General DiscussionMMR A bad representative of skill?

MMR is based on wins. They should do TSR. I remember back in hon, there was like a 90% accuracy of say knowing which team would win based on TSR as opposed to MMR. Just think if we had public access to each account's Hero rating and quantify it as a rating of say 0-10 for the whole account etc..

Here's one sample I remember:

Your team:
You: 4.71
Ally 1: 3.68
Ally 2: 2.81
Ally 3: 2.24
Ally 4: 5.23

vs.

Enemy 1: 5.28
Enemy 2: 3.75
Enemy 3: 3.62
enemy 4: 4.82
Enemy 5: 3.92

Everybody is in 100 MMR of each other... Cuz u got 2 low skilled players (2.x rating) u will most def. lose if u didn't out draft them.

^ Source on that?

1,000+ games with TSR mod.

win is all that matters

Just imagine if you're sitting at 5k mmr and a 2k mmr kid is telling you how mmr is a bad representation of skill. It is not. Do you think a 6k mmr player gets 6k just by pure luck, or do you think he's figured something out by now? I guarantee you a 6k mmr player will slaughter a 3k mmr player. It's a far-fetched example, but it shows that at varying levels, MMR really is a good measurement of skill.

To address your point further, KDA, xpm, and gpa mean nothing. Only the win matters. It doesn't matter that you go 25-3 and you lose, because your skills weren't enough to bring your team to victory. And in that case, your team probably didn't deserve to win / deserved to lose MMR, including you (because there are people who *could* have carried your team to victory and truly deserve those mmr points, but you failed to bring your team to victory).

win does matter. It's like casino, how u know this 6k guy isn't lucky and hits 4 man stacks every time etc.... Take for example Fnatic Blunt etc...

He coulda been a 5 man stack that his team carried him to 6k per ranked and sitting at 6k mmr due to 50% lose rate.

Taking it seriously now dota is 5v5 man game. In a tough game just 1 weaker player can mean win or lose. I don't think that's an accurate reprensentation of skill.

Volvo has millions of replays, they can quantify how much XPM/GPM an average furion does at X MMR range etc... how good he does at high and lower ratings etc... Sort of like our hero charts thingabob.

So use that as a gauge to show hey... you are at 6.5 dota scale vs. 10 (dendi level) etc...

Just to counter Jjang's argument. This isn't call of duty, you can't simply win the whole entire game for your team when your team can't do anything. Dota is a team game and as FyrionPlax put it, just 1 weaker player can mean win or lose.

The only thing that affects your MMR is wether you win or lose.
I went 19/4/14, lost the game and still got -25 MMR, same - MRR as if I feed like hell.

^
the only thing matter is you get the enemy throne, not killing enemies.

MMR is subjective, but it does provide a good starting point to gauge skill. KDA-based computations are irrelevant because as Flop put it, the objective is the game is to destroy the opposing team's base. Teams can stomp in terms of kills and still lose to split pushes. Valve has elected to go for the bare basics, which is wins/losses. It's not ideal, but it's sufficient for the time being. As time goes by and more data is gathered, it'll become easier to fine tune MMR to be more accurate.

To those making comparisons with HoN, please don't. The ratings system employed there was utter rubbish. I was an early beta tester, and we had a leg up on everyone else because we formed teams earlier.

Exactly, they need to come up with some skill rating that actually takes into account multiple sources, not just win/loss, and not just KDA, but rather a mix of all the different stats that can come out of a game. If you were a support that bought wards and did everything you could have done to win the game, but you still didn't, your rating should still go up. Also, if you played very badly yet your team managed to win, you shouldn't be rewarded for that. I'm just saying that they need to take into account more than just win/loss. Because any stat that represents skill should not be based on just one basic overall thing such as win rate, even if it is a team game.

Defining such a rating is considered an ill-posed problem, one which is not likely to be solved anytime soon. My day job involves working on the mathematical modelling of complex systems, so take my word for it that you best get comfortable with the system we have now, because a better one is way off in the horizon.

