What is the point of MMR?

[same]

The goal of this post is to try and clear up some confusion.

MMR in a competitive game is not meant to be a reward or punishment. It's also not meant to be a carrot on a stick used to keep players engaged. It's meant to attempt to provide players with balanced matches.

What does it aim to measure? Two things:
1. A players engagement with the matchmaking system. This one is pretty obvious, you need to play games in order for your MMR to change.
2. A players relative skill level within the system, or maybe more accurately, their likelihood of winning against another player with an established MMR.

I'm not saying the system as it is implemented currently is perfect, the swings in MMR seem quite punishing for a player rated 200-300+ points higher than their opponent and likely don't reflect their actual chance of winning the match, although I don't have the statistics to really make that claim in earnest.

If you are dissuaded from playing the game because of a number that's only meant to provide a basis for balanced matches, I don't know what to tell you. But I might ask, 'Why were you playing the game in the first place? Is it because you enjoyed it or because you wanted some participation award for playing?'

[Adam Earlston]

If any1 is confused about MMR and dislikes it, they need to stop taking the numbers too seriously and continue playing like how it was before MMR existed.

Just start thinking about a player with a higher MMR as a bonus opportunity to get more MMR for urself. It doesn't mean they are going to be much better than u and win, it's just a number and if they matched against u, then their number is probably irrelevant anyway.

[Fishy]

Originally posted by Adam "Yuudai" Earlston:

Just start thinking about a player with a higher MMR as a bonus opportunity to get more MMR for urself. It doesn't mean they are going to be much better than u and win, it's just a number and if they matched against u, then their number is probably irrelevant anyway.

People are actually facing the opposite problem, and want to get matchup upwards, because the MMR weighting is so broken.

MMR in a competitive game is not meant to be a reward or punishment. It's also not meant to be a carrot on a stick used to keep players engaged. It's meant to attempt to provide players with balanced matches.

Outside of matchmaking, It's also a measure of your relative skill level, and does provide an incentive and reward for people who want to improve at the game and measure that improvement (and can handle a number that won't always go up.)

Right now, it's badly broken, so people who are used to other games' fake MMRs and want "participation" points aren't getting it, but also people who want the MMR to be a measure of relative skill *also* aren't getting it.

[Vegas Dad]

Originally posted by same:

The goal of this post is to try and clear up some confusion.

MMR in a competitive game is not meant to be a reward or punishment. It's also not meant to be a carrot on a stick used to keep players engaged. It's meant to attempt to provide players with balanced matches.

What does it aim to measure? Two things:
1. A players engagement with the matchmaking system. This one is pretty obvious, you need to play games in order for your MMR to change.
2. A players relative skill level within the system, or maybe more accurately, their likelihood of winning against another player with an established MMR.

I'm not saying the system as it is implemented currently is perfect, the swings in MMR seem quite punishing for a player rated 200-300+ points higher than their opponent and likely don't reflect their actual chance of winning the match, although I don't have the statistics to really make that claim in earnest.

If you are dissuaded from playing the game because of a number that's only meant to provide a basis for balanced matches, I don't know what to tell you. But I might ask, 'Why were you playing the game in the first place? Is it because you enjoyed it or because you wanted some participation award for playing?'

You are half right and have wrong. You totally missed the point of the MMR rating is to reward people allowing them to climb a competitive ladder it's not just a number. A big problem is people are trying to force MMR elo systems in random video games because it works in chess which has an extraordinary confined playspace which results in a dramatic skill expression, chess if you are 200 points better you will literally beat someone 95 out of 100 games because there is zero luck involved, and every possible variable is set and accounted for. In many of these computer games all of that reliability is missing which results in a very high variance and low reliaibility "skill" expression". In this game your starting spec and units can literally decide the match, a random placement of a unit placed blindly can wildly alter the outcome and you have zero ability to use your superior skill to account for it, you have zero ability to account for the random cards, the random placement of spells, and some of the random interactions with units (literally a pixel to the left will alter the outcome of a round.) So as a 2200 player playing a 2000 player you can easily get a 60-40 winrate. To say otherwise is delusional.

