Confused by matchmaker? Join me to learn more about ELO and MMR systems.

I will run out of room fast so pardon a lack of introduction. I'm going to hit what I consider key points and will reply to questions as best I can.

How do you say Elo? Its from a name not an anagram for anything. So you pronounce as his name. When I checked what I was given is E-lew. Not E-L-O. Hungarian is hard. If some one Sneezes the response is Egészségedre! (to your health is the translation) Which I know because when I get curious I study and research things.

What is Elo/MMR? A principle and then the math to make it work. The principle is that Skill in a skill game has a direct impact on a players ability to win. Thus in a game of skill we can track wins, loss, and draws to rate the player for a rough but workable estimate of their level of skill, because we do this for every player and adjust when something does not match the systems expectations. How It works is of large numbers, a lot of data collected with the goal to be self correcting and increasingly accurate over time. The benefit of this means that we can create a more fair game then randomly selecting players from the entire pool. By looking for skill rather then # of games played we lower the number of mismatched games. The cons? this takes more time then a lot of folks realize, and the systems starting point for any 1 account is highly likely to be inaccurate. This means the system will see a match as fair and can be completely wrong, but it takes each instance of being wrong to adjust to become more accurate. In Elo and MMR you will have grossly mismatched games because the end goal is to place accounts where they belong, it is not the starting point.

Wasn't Arpad Elo's system made for chess, can it even work for other games? Yes, because the system does not look at the game itself but wins, losses, and in chess draws. Its not rating you off your plays or how well you moved queen or knights. Its rating you off if you beat a player above you or lost to a player below you from the rating it gives you and adjust over time. The system its self has nothing to do with chess. Instead of watching the game start to finish then trying to rate performances what you ask is how good where the players? who did we expect to win? who won? How certain was the system that both players are accurately represented by the number the system has assigned to them? It then will Adjust accordingly.

What's the difference between Team Elo and MMR? Nothing really, Devs use the terms interchangeably for the systems. Team elo was confusing because a lot of folks looked directly at the 1v1 chess and tried to apply It to team games as is. That does not work, it needs to be adapted for team play to account for individuals on a team. So MMR became a popular term. End of the day the systems need to function off Skill matters and self correction, the names Team ELO, MMR, Or even Microsoft Trueskill set expectations on how the system functions. 1. It starts with high uncertainty on an account. 2. accounts have the same start point and move faster with high uncertainty then slow down over experience and at various points on the ladder, if your system sees data that makes it have more uncertainty again they move faster. 3. The average player will be near the starting point, by virtue of the worse players falling down and the best players climbing up. 4. We do not start players at the bottom of the ladder, or to be more clear we do not start accounts at different percentiles randomly. (Mind you its entirely possible to have "bronze" be your start point in a game. Just means that 50% of your games players are going to be in bronze on the ladder. the percentiles don't change just the arbitrary Medals devs get to pick does. A "bronze" player in game A could be at the same Percentile of ladder as a "gold" Player in game B, Trophy's and Tiers are a distraction because its all decided by devs. Some devs want folks to feel challenged so their going to put their start point in a "lower" Ladder. Some devs will want players to feel like they are pretty good and skilled so theirs is "Higher" Ladder. )

So what's my MMR score mean? It depends on the game your playing for its exact meaning because the devs can pick their numbers. On the whole the first thing you need to know its how match maker builds games. In most Mobas it takes a range for acceptable MMR to build a match, fills the game with players in that range, factors information for each individual player into its variables then adjusts individually for the players on the result of a match, despite sharing the same name not all Elo/MMR systems do this the same. The second thing you need to know is that on the ranked ladder different MMRs represent an expectation of a level of skill and will account for a certain % of the player base. On the whole 50% of the games player base on average is above and 50% is below the starting value all accounts are assigned, Some variance applies, also Most of the games, players will end up being in fairly close to that starting value. It largely represent the average level of skill of its players at any given time. Humans cheating the system, as well as outliers do exist but are highly improbable to represent a majority if your player base is high enough.

