General Discussionmmr is and should be determined by win/loss

^agreed.
Mmr is already fine as it is. As a businessman, leader,president etc. you cant decide everything that make everyone happy, so try to accept it.

Majority of dota 2 players already accept it, so to the minority who still complaining try to "move on"

An MMR based on wins/losses is just the beginning. To base it solely off this attribute would be highly inaccurate. The existing system is sufficient, but not excellent. What most players don't realize is that skill isn't something which can be measured numerically (at least not yet). No matter how complex a matching algorithm is, it still bases its decisions on numerical values, values which cannot be accurately assigned because we don't understand every contributing factor.

well i just feel that dota should be like the rating in chess - when you win your rating goes up and when you lose your rating drops. in the long run the rating will incorporate all aspects of your skills and each aspect will be weighted against how much it contributes to winning

thats my thinking too

agree

@ 2 above chess is entirely different as it measures individual skill to begin with (1 on 1) whereas dota is 5v5 which i believe is what the op is getting at.

My 57% win rate does not agree with my 1900 MMR. Started with 1600 after calibration. =\
Been slowly climbing up, but it's tough, especially starting at...well, as I said, 1600.

EDIT: Oh, and it's worth noting that a majority of people I play with in matches has <50% win rate. Unsure of why this is. Though I do agree that the current + and - system of MMR is good as-is.

the issue is that you get +-25 for almost all games unless you are very low or very very high on the mmr scale.

everyone else gets the same points, regardless of who the lose to or beat.

Heroes of Newerth actually took into account the mmr of your opponents.

^are you sure? i had one game that gave me +19, which surprised me cause it was pretty close. I think it does depend on your opponents, except you are matched with people with the same average mmr

win/loss should not determine mmr. Because it's a 5v5 game. Games like chess, checkers, pool, bowling etc.. where its' 1v1 etc.. where elo rating will work good for it. But in a 5v5 such as basketball and dota should be more of a TSR rating. For example:

TO score a 10 rating on abbadon u need 150 creep kills, 6 denies, 3000+ HP healing done and 4 kills and 4 death per game etc.... lets call that a 7/10 abbadon.. Anybody who plays abbadon whould be weighted on that scale.

This should protect people who are always 'cursed' to play with 4 noobs, who constantly lose with scores of 23-4, or 40-10 etc....
If you ever played a really good match. Just 1 weaker player will show in the end or mid game etc... having 2 or more is just really unbalanced in terms of ranked MMR.

But having MMR based on wins is totally stupid. I mean I was so glad when ranked came up. Then all sudden people started to complain that their allies suck in solo que, etc... not realizing they are the one that sucks even more. Then there's people who only play party mmr and totally suck at solo que etc.

In conclusion ranked games are not chess. It's not 1v1 it's more than 2 players so they should at least look at other games and see how they do ranking/ratings. Such as TRS in HON, to Leagues in Starcraft 2 & LoL.

I never played HON but from what I know about TSR, it has a glaring weakness. Abbadon can be played both as a carry and a support. So is alchemist, leshrac, kunkka, vengeful spirit and the list goes on. On such a scale a support abbadon and support alc will score extremely low when compared to a database of average player stats on these heroes. TSR punishes role swapping.

The problem about trash teammates is a very real problem, but it is an inevitable problem with the game, not with mmr. You see we don't play consistently. I can own in one game with this hero and feed the very next game with the same hero. Players perform differently and arguably, there is no statistic in the world to predict whether they will be playing out of their mind in their next game.

Now I kind of agree with cartographer. MMR based on win/lose perhaps isn't the most outstanding ranking system, but it is sufficient and effective.

if people rly cant understand that TSR is fucked up and the correct measure of skill is mmr then ur just fucked in life and i pity u

@我是不哭鱼
close games actually give you less MMR than stomps. because
stomps= you were much better than the enemy team, so you should be on higher rating than they are.
close games= you are nearly as good as the enemy team, even though you won, you let them dominate earlier stage of the game, so your rating should be somewhat close to theirs'.

The problem is not how mmr is determined -W/L, but how imbalanced MM can be sometimes.

@[HMOOB]FyrionPlax
And what if the game ends in 18 min? You still need the 3k heal with abaddon? And what if you're in team with a necro with mek and your heal is no longer necessary? Should you be penalized because you use your heal as nuke?

And if kda would be important, whenever there would be a stomp, instead of ending the game, people would camp the enemy fountain. And as bonus, supports would buy sentries and put them on the ground just to increase their "warding score".

Winning is the only relevant factor.

I don't know the exact MMR points spread everyone has but the +/- effect on your teams MMR versus your opponents will determine your MMR points outcome.

Your team has a lower MMR than your opponent and you win you get +/- a few points

Really I see this MMR is really equivalent to Blizzards Starcraft 2 MMR settings. You get about 50 % win/loss and as you keep winning, you keep getting harder opponents.

Example: I played SC2 and am in platinum rank 25 in my league. I played 6 games in the past 2 days of solo MMR and won the first 5 in a row. At first the first 3 opponents were about the same around rank 25 to 15 platinum.

