death.

All As or never lower than 3rd, that only means that you pretty much always outcs at least 1 persons, since support will nearly always have 5th cs... Since most probably 1 of you lane will lose, and that you can hardly lose your lane, that makes you a safe 3rd. safe 3rd with still a lot less cs than your 2 winning lanes (unless you split push)...

Also, I do believe that in most cases, when the initiator dies fightig, he ends up doing a better job than when he runs away. Since you are tanky, and have CC/brust (since you're an initiator), you can probably save 1+ carries by dieing. The fact that you always have the least deaths point me toward the conclusion that maybe some carries died because you saved your life...

BrunikRokbyter1 votes

Actually, you will find as you play tanky initiators, a LOT of the time you can save more lives by being at < 1/4 health running away than still in the fight. This is because the enemy sees low health and tunnels on catching you (ESPECIALLY if you have done the job of annoying them to death all game and not dying, as any jungler should). You can bait kills, force extended chases, and draw more attention to yourself at low health than at none. I have gotten more kills for my team running with 200 health than i could have possibly gotten or forced to have happened had I actually stuck around.

In really high ranked and levels of play, these tactics may not work, but i'm not there... yet. A tanky initiator, ESPECIALLY one that knows their limits, can have good stats, even in a loosing game.. you just need to know your limits, and your team mates need to trust you (which happens through you having a good early game and showing them you aren't incompetent)

All of that to say, I have always been frustraited at loosing the same amount of ranking as that teemo that was 0/15/1 even though I was possotive KDA, and all my other stats were solid... and most importantly, I did my job. There should be some disparity.... even just a little (same with winning).

ploki1222 votes

This is a reply to

Right now all the system promotes is grinding. And more grinding. Yeah, after 1000 games maybe all those teams of feeders starts to balance out. But that is a long journey that doesn't feel fun.
If I felt like I could still improve my score by grabbing some objectives and fine tuning my CS despite my team being down 5 - 30, it would make the game much more rewarding.

Source


So yeah, sure it may feel more rewarding to win more points because you feel you did better. That,s a given.

But will you feel good after winning only 4 LP because you were on a bad game?
Will you feel good when the enemy team drag it down since you cannot overtake them as of right now? So they just get kills and cs to boost their LPs?
What if you've got an AFK and 2 of your teammates refuse to surrender? How enjoyable is gonna be your 50 minute losing 4v5?
What if you end up on the other side? you're getting camped, yuo're losing the game, and you lose the game... -35LP because you struck a bad onee?

Ovarall, I think I said what I had to say, and I don't think this MMR system is healthy for the game.

General Elephant1 votes

I should have clarified what would happen in the event of leavers (i never said leavers would simply lose a flat large rate, probably 20 points). basically the bottom spots are dropped in the cs, ka, d, and ka/d has the leaver's stats ignored, and ka/d is increased by the average of the remaining players ka/d. essentially removing the places or mos points lost in every category, so the most you can lose is 5, not 10, assuming you did the worst. if you have a bad game, then you only really earned 4 lp, if you had a good game but lost, you deserve a less severe loss, it makes sense to me. Yes there are flaws. you have done a much better job at laying out issues you see in this design, however I feel like there is a better, more efficient system out there. I spent an hour and a half concocting this entire set up... it was meant to show what should be done as opposed to leaving the current w/l is the only thing that matters.

RiotRiot LilyHaze1 votes

If a system cannot judge players accurately across all roles, it should not be implemented. Riot has not said that they want Supports to get more CS in season 4, but instead to have more gold to spend. By running a system that discriminates based on CS, you unfairly punish supports and junglers, as you said. This only encourages supports and junglers to farm, perhaps stealing CS, to increase this rank.

Additionally, you encourage players to not work together, instead farming their lane long past laning phase because they don't want to risk losing points.

Similarly, basing things on KDA is problematic to many roles. While supports can get many assists, not all do. Sona will skyrocket in this system because her aura is always on around her. Leona, on the other hand, might not be able to hit all enemies, getting less assists. Additionally, top lane, a lane with less game-wide kill potential, will get a shorter end of the stick here as they are unlikely to be able to participate in many engagements in the first half of the game.