So my MMR is my true skill exactly how good I am? No, its a range as ya don't expect players to be consistent. You can think of it as a mean average for your skill, rather then the median or mode. If you check the Microsoft TrueSkill link they have a nice graph of this.

Kai, i'd like to know the math so I can check this, pretty please? Fair enough, but I cant provide that because basically every dev does not share this information. The systems are built around variables that change on a match to match biases. The system creates an expectation with the information it has, uses that to set the variables, then adjusts after the end of a match. So knowing any Factor with out knowing the entire system means nothing because the values are basically arbitrary, so long as they don't screw the system completely. Your going to need to look a lot of things up on your own time if you don't think you can trust me. Hit me up and I can share some sources, mostly checked FAQs for mobas. Checked ranked ladder and look at Percentiles for them. Checked a lot of pages on how Elo works. Theres the True skill link and a video down the road for some one testing a team Elo. Everything is publicly available and I invite you to take your own time if your skeptical of me as source.

This sounds kind of unfair, if I am an average player my games are going to be largely decided by players on my team? No, its not unfair but Yes, That exactly what this type of system strives to do, get to a point that your skill is so accurately represented by the assigned number you carry its expected amount of weight on your team. It wants your skill and assigned skill to be close. Its not unfair how ever because the systems not trying to give you wins or losses, it doesn't give a dam about you honestly. It just wants to divide players up by their consistent level of skill. Eventually your going to hit a wall. Weaker players will go down until the majority of players there are at their level. Average players will usually hover near starting points, good players are the only ones who climb, and they only climb how far their skill can bring them. There Is a luck factor but its set so that mismatched players cant rely on just luck, you get enough matches in and it balances out the unwinnable and free win games.

"HA I GOT YOU KAI YOU JUST SAID IT. TEAMMATES DECIDE GAMES! RIGGED 50% WTF RITO" … uh...no. See rigged would be the system has to care about you and try to affect your matches outcomes in advanced. MMR and ELO gives 0 dams about any 1 player. A loss from a trash teammates just 1 loss. If any 1 player is so bad they lose the game for their team at a MMR range that dudes going to lose 100%, They will drop out of that range faster then you will. If 1 dudes so awesome he can carry anybody, he is going to go up until the MMR range is harder and he cant carry like that. So Its only if your near your average skill in A mmr, that's when your expected to be floating near that magical 50% win rate folks like to talk about. Lets say your on a new account or its a new season. This adds uncertainty, that called the K factor and I will chat about that in a moment. If your K factor was really high and you dropped HELLA far down the ladder... Then the system actually favors you because your no longer the average level of skill player there. You actually much better then your being accounted for by the system. The Mean Skill of how you play is above the Mean skill the system assumes for you. In Elo and MMR The system wont try to load teammates to make you win or fail, its self correcting so it expects if your skill is good enough being mismatched with weaker players just gives you more opportunity's to shine. Lets look at people who cheat the system. Smurfs show a pretty clear example of higher skill wrecking low skill. Purchased accounts tend to show a example of Low skill being a handicap in 100% of their games.

"Well... That's... Logical but I'm pretty Freaking positive I am better then the scrubs I am playing with and vs. why cant I climb? Few reasons. 1. Inadequate # of games played. 2. In league MMR is hidden to keep folks from getting discouraged 3. You might be slightly wrong about your own skill. Its not that your not a bit better then the average level of skill in your games but that it has to be high enough to climb, just being better on some mechanics alone is not enough. KDA is not enough. You need to convert your skill into consistent wins vs the people in your range. If your only slightly above average you can expect to need to play 100+ matches with a bad start, K factor (I promise ill explain that in more detail and why it matters.) Has set you back. High K factor makes folks rubber band over MMR ranges. Zooom, up they go. Few lucky games your now out of your skill range. You handicap your team slightly ever match. So Zoom down you go, Few unlucky games they are now consistently the strongest player in a game. (your always favored in games your strongest player btw. 20% odds the worse player is on the other team of the best player. This is literally true for every match, that's probability's for ya.) Zoom Back up you go. The K factor lowers with # games played, so over enough matches this rubber banding becomes less dramatic.