Not to name names but sacco a prominent streamer loved to argue there was some huge gap between 2200 and 2000 when he was 2200 :P not he consistently sits around 2k, and over his journey down lost to many people under 1900 and one as low as 1400 if it was truly a big skill gap you would never lose to some one that much lower. This isnt basketball or a hard hand eye skill game like first person shooters or a game that needs 120 APM. It's nit even highyl strategic or tactical in its current iteration its just a base knowledge and go game.

The point being finding a match of players of "equal" skill is a bit of an illusion because as i listed above there is not alot of skill expression in the game so there is very little room for gaps to truly have skilled players excel over unskilled players unless the game is huge like a new player or a very casual player vs a player who has a spreed sheet with every match up. To your point if the number is meaningless then what is the point of me losing 300 points to a player i am matched up against and only gaining 6 points since it was decided the player i was matched up against is of appropriate skill? If they were not of appropriate skill then i should never been offered the match it can't be both so sorry the numbers do matter.

[same]

I never said the number is meaningless. I said it was meant to be the basis for attempting to balance matchmaking.

As to your point about win rates at rating differences for Mechabellum vs Chess, I agree that it's different. But the weighting of MMR could be changed in order to reflect the actual chance of a player winning their match. It seems like the way it is currently implemented, it's as if the win rates are supposed to be Chess-like, but it was fairly recently introduced and I think it's pretty reasonable to give the devs some time to look at the data and then they can make some adjustments to the weighting, if it looks like that should be done. You don't need to throw out the entire system because it has some flaws.

[Vegas Dad]

Originally posted by same:

I never said the number is meaningless. I said it was meant to be the basis for attempting to balance matchmaking.

As to your point about win rates at rating differences for Mechabellum vs Chess, I agree that it's different. But the weighting of MMR could be changed in order to reflect the actual chance of a player winning their match. It seems like the way it is currently implemented, it's as if the win rates are supposed to be Chess-like, but it was fairly recently introduced and I think it's pretty reasonable to give the devs some time to look at the data and then they can make some adjustments to the weighting, if it looks like that should be done. You don't need to throw out the entire system because it has some flaws.

Fair enough but my point still stands the number is for more than just match making. Many people play games for their competition fix. So i was just adding some clarification to your clarification post :P

[Fishy]

A big problem is people are trying to force MMR elo systems in random video games because it works in chess which has an extraordinary confined playspace which results in a dramatic skill expression, chess if you are 200 points better you will literally beat someone 95 out of 100 games because there is zero luck involved, and every possible variable is set and accounted for. In many of these computer games all of that reliability is missing which results in a very high variance and low reliaibility "skill" expression".

There are so many things wrong with this paragraph it's hard to know where to start!

Chess doesn't have an extraordinarily confined playspace. There are more possible game states in chess than atoms in the universe.

200 points of ELO difference in chess is ("literally") about 3/1 odds, not 20/1.

There is lots of luck in chess. This one may seem unintuitive if you're used to modelling luck purely as overt randomness, but have a watch of Garfield's talk on "Luck vs Skill in games" to see where this statement is wrong.

Having variability (or instances of overt randomness) in games doesn't necessarily equate to less skill expression. In fact they often create the circumstances for greater expression of skill and more games going to the more skilled player.

And the main point - Elo is absolutely fine for games that have some amount of randomness in the result of a given match, as long as it is weighted correctly. It will still produce robust and fair relative rankings of players.

[pjnt]

if one side has twice the MMR, it wins almost always. It is not difficult. When the MMR is equal the fights tend to be equal, too. Match making should reflect this but it doesnt, so the game will die, unless the devs fix it.

What I dont understand is why in 2v2 there are 2 600MMR on one side and 2 300MMr on the other?`Why not make it even?

[puschit]

VegasDad, this game isn't a "random video game" that MMR is forced into without reason, it's a highly competitive PvP game that absolutely needs and MMR to pair players of equal skill against each other, otherwise it would be frustrating for everybody involved. Besides, almost every single player WANTED an MMR based system. Beforehand, during the Beta days, we only had Combat Power and it sucked hard because it only reflected amount of time spent on the game but not skill.

