MMR Experiment
Currently there is no better way to determine one's skill than MMR. Though the system has a multitude of flaws, it is often stated by both professionals and pubstars alike, that in most cases you are where you belong, when it comes to MMR bracket. You win or lose as a team and sometimes people on your team misplay. But you will be lying to yourself if you think that there was no game in your recent history where it was you who has failed to have enough impact to achieve the victory.
Destroying the ancients is what this game is all about - after the calibration process is over, it does not matter how much damage you have dealt to the enemy heroes and structures, or how high your GPM was. If you win as a team - all of you will get your points, regardless of impact, perceived or objective. Strangely, though, the initial calibration does take into account these factors, and I have always been curious about the extent to which these metrics can influence the freshly calibrated MMR.
I am a lucky person - I have managed to convince my girlfriend to play Dota 2. In fact, the idea of this blog post came after one of the discussions about ranked play I had with her. Both of us have created fresh accounts for the experiment and have only played party matchmaking AP until we reached level 13.
Initially, as I was working on this side project of mine, MMR and calibration were pretty much the only thing that I wanted to discuss, but playing with a lot of newer players that have not yet been turned into your stereotypical hardcore gamer proved to be an equally interesting topic, hence some of the post will be dedicated to discourse on my personal experiences as a Dota/HoN/Dota 2 veteran, as well as the experience of my girlfriend during the calibration.
Experiment
At the time of writing my Solo/Party MMR on the main account are 4281/4217. I am pretty confident in this MMR representing my actual skill, since I have been floating around 4.3k for quite a while now, only reaching around 4.6k maximum in either.
My girlfriend, on the other hand, had no Dota 2 experience whatsoever - no LoL, HoN or Dota 1 neither. She was not completely unfamiliar with gaming though, being a huge Borderlands fan, as well as having some previous Warcraft 3 experience, so she was pretty confident in handling the mouse and point-and-clicking.
She has quickly grasped the concepts of laning and did not make most of the mistakes newer players make - such as auto-attacking creeps or drawing the aggro away. On top of that, she started being a decent support around our 40th game, with somewhat effective (though not necessarily thought-out) harassing and denying.
When it came to fights, the panic mode managed to do wonders - pressing all buttons to focus the nearest enemy turned out to be a rather effective strategy for 
Skywrath Mage (hero of her choice) at this level of play. Slowly the concepts of timing, ganking, focusing and positioning have started to settle in and by the time we reached level 13 (roughly 100 games) I had a decent, albeit at times unreliable lane partner.
When it came to actual hero damage in games, she was never at the top, but rarely at the bottom of the list. In most of my games with her, I have managed to deal the most damage to the enemy heroes on the team, even playing supports. Our KDA were significantly apart, with her average being at 1.8 and mine at around 5.1. Another huge difference was in average GPM and XPM. Mine and her were 523/251 GPM and 610/340 XPM respectively.
We have won 7/10 of our calibration games for party MMR. I have calibrated at 3648 while she got 2358 and it tells a rather interesting story, that even given the same Winrate against same opponents in the same games, there can be a lot of disparity between MMR's after calibration.
I did not expect the difference to be that significant. Also, I did not expect her MMR to be that impressive - I know a decent amount of people around 2.5k who are objectively better at Dota 2, compared to my girlfriend, who has only played for a month.
As I have stated previously, I am pretty confident in my 4.3k MMR - this is where I belong and it is the level I enjoy playing at the most (early in the morning I occasionally get matched up with people either way over my skill, or way lower). Since all of the games we have played together were in the "normal" skill bracket, I believe I could not have calibrated higher than 3.7k - more or less a cut point between high skill and very high skill games.
Similarly, I do not agree with almost 2.4k party MMR on my girlfriend's account - it is currently extremely hard for her to play alone or in a party without me and her Win Rate has dropped from ~60% to ~53.5%.
What conclusion it leads to is that depending on the people you play with during unranked play and calibration, your MMR can be either inflated or deflated, drawing both of you closer to the average. However, a lot of it is still based on your actual contribution to the games you play - all of the metrics we calculate on our website do matter and, in a sense, are representative of the skill of a player. These metrics include, but are not restricted to:
- KDA (5.1 vs 1.8)
- Hero Damage
- Hero Healing
- GPM (523 vs 251)
- XPM (610 vs 340)
So next time when you think of complaining about your teammates being terrible or you belonging to a higher skill/MMR bracket - remember that there are always things in any given game you could have done better, and there have been ways for you to make more impact on the outcome of the match, and you not utilizing these ways and/or failing to coordinate as a team is what led to an eventual loss. You cannot improve the teammates you get matched with unless you improve yourself first.