The next two that I won was a rank 1 platinum, then a rank 75 diamond league. The 6th game I had last night that I finally lost my streak to was a rank 40 diamond league. Now typically from here I will be stair stepped back to around my rank (increases as I win, and increments points higher for winning against a harder opponent.)

I think if they had a more open view on peoples MMR in DOTA 2 as you're playing, you would see the same effect. As you win consecutive games, you will get stair stepped to harder and harder opponents until you plateau (or get weaker teammates or both) then it stairsteps you back to keep on track with the 50%.

I really don't get why this is such a unusual concept to everyone. Granted there are tons of variables compared to doing 3 different types of "characters" I can choose from in Starcraft 2 versus 110 on DOTA 2. I think the MMR is valid and places people where they belong. I think you could measure this better though if everyones MMR was visible POST game so you could see "oh I had some really crappy mates that game and "I" was supposed to carry the team because my MMR was much higher than the rest. Or "damn that opposite team had 2.2k average MMR, and my team was average 3k. We must of all lost a ton of games in a row to be here"

This isn't theory crafting this is the truth :P

Ive played over 2000 games and i get +- 25 mmr win or lose every single game, doesn't matter how well I did its the same rating. A ranking on a team game just based on W/L is retarded and shows what a slapdash lazy job valve have done with the MMR system. I had a game where I went 20 kills with lycan really early, and because my 4 other team mates were so brain dead and inept at the game couldn't follow up on the advantage we lost the game why should i lose the same mmr as those trashcans that lost me the game by feeding and doing nothing? I should get a reduced MMR loss say 15 instead 25. Same goes when you have that one guy who goes 0/9/2, you know the guy, the person who ruins the game with his badness that player should be punished with extra -mmr

^ Ive seen lots of games when the 22-2-10 guy lost it single handedly by being a retard and going in as rambo, rage buying back then being down for 120 seconds while the team cant defend couse every farm went to that one retard.. Should he lose less? How can the system determine this?

well thats not me, it should at least reward me because I had a high kd or gold gain, or ward whoring or something not justs a flat rating its just a lazy system and a bit pointless

Thats why im playing lycan in my roadbto5k lol

^^ Im okey with that, but tell me how can the system decide it :D Just having a high kda and gpm doesnt mean you are playing to win. Maybe you do, but not everyone. So the system would reward retards, it rewards retards now too, but its still LESS unfair, I mean, you cant actually exploit, if it would take kda into consideration, it WOULD be exploitable...

And there would be no reason to pick supports. Just pick carries with escape and should be fine.

Also, it would be easy to comeback: you fucked it up 3 games in a row? No problem! Fountain camp next game!

i dont know the exact way to program a good mmr system for dota 2 im not a programmer nor am i payed to work it for them, but there has to be a better way than just wins. Some influence to you total MMR win loss should factor in your kda, gold gain and and for supports warding and assists i dont know but you cant think that current way is reflects your skill lv at all.

Why don't you think of something then? Your ideas as have been explained are easily exploitable and also fail to take into account role swapping on heroes. That's why they're pretty horrible. Think of something better instead of just complaining if you really want to improve it.

MMR is so important, they should pay some genius from the nasa or some shit to study the best system to use in dota.

^^^ You dont have to be a programmer to have ideas. Im studying programming, im interested in machine learning algorythms too, but I still think ANY other system would be exploited.

Ok, lets say for carries, you somehow calculate kda and compared to the hero... So if a retard gets a good team creatin space and all he has to do is last hit (or as un ursa, right click, couse hes farming the jungle), then rambos into 5 men twice, (buyback between..) lose the game, then the perfectly playing supporting team should lose more than the retard?

Lets say you watch warding etc on supports. Sometimes there are 2 supports, sometimes just one. How can you decide if its whos fault to not buy the wards when you have a real support + Venge + Windranger? The real support buys most of the wards but needs boots and stuff, so who will you punish, if they arent bought on cd?

Also, you can consider stacking camps for supps, but sometimes you stack all game and not participate in teamfights, then your stacking was literally useless and you dont deserve it to be even taken into consideration in any means. Other times you usefully stack camps then your carry is a retard and not taking them (so it was wasted, but its not your fault..) so cant just check if the stacks were taken or not.

As heroes that can be played in mid and offlane, you will die more if you are offlaning. Also, you will die more if you are unlucky and gets ganked a lot.

And whats when someone plays support alche, then transitions into carry lategame? Ofc hes gonna have better KDA than a support alche who has to stay supp all game long couse of reasons... Also same can happen to venge, veno, etc. So what? You cant just decide at the beginning of the game whos playing which role... And in pubs, you cant use the 1-2-3-4-5 system, its just not there...

Also, why would you punish the guy who is dived for 15 seconds then dies literally hugging tower, when the team is at fault for not tp'ing? Sometimes you shouldnt even have gone to that lane, its your fault. Sometimes its the teams fault.