The problem with the current system is apparent and twofold: Too harsh punishments for losing against someone with lower MMR and the ability to cancel a matchup that you don't like. The latter makes zero sense to begin with because it undermines the very concept of MMR based matchmaking but at the same time it exacerbates the former problem because players now tend to cancel all matchups against lower ranks.

Your points about randomness are drivel, the amount of luck in this game is very small and skill will always compensate for whatever small disadvantage you might have from starting units. Games with MUCH higher RNG have successfully implemented MMR based matchmaking systems.

This game NEEDs MMR based matchmaking but it should be a working one. The opposite of what you said is true: Mechabellum would be better off introducing an MMR system like chess does instead of trying to build their own version with their own calculations and quirks because, as it seems, they have no clue how to make that work, the ability to cancel matchups is all you need to know regarding that.

[Fishy]

Originally posted by puschit:

This game NEEDs MMR based matchmaking but it should be a working one. The opposite of what you said is true: Mechabellum would be better off introducing an MMR system like chess does instead of trying to build their own version with their own calculations and quirks because, as it seems, they have no clue how to make that work, the ability to cancel matchups is all you need to know regarding that.

They basically have just used the same one from chess, but that is part of the problem. It does need tweaking. The chess Elo algorithm assumes a 200 MMR difference gives 3-1 odds to the higher MMR.

They also seem to have doubled the base point gain/loss (from 10 to 20), which just makes things more swingy. Not the worst, if things are balanced, but it does exacerbate the other issues. I think they are also doing some win/lose-streak matching too, which is also exacerbating things.

Turns out that if you try and apply 3 sets of balancing algorithms simultaneously (and one of them is broken), you get weird effects!

[Vegas Dad]

Originally posted by Fishy:

A big problem is people are trying to force MMR elo systems in random video games because it works in chess which has an extraordinary confined playspace which results in a dramatic skill expression, chess if you are 200 points better you will literally beat someone 95 out of 100 games because there is zero luck involved, and every possible variable is set and accounted for. In many of these computer games all of that reliability is missing which results in a very high variance and low reliaibility "skill" expression".

There are so many things wrong with this paragraph it's hard to know where to start!

Chess doesn't have an extraordinarily confined playspace. There are more possible game states in chess than atoms in the universe.

200 points of ELO difference in chess is ("literally") about 3/1 odds, not 20/1.

There is lots of luck in chess. This one may seem unintuitive if you're used to modelling luck purely as overt randomness, but have a watch of Garfield's talk on "Luck vs Skill in games" to see where this statement is wrong.

Having variability (or instances of overt randomness) in games doesn't necessarily equate to less skill expression. In fact they often create the circumstances for greater expression of skill and more games going to the more skilled player.

And the main point - Elo is absolutely fine for games that have some amount of randomness in the result of a given match, as long as it is weighted correctly. It will still produce robust and fair relative rankings of players.

actually there was nothing wrong with the post beyond the "random" example of gave of what chess elo difference might make to a win rate. Other than that you are so interested in arguing it overrides your ability to read.

without getting into a silly philosophical what actually is luck or random debate you are wrong. Chess has very little "luck" or variables to the state of the game beyond "what my opponent chose to move last turn" the play space is extraordinarily confined. You know all the peices, you know EXACTLY what they can do, EXACTLY where they are, What possible spaces they could move to everything possibility is easy to determine with 100% certainty. You will know with 100% certainty exactly what occured on your opponents move before you take your move. There is zero "luck" there is zero "random events" that can impact the board state before you make your move.