Experience
Disclaimer: all of the points in this section are personal experiences and as such are quite subjective. All of the games have been played on Russian servers and the experience may vary for different regions.
People actually listen if you talk to them nicely in the lowest of brackets. If my teammates were as cooperative in games from my main account, I am sure the overall experience would be a lot nicer. My belief is that to a certain degree this inability to listen to your teammates comes from the fact that once you have reached a certain milestone, be it 3 or 4k you start thinking that you are good at Dota. And it is not the case.
At first, when you come to Dota you have a full understanding of the fact that you are terrible at it - the game is hard. Then, slowly but surely, things start to get more understandable and you start forming your own opinions about things, as opposed to listening to guides and/or your teammates. This is where a lot of thing go terribly wrong - not only are a lot of guides, which serve as a backbone for your opinions are terribly outdated, but there is also a problem of them being terrible in the first place - watching my girlfriend watch another video on how to play Skywrath Mage makes me cringe every time.
I am not going to lie - I have suffered from the same disease - not being open-minded enough is something I am trying to tackle. But as our journey through unranked play to level 13 progressed, so did the certainty with which the players on our team have been defending their ideas, which in most cases were objectively suboptimal - be it itembuilds or strategy. I know for sure that I am not a great player, but I have coordinated a fair share of teams and have played the game for more than a third of my life - there are things that simply don't work in certain situations and failing to understand it and failing to listen to a well-constructed argument is what makes a lot of people in the middle MMR brackets lose.
Another interesting point about the players in different skill brackets is their attitude towards women in Dota. While the first games where my girlfriends identity has been revealed have not exactly been this, there was an appropriate amount of respect towards her, sometimes on the border of "white-knighting", which was nice and kind of cute, to be honest.
Once the smurf detection has kicked in and we got into what felt like 3k unranked play, things have drastically changed. The amount of sexism that would previously seem normal (by Dota 2 standards) to me, under the fact that it was directed at someone I knew and respected has completely changed my point of view on the issue. Dota 2 is primarily a male e-sport, dominated by male players, casters and persons of interest. It does not, however, justify feeling of entitlement and gender exclusivity a lot of players tend to develop under the influence of something, which is yet to be discovered.
Given the very noticeable change in attitude and a decently sized sample of around 30 games before things calibrated to ~3k, I believe this sexism is not innate to the gamers who get into Dota, but is somehow learned in the process of advancing in the MMR brackets. Long story short - whenever we play a game together, we generally mute all of the opposing team at the start and have a "one-strike" policy on muting teammates as well.
https://www.youtube.com/watch?v=zparciz8Res
MMR are irrelevant ....like this site statistics ...pub ranked is same as pub ....statistic only work for pros not for casual users.....
My issue is, pre calibration you and your girlfriend are getting a different number of "points" per game, and after the calibration you will forever be the same amount of points ahead of her regardless of play since everyone on the team always gets the same points and what for? to save peoples feelings? you could even switch profiles you play on hers and vice versa and it wont change anything in the disparity of points as long as you always play together, and that is an inaccurate representation of skill.
MMR Calibration is all about GPM/XPM before you get 13 and after it, if you play Meepo/SF/Alchemist every game even if you lose or win, you'll get 4.5-5k MMR range.
That's why people would sell their accounts cause they would play against other people when leveling and when reaching at level 13 having an average of 800-900 GPM/XPM they will go straight to 5k-5.5k and just do a few games more and get 6k easily.
Играл давно в первые разы ММР и вышло так что из 10 матчей 6 выиграл и 4 проиграл. Но после 10 игры мне дали 4.4к ММР не взирая на то что я проиграл 4 раза) Видемо вы правы)
ALL THESE BUTTHURT 4k GARBANZOS HERE TALKING HOW RMM IS BULLSHIT Kappa
OP, you miss exactly one crucial point with your MMR justicifcation. Imagine a clone of you with exactly skills. Imagine you play with him together from 0-13 and then ranked. Your role: Carry (#1-#2). His role: Hard support (#5). Both of stick true to your roles.
Obviously your clone hard support will have lower:
Hero Damage
GPM (523 vs 251)
XPM (610 vs 340)
And Hero Healing is lot less ratet then Hero Damage.
Resulting in ~500-1000 less MMR. Fair?
the MMR system fails to reward active players and roamers.....please sit on lane and kill creeps for high MMR
I usually playing only supports... my win rate... my MMR... dammit.