And now you've created an awkward scenario for junglers. If they don't gank early, their KDA score goes down. If they don't spam farm, their CS score goes down.

General Elephant1 votes

The point of this system is to give them choices... a healthy balance of farm and ganks is ideal, which lands them in the middle/lower end, but by staying on par level wise, and bouncing between cs and ganks helps your team the most. If you do a good job as jungle, then your team will be able to take more objectives, and thus, you will win, if you fail at your job, then you lose points. It brings personal accountability into the game, which prior, it has had none. This also PREVENTS ELO BOOSTING ALL TOGETHER. :)

as far as supports go, they take a hit in CS yes... that is one of their weaknesses... but if they do well in lane, they will recover those points via most assists, and least deaths. It also prevents the whole "i will die to save you sen-pai!" feeling supports get when their adc is in trouble, which more often than not leads to an enemy double kill.

SmokingPuffin1 votes

Supports die more than anyone in high elo games. Supports should absolutely die to save their carry from death, if this is the only possible way to prevent the carry from dying. Failure to do so is failure to maximize your team's winning chances.

I have a hard time believing you will ever construct some statistical model that accurately captures game impact of a support in proper relation to a laner.

Also, this definitely doesn't prevent elo boosting. I'd say it probably encourages it. Suppose I want to elo boost you. I have you pick top, then I pick jungle and camp your lane. It's easy to make you look good statistically.

General Elephant1 votes

if you support, and you choose to die for the good of your team... you will still net a huge gain... if you lose 10 points for deaths, and 10 points for lowest cs... thats okay.. because those 20 points will be more than made up for by team things like objective and team gold, and K+A/D. (also winning matters a lot...)

ploki1222 votes

Do you honestly belive that :

  1. Your system prevents elo boosting?
  2. You system judges all orles equally?

In my mind, Top laner's job in lane, in most cases, is to soak up ganks from teh jungler since he can. Whereas, the ADC's job is clearly 100% to farm and kill to get ahead. One of them will most probably end up with more CS and/or kills. Otherwise, they will both probably be on par for everything else.

As for the jungler/support thing, jungler gets more farm, jungler gets more assists in most cases... how does teh support compensate.

Lastly, yes, a support dieing may net 6 towers and 3 barons a 22 kills for his team... but overall he will still "lose" elo simply because he did what was best (in this case, lose means win less). So no, supports are not promoted in your style of rewarding, and no initiators are not rewarded... Mostly everyone ends up being penalized for not being selfish.

Oh but wait, not being selfish will net you more wins, and then you will rise by not being selfish, you simply won't win as much.

Which leads to ELO boosting... If you still win by going against the system, the main way to win fast is to get a much better person to carry you, while playing selfishly. that way, you have a selfish score in a win.

Also, I would love to know how making MMR move faster prevents boosting. Since it's all absedon league, MMR's slow movement is a limiter to abuse cases. You can't get boosted from Bronze to Plat in 1 week, unless you play a few hundred games.

Sorry, but this ssytem is heavily flawed and promotes selfishness over teamplay, and plays over wins.

SmokingPuffin1 votes

No, it won't be okay. All of the points you mention for objective gold, team gold, KDA, they will be available to the other roles as well. In fact, given that supports are the role that die the most in high level games, they are more available to other roles than to supports.

If you're telling me that support will consistently get fewer points than other roles, people will fight to not play support to an even larger extent than they do now.

The absolute last thing I want from a rating system is for it to increase the amount of drama in the team. I don't need another reason for people to play selfishly rather than doing what the team needs.

DocSmoothJew1 votes
General Elephant2 votes

I said I was going to make my own post with my answer, and I did. It is not spam.

DocSmoothJew2 votes

Sorry, that was a bit harsh of me. :(

Math Test1 votes

I think this links in with the "League" system overall. Before this new system, the unit of measure of skill was ELO. These days, it seems that MMR is the new elo. Even if you are Plat 3, if you get matched with Gold 5s then your "hidden elo" or MMR is clearly not Plat level. It makes me wonder what the point of Leagues are except as a psychological "feel good" type of system with MMR being the hidden ELO in the background. There definitely needs to be some improvements to MMR. Maybe these improvements would help make the League system actually means something, but I am not sure. Just wanted to get that in there that MMR also has consequences for the League system too.