MMR is a rough approximate for your overall skill, from the hero you pick to the tenacity you display to the success of your plays. Sometimes you may play well but a destructive attitude lowers the team morale, or the hero you pick does well but is unable to carry the game. And of course there will be times when you lose the game despite having done everything correctly.

But this is inevitable. It is a result of the rating system being only an approximate gauge of your skill and being unable to predict exactly how well you will play the next game. It is an inherent part of any rating system, from the Elo system in chess to the Elo system back when I played Pokemon Online competitively. A thousand variables affect skill and your rating is a simplified version so that matchmaking can at least try to balance games to the best of its ability. Even in Chess when you have different playstyles/favoured openings matched up against each other a lower Elo player may consistently beat a higher Elo player. But the Elo system doesn't aim to account for how likely a player is to castle his king to the left or move his queen to D6, for such a system would be unnecessarily complicated and convoluted for relatively little gain.

It's just a number. It's fun to look at; to manipulate; it doesn't necessarily show your true skill level, but it works as a decent approximate.

mmr is fine!
you good u win, you bad you lose -->+-mmrr

Benao you are just a trashcan that needs to be melted when going on MID, long enough to hurt your brain functions and prevent you from posting mindless comments on these forums. Stop spamming good threads about mmr, fucktard, if you are not in the mood for discussion or have anything substantial to say.

And yes i totally agree with PyrionFlax about TSR-type ranking system. Only guys like him (with his mindset) can make this game improve further and reach its true potential.

MMR is luck based in the sense that your teammates could be peruvian or your enemies could be peruvian, and the team with the least peruvians wins. Yeah you outskill the enemy team, but it's because they're peruvian.

i'd rather have peruvians on my team (10% of whom are retarded) when i queue on us servers than shithead americans who try and tell the team what to do obnoxiously over the mic and then flame you if you don't listen to their bad advice (99% of whom are retarded)

This is easy to answer.

Is W/R the best way to gauge skill? No, but its the most consistent that we have

Is W/R Luck based? Yes but slightly. Not everyone will be compatible, competent, and serious.

Will there be another mmr rating system in the future? Probably not and def not in the near future. The money isn't behind it

How do we get valve to change it? Make a 3rd party client that uses the TSR system and make it so everyone goes to that client and doesn't queue up on valves matchmaking. However TSR TAKES FOR EVER to get a accurate representation.

Plus, making mmr only about W/R makes the game quite stressful and part of the reason the community is garbage today. People taking winning too seriously. Id rather not worry about the win and try my little fucking ass off. If I lose, I know I'm in good hands cause of my Tower damage, last hits, gpm, kills, assists : team kill ratio etc.

^^^ LOL!!!

Yeah you know I was surprised to see how many people from peru play and I get during the night time. I was like "uhhh i'm a dumb american so lets see some stats on these guys:"

Population: 30 Million
8% unemployment
Average pay per year: $6700
Literacy rate: 98%
Primary school enrollment: 77% (elementary through highschool)

IDK why I looked all those stats up besides my ignorance of what the country was like. Looks like they are doing good except the average pay per year is disgusting which leads me to another question. WTF kind of computers/internet are these people using if their base salary per year is that dismal? Am I playing with the "middle/upper" class there? Or are they using hand me down old ass PC's from 10 years ago to play this game? Beats me.

@ op

considering nobody in that game has > 5 cs/minute, i'm guessing the mmr range is 2.5-3k, so this is the wrong forum to whine in. playdota will have people who are more likely to sympathize with you (they're still wrong, but you'll feel better amongst your kind)

if you want actual advice: learn how to farm better

@ whoever talked about tsr

if you hate the mmr system as it is now cuz you think it takes too long to improve your score, you would jump off a building in disgust at the improvement rate if the tsr system were implemented.

what a thread

TIL beesa (6k rated) just "got lucky" when he solo queued. it isn't because he's actually better than the vast majority of players.

damn wish i was just as lucky :(

Zenoth what rating were you on Pokemon Online? I capped out at 1800 LOL. I hated the critical moves (Stone Edge/Focus Blast) that were necessary but had shit accuracy; they always feel like 90% accuracy when opponents use them and 30% accuracy when I do.