@NoName
"And Hero Healing is lot less ratet then Hero Damage."
Thats something you cant know.
But i agree, not all supports have healing spells, and mesuring the game impact of a support who is not feeding can be hard and is imo impossible for an automatic system.
I liked the experiment, but unfortunately there were 2 players with a different skillevel, so the conclusion is hard to make.
You dont know how different heroes are rated, or if supports get a "bonus" over carries, or if the system tracks who is farming and who is buying wards etc.
Even though supports will have lower KDA, 1.8 is not great and good supports will take free lanes to get at least some farm.
I would be really interested in seeing the experiment and the MMR with 2 equally skilled players.
great guide
MMR Calibrations varies. The best thing to experiment is creating a smurf or fresh account. With only 1 hero to play. Given sniper for example. If you're kda with that hero is good. You will land on 5~6k at most. My friend tried this and boom. He got easy 6k mmr. Then he sold it to someone else.
I acknowledge my personal failures (many of them, at least) and assume 1/5th the responsibility for every lost match. But I do feel that my solo MMR is currently less than it should be.
I started at 2.6k solo MMR, and was floating within 2.4k-2.5k for a pretty long while. Until I caught a losing streak of like 10 games in a row that brought me to ~2k MMR. At which point it felt like the average level of play did drop, I played some more games with WR about 45%, if not 43, and now I'm at 1670 solo MMR where sometimes the teammates are actually good, and sometimes they're HORRIBLE. It feels like I'm now in a pit that I can't get out of, and my WR keeps plummeting. I think I've had around 600, perhaps even 700 games on record when MMR was introduced, and WR of over 49.0%. Now, after playing 164 ranked games, I have 870 games total and 46.5% WR. The ranked WR is 38.4%, non-ranked WR is 48.0%. And I was mostly playing heroes that I actually know how to play, with only a couple exceptions.
Any thoughts? Did I get unlucky and got trapped in a pit, or am I where I belong?
"But you will be lying to yourself if you think that there was no game in your recent history where it was you who has failed to have enough impact to achieve the victory."
Sorry but you're talking absoleute shit. Any of my games recently where I've lost have been either of the following
A) I decide to be nice and pick a support and instantly fucking regret it when I realise our midlane/carry is a slavic piece of shit with barely 1 cs/minute who says 'omg omg omg omg omg' over and over again and 'gg' if so much as a trade happens. I make ganks happen. I try to shut down carries. I buy all the wards and smokes and dust. Nothing comes of it.
B) I do make a huge impact on the game, stomp hard and can't win because of throws. Lemme give you example. Pop BKB, initiate, friendly naga uses ult, suddenly no more BKB.
'Enough' is such a blanket statement. It's meaningless. Is 99% 'enough'? Because it's how demanding some of these games are with slavic retards.
REGION LOCK WHEN
@Khanzeal: Every retard has a name. Not nationality, religion or anything else. It is true, from my experience, that the percentage of retards is higher on RU and EU East server, but there are also lots of good and great players. Conversely, there are many idiots from US. Blocking players from accessing certain servers based on their region of residence is counter-productive. What we really need is a black list, so that you can tell the matchmaker to never ever place you on the same team with a specific player again.
"under the influence of something, which is yet to be discovered."
That my boy, is internet anonymity combined with a dash of good ol' group behaviour. These things are well documented in psychology and unfortunately it isn't even an indicator of who is or is not a nice person. You find this especially in matchmaking parties with friends - they'll say stuff they'd never say to someone's face- especially with team mates who make mistakes. But if one of your mates makes a mistake, "it's OK but you should have done this".
Dota is a great, fun game but unfortunately its skill level creates a sense of elitism that in turn creates more and more negative behaviours creating a cesspool. Which of course turns away nice(r) people from playing and you get a higher percentage of douche-bags.
http://www.simplypsychology.org/zimbardo.html
could you explain why smurf accounts have higher mmr?
https://www.dotabuff.com/players/204094407/matches How much mmr this accaunt can rich?
my boyfriend always make pilot on my games tho
If 45% winrate how many MMR ??
http://www.dotabuff.com/players/238330885/matches HOW MANY MMR THIS PLEASE TELL ME :) IM LVL 12 :)
I'm surprised that you only got 3.6K MMR when you have 5.1 KDA, 500/600 GPM/XPM, 65% W/L and I'm surprised that you got those stats while you have a person who just started on your team.
aw why always loss all my game :(
aw why always loss all my game :(
this article is pure bs