They have bots, so they actually have some kind of game-sense intelligence in the game. They can use that to derermine if you played well or not. If you did what the bot would have done, you get more points/lose less points. Do you really want this? If they could make better intelligence that has better game sense, you could play against those in bot modes. But they cant. And even if it would work well as a bot mode, couse better than these bots are doable, still, it wouldnt be able to give points for thinking out of the box, and thats a really important part of dota, to invent new strategies, etc.

If you lose, its sometimes really not your fault. But some games you will be carried, even if you fail hard. The idea behind mmr system is that your rating will slowly get to where you belong, where you can maintain 50%ish win rate. Dont think of your mmr as a number, rather its a range. 200-300 range, so if you are there, you probably belong there. If you think you can do better, tryhard your best hero for 100 games, you will be higher. Then start playing all your heroes again. If you keep losing mmr, you dont belong there. If you can stay there, thats your place, just you were unlucky. This system is good actually, just needs time. And NO system can figure out players mood, and it cant do anything with people that usually random their heroes...

The way how mm is set is ok, but i cant stand the fact that no matter how well you do (even with intentional feeder on your team), you still lose MMR. Individual performance should count (said by Valve alone), but it DOESN'T.

hey minazuki guess what? i dont have to think of anything i dont work for valve and i dont write the algorithms for the MM, but I can moan about it, god knows Ive spent enough money on the game I can at least hope that Valve can come up with a compentant mm that isnt based on 2 black and white stats with all the millions they have made of this game.

^ Have you even read my post? Its not fucking possible, can you get it? Anything not just based on w/l is NOT FUCKING POSSIBLE. Or the whole meta would completly change, noone would play carry-able heroes as support anymore, either not pick venge/veno/alche/etc or play them as carry, no more heroes without escape, supports will not save your life if it means risking a suicide, no more teamfights, just pick-offs and 5men doto all the time. No more split push only with cancer and tb (so more more cancer and tb pickers), etc... And even more flaming. You simply CANT do it. You can score people outside of mmr pretty well, mydotaskill does it with not that much flows... but if it would be in mmr system, it would change the game. Can you understand that? Completly change the game, not just slightly as a new patch, completly...

@vaeldiithia
When you can't argue with idiots you do this.

@Who The Fuck Let Scout In?
that 49.58% wr..

Boooobs

@luxon
wow if that is true it is fucked up. imagine a 4k mmr team matched with a 3.5k mmr team. the lower mmr team will be easily stomped and they still lose the maximum mmr.

im amazed by vaeldiithia D:

One does not simply make a better system.

Nah, a slightly better system is doable, ofc. But problem is, if it goes public, people will try to exploit. Even if they cant, they will try, it will ruin the game. If they will have a better hidden system, people will start to try to find out how it works, then even sell the possibly working exploits...

Btw you arent dont lose 25, I usually win/low 24-26 but ive seen 19 and 31 too in my own games and there were lots of topics about higher mmr players getting only 1 point or even 0 per wins, if mm cant find a balanced game.

@ 6_din_49

IF games ends early They have a scale for that. Say it ends in 18 mintues instead of the average 43 mintues.. then all the last hits & heals are scaled down. So it would be only 1/3 less etc...Also works in scoring if say... you are matched with way higher skilled players..

Like team A are 1.3x ur team B ratings.. so thus u will not be required to hit 100% required, but scaled down to say 70%.

I'm sure tons of geniuses can figure this out.

Hon's TSR too into considering wards per game & also assist. Which really changed the way people view KDR. Which is now shifted to KDA in dota 2.

@[HMOOB]FyrionPlax
And what level do you expect abaddon to be at min 18? What if he decided to max shield first? Then his "massive heals" are delayed by at least 5 minutes in a game that ends at min 18. What if he wants to go carry and max passive before heal? There's really no point to restrict skill builds.

Edit: also in a fountain-farming game he would be able to unleash tons of heals because his allies wouldn't be full hp most of the time. Or even better, some ally buy soul-ring or armlet to give aba something to heal, lol.

Dota 2 has a huge database of what is expected for a '10' abbadon, I don't know the numbers etc... but check this link out.
https://forums.heroesofnewerth.com/showthread.php?126454-True-Skill-Rating-(TSR)-mod-v-4-4-(01-20-12)

Sure TSR has flaws but it's 2x-5x better than win/loss MMR.

no ^

^^ TSR is useless. The only people that cared about TSR in HoN were the people who were shit but believed they were good. Typically 1700 rated carry/mid players.

KDA shouldnt be the determining factor. I have had plenty of games where I got slaughtered as a support because my carries were terrible. All that really needs to change is the +-. You can do really well and be the only reason you are in a game and your team can still lose and you get -25. It seems unfair. I understand getting - for losing, but a flat out 24 or 25 for losing regardless of how you played really isnt a reflection of your skills.

i blame gookstars for every loss

Lol you guys are all delusional.

MMR SHOULD BE BASED ON YOUR PERFOMANCE.

KDA, GPM, XPM, shit like that. W/L relies to much on your teammates

stop trolling u stupid fuck

i think what people dont understand is how quickly the MMR changes during the first 5 games... im fairly sure its about 1000 mmr in change per game.