Mechabellum on the other hand is filled with luck, randomness and uncertainty when you make your move in part by the fact your turns occur simultaneously and you are blinded to what your opponent employs that turn, on top of that there are random cards that the literal definition of luck and RNG (some might help a player more than the other, there are 5 choices each of which might require a different counter play but you will have ZERO idea which to employ because you are blind to what your opponent chose). You have no idea if your opponent bought shields, any tower abilties etc when you execute your move , you have no idea what units they bought nor where it is positioned. All of that means you cannot possibily make a highly "skilled" decision on what to do your turn and on top of that many times an arguably "skilled" decision can have very poor outcomes. This means your skill has a lower impact on the outcome than other instances of skill measurements.

you confuse variability and random events you have no control over nor even information to make a proper descision on. Chess has alot of variability in the board state because it changes with each move, but the board state is 100% certain there is not a deck of cards where we flip a random card that might have a random instruction that must be executed to alter the board state, if that was in place chess would instantly have less of a skill expression as your opponent might turn over a card that says "instantly kill your opponents queen" and suddenly you are able to check mate your opponent that turn as there queen was removed from play. there would be no possible reliable way your opponement could account for all the possibile actions in that deck of cards or have an option to employ your skill to counter play it and at times they will lose to an inferior player because of the luck of the cards not the skill of play.

It is part of the magic of these types of games if you remove it the game would become far less interesting to most people much like chess is.

[MutantBison, PhD]

Originally posted by Vegas Dad:

Mechabellum on the other hand is filled with luck, randomness and uncertainty when you make your move in part by the fact your turns occur simultaneously and you are blinded to what your opponent employs that turn, on top of that there are random cards that the literal definition of luck and RNG (some might help a player more than the other, there are 5 choices each of which might require a different counter play but you will have ZERO idea which to employ because you are blind to what your opponent chose). You have no idea if your opponent bought shields, any tower abilties etc when you execute your move , you have no idea what units they bought nor where it is positioned. All of that means you cannot possibily make a highly "skilled" decision on what to do your turn…

The idea that a player can only make skill based decisions when all of their opponent’s moves are telegraphed to them is laughable.

I guess every strategy game in existence with highly competitive ranking systems, world class, professional tournament-level teams, and fog of war in game must therefore lack any ability for those players to be skilled if their expression of tactics and strategy depends completely on being able to see what their opponent is doing. No one tell StarCraft E-Sports.

Or all those FPS titles like Siege that function under the same conditions of an unknown environment and enemy actions/characters, with worldwide professional e-sports, are also inherently luck based…

Not to mention the fact that strategy games are just a further, creative expression of the principles of war as demonstrated by the greatest military minds throughout history: Rommel, Sun Tzu, Napoleon, Alexander, etc. And the real world application of knowledge and skill is what allowed them to win and develop the art of strategy specifically because they could win irregardless if they knew exactly what their opponent was doing. But I guess they just got lucky.

Mechabellum must therefore be purely luck based without adequate expression of strategy or skill, since it falls under the same category as all of those previously mentioned. So I guess that means we better remove MMR entirely as it serves no purpose to define player skill level in a game which lacks it completely.

Thanks RNGesus, you’ve ruined yet another game!

[Fishy]

actually there was nothing wrong with the post beyond the "random" example of gave of what chess elo difference might make to a win rate. Other than that you are so interested in arguing it overrides your ability to read.

Nah, you're wrong mate. And like you always do, you just resort to insulting people. It's so tiresome.

If there was 0 luck in chess, then the win-rate between someone with an Elo difference of 1 would be 100% / 0%

I don't need 3 paragraphs telling me where the randomness in Mechabellum resides, from someone who doesn't understand that luck can come from sources other than just 'RNG', and that it can increase the expression of skill in a game.

Why don't you go and win the Poker World Tour or something. You'll be in with a decent shot, since it's all just luck right? There's no possible way someone could account for what card comes off the top of the deck. And that random card draw happens thousands and thousands of times in a tournament! That's the most luck-based game imaginable, right? You literally can't win if your opponent always draws out on you. No possible way to play around your opponent drawing the nuts every turn. And if there's that much luck, that means there must be 0 skill, right?

Those people who make a living off of it are some lucky, lucky folk!

[Shaia_Tempari]

I think there's a trench right now, where if you get into a certain point it's really hard to climb out of the MMR hole, but if you do, you get easier opponents.

[Purple Sauce]

Honestly adding MMR to the game was a god awful decision. They should remove it until they properly fix it.