Zynz1 votes

As far as I'm concerned the problem with mmr is it's hidden. Show us mmr that's it. But I also much preferred the elo system to the leagues system soo

SmokingPuffin1 votes

I hate this idea.

You're going to inevitably reward some roles more strongly than others and some playstyles more strongly than others.

This will cause players to make decisions that aren't the best for winning the game, but are the best for maximizing their personal stats.

General Elephant1 votes

this is set up in a way where doing what is best to win is rewarded, while being selfish might delay losing points, but being selfish will lose you more games over all. Personal stats only account for 40-ish variable points... out of 300. The rest is based on team related goals such as team gold, team objectives.. team K+A/D (meaning assists will net you this faster).

frostygutz1 votes

Personally i enjoy the current system. I do not think that is that flawed and if there were "flaws" it would be the fact that wins/losses in promos don't affect MMR. They are essentially a 3 game "void" that tests if you deserve to continue. Even then once you win the promos you gain mmr so it's not that big of an issue. Me having 6 losses in gold 3 and 18 wins means very little and i am not frustrated by this. I look at my mmr on lolking and think "this is good but i want better" and keep trying. I agree not hiding mmr makes progress felt more but i think the current system is great. You don't feel the non-stop chaotic fluctuations like elo and you still have promos to make sure players earn the tiers they are in. I feel an incoming down vote on my first post because it is a different opinion based on the voting on this post but just some insight. I like what we have now, maybe showing elo and tiers may be what we need and introducing a new system may just take time out of more productive changes such as reworks, magma chamber, team builder and pre season changes.

I really respect what you've done and the thought you put into it though. As a support main i enjoy this system and disagree with any who states it's impossible to carry as support.

Dragoon Feld1 votes

the only issue ive had in ranked. is when i finally climed from B1 to s5, i was +20 games. the MINUTE i hit S5 i was matched with low silvers/LOW bronze, playing compleatly differently then the HIGH silver, low gold i was getting in my B1 promos.

thats the only flaw and agrivating thing ive seen in this system, it intentionally pits you to lose which is not right in the least.

OneFreeMan1 votes

I sort of disagree.. I worked and worked and my work paid off when I finally got my promo and then WON THREE IN A ROW! I know the system isn't perfect, but mmr rocks compared to elo... imo...

m641 votes

I think the biggest issue (I've made a new account because of MMR) is that when you're climbing between tiers, silver/gold gold/plat plat/diamond etc. You tend to have horrible teams who're stuck in the 5th divison of the upper-elo that troll because "i cant be demoted". If riot fixes that and changes division 5 to any other division, you'd see alot more balance in matchmaking.

surferhtc211 votes

I think that matchmaking is also flawed in that a player with 8 ranked wins, noob to ranked, gets put into a game with someone who is trying to climb and has 200+ wins. I climbed from bronze 5 to silver 3 and dropped back to silver 4. I am not placing blame. I make bad decisions too, but I keep getting 1 or 2 players each game that have very little ranked experience and wont take help from team mates. This keeps leading to one or two 0/8 or 1/6 players on one team and another team filled with silver 3's. How is that equalized matchmaking?

I like your idea on the MMR. Part of the problem is that you cant see it. We see W/L and LP. I use Lolking a lot. I was holding around 1300 but this past month it has been hard to piece together any wins so I'm holding at 1220. I feel that if you can at least see a number you feel more secure.

Red Ryu1 votes

I don't like it, as a support man I'm gonna get a lot of assists but may also die a lot if I am trying to make plays that my team needs to win.

This basically promotes people wanting to get Top/Mid/ADC because they will get CS/Kills/Assists and then some more pros, where as supports get just assists and only deaths.

MrSc0tty1 votes

Support mains get robbed in this system, as well as junglers. You can't say that team gold will make up for it, an LCS level support simply will not have as much Cs or gold than a gold level carry. In a game you win, through this system, the support would always be rated as the "worst" player on the team.

More from [ARCHIVED] Miscellaneous

More by General Elephant