I don't know man... sounds like a lot of impact from luck. Oh lucks a factor but all it does is affect how much of a grind ranked ladder will be, rather then prevent you from ever being able to get accurately close to your Skill level. There's a pretty easy and objective way to check for skill impacting a games outcome. Compare yourself, to yourself. Ask, when I started champion X how good was I? When I started playing this game vs how I play it now? If you had a team of You currently vs you in the past when you started, who would win? If you Mirror lane your Main champion now vs Old you who would win? I cant answer that for you. I can say that for me season 1 Kai was a Bronze Player. I am using the icon to prove it. Back then I was awful and toxic AF as well. AGI dota allstars Carry mentality of farm farm farm KILLS KILLS KILLS! I wanted to be a 1 man show once I hit items... Looking back i can say objectively i had Awful teamplay and expected everybody to work around me. Works well if your a Phantom lancer or Faceless Void main back in allstars. Awful if your AP trist or Ashe. I would often lose games with 12+ kills because all I did was go for kills all day long. Take tower? Nahhhh roam for more kills! Help OBJ? naahhh ima cheese some one with a 1 shot combo in this bush! So luck is a factor in how much time you need to commit, But skill is what determines the max range you can fall or climb on the ladder. One of the reasons know about Elo/MMR systems is I tried then failed to find proof Elo Hell was real and not my fault but teammates.

Finally, K factor? Yep. As I said early K factor is the variable in adjusting MMR after games used to represent uncertainty. the Less data the game has the higher the K factor. K factor = uncertainty, can also think of it as K factor = how much the Max MMR you can gain or lose after a match will be. A basic 1v1 adjustment would look like New MMR = Old MMR +Kfactor(match results - Expected results). The stuff in () is variables, I am using terms that should help folks understand what those variables represent rather then leavening them as vague variables. K is higher for new accounts. Always, this is CORE to MMR and ELO systems, the exact value gets debated and needs to be game by game. Other modifications can get applied, one of the most common ones is that it slows down in higher and lower Tiers. That way a new account does not get excessively high or low in the ladder when its still new with a higher K factor. You don't want good players falling to the bottom and bad ones mucking it up at the top. Makes mismatched games, hurts the system over all. Lets give an example. All numbers are arbitrary decided by me and have no actual meaning besides to be an example. 1200 is your current MMR. You play a game and win. Formula looks like this. 1200 + 32(4-1) = new MMR of 1296. A different player has a lot of games played so their K is lower. They are also at 1200 and play the exact same match. 1200 + 10(4-1) = New MMR score of 1230. Same Expectations same starting MMR diffident gains from the win. Over time this lowers rubber banding you see with high K factors and narrows in on an accounts MMR. Theres a lot of ways to build your Adjustment formula and in a team game my example does not work for team games and needs a lot of modifications to account for individual impact. Also in my example K is an invariant and some people use it as a variable built off both players K factor. I assume you can also make it a Variable off gap in skill size thou I don't have a good understand of what that would do to a system. As the Matchmaker should be pulling accounts only from a set range in a team game. Like I said lots of room to adjust arbitrary numbers in a variable system.

Humm… well... Is there any way we can test the systems or know for shure ELO and MMR works for team games. Im still thinking luck is going to be a big decider. Oh lucks going to Impact how good of a time you will have. Riots last official # I saw was they expected it to take about 150 to 300 games to hit your actual skill. Rather then another wall of text. Please look at this guys video as he built and tested his own team MMR System. I don't think i'd use exactly the same system as his but its pretty useful to see how you can make and test things. Its pretty easy to understand and digest.

"Hold on, this is just MMR and Elo systems. League of legends uses riots League System. So uh...That's different right?" Yes and no. Like, yea, It is different in several ways its BS to claim its not. It has some further modification on it then just MMR. Objectively, Its more inaccurate then a purist MMR/ELO system. Yet It has some perks, players feel better using it because they don't know what percentile they are in or how the system exactly judges their skill. For the Devs It also gives them the power to apply a penalty to Lobby dodgers with out it affecting the accounts MMR. But I have to say no for it being significantly different we should treat it as a completely independent system because Ultimately its just a Derivative of a team Elo/MMR in how it builds matches and adjust the players values. So we can set expectations for how it functions by knowing about MMR and ELO. In Riots League system I think its fair to say that Lucks gonna result in folks having a higher or lower tier then they deserve more frequently but its going to be in a workable range assuming they play enough games and Riot did not completely FK their system up. As you can test your system its not unreasonable for them to know if they completely killed it or not and we can always just look for signs MMR is working as intended. Weak goes down Strong goes up, can you check the players in the bottom of the ladder and top and see a gap in skill? That's a basic requirement. Again with Tiers, those are ultimately meaningless so seeing 1200 = bronze means nothing. Its arbitrary devs pick it. Some one else could just make their game and say 1200 = Silver. Hell with new tiers like Iron riot can completely adjust what the average player tier is going to be, just make them reflect different MMRs then the Old one did and presto majority of the player base is something new. Can be higher, can be lower, entirely up to what devs set.

So what take away do you want me to have from this? They work at building ranked ladders off skill where # of games played has no meaning with out the ability to back it up. They are fair because they don't care about individuals. Luck affects the speed to get to a workable level of skill to assigned mmr rather then being the decided on final MMR alone. Good players are favored, weaker players are not. Forced win Forced loss systems are more complicated to build, and any end result they achieve can be accomplished by just tweaking your variables.

I have more questions! Go for it. I will do my best to answer.

Kai you derp you screwed up on X Entirely possible. I am a bartender by trade. Games are just a hobby and I like breaking systems, been a geek most my life and love Vid + board games. In MTG terms im a Jhonny style player. In Leauge i'd be a Off meta main. This is just the results of me learning and checking and double checking how systems work then trying to make that information available in less cluttered ways to other players. Feel free to call me out but try to avoid purely anecdotal evidence (feel free to use it when accurate to support a logical point or for illustration) I also would appreciate it If folks avoid discussing things they have not studied the theory for and are just making a stab in the dark.

Kai I swear to god its rigged 50% and I will fight you! By all means go ahead. Say why you think its rigged & what riot is gaining for doing this. Here is exactly what I will do in my response. 1. Ask what gains they get if you don't list them. 2. Show how you can get the same end result very easy by just playing with a MMR system. 3. Point out the very high level of complexity and massive number of RNG involved in forcefully throw matches. 4. Show that Forced systems are going to be weaker then MMR at consistently getting the results the devs want even if we assume they are doing it for greedy/selfish reasons. 5. Link you to other Message boards and forums for mainstream mobas/games where their player base is also bitching about forced 50% matchmakers and ask you, is this the industry standard then? If riot is the only company doing it why do so many people in Dota 2 have the same complaint? Why are you right about Riot and they are wrong about Value. 6. I am going to ask you about the clear gap in skill of bronze 5 players to Challenger and how that happens if its forcing Random accounts to win or lose. 7. Where is the logic for your view? Because if it was 100% forced 50% then no one would climb. So it has to be "RNG" Forced 50%... then how do the same players constantly climb on multiple accounts year after year? 8. If you throw Anecdotal evidence in my face I will throw it right back at you + point back to the LCS smurfs riot unleashed on us a few years back.

Look, you can save us both a lot of time by learning about MMR and ELO then think up ways to get any end result you want just by manipulating all the arbitrary numbers you get to play with when you control the entire system.

Well long ass post but I do hope its helpful to some one.