- 25/26
- no
- u wont gain more than +25-28 per win regardless of winstreak (after placement games)
-
- 20 wins in a row will give u around 500 rating
- if someone leaves game teams usually gain/lose 12-16 rating
they dont do much really difference between placement and normal games was mainly this = 50/60 rating for loss/win instead of 25-28
http://dotabuff.com/matches/452103426
usualy i get/loose around 25. but this match lost 36, wo te fuq. radiant was 5 stack, me solo.
..and if game ends before 25th min you win/loose less points
+25~ per win in solo on average
-30~ per lose in solo on average
and it team one depends who i stack with
@ Iareyou 4543 urs ?
- only +23 last game :( , searched for hours and puts me into shitpool were i cant earn anything
(" ..and if game ends before 25th min you win/loose less points" that might be a reason...game was 25.35 - so i better w8 5 mins and fountain farm next time xd )
- i play solo
+-10-40 per game.
It seems like the entire team will get the same score, which depends on the average mmr of your team against the other (probably +-20 for same mmr). I've seen occasional games where one person got a much bigger or smaller change, possibly a high hidden uncertainty value.
^ then a winning streak would do it right ? win alot = high uncertainty value = loads of mmr not ?
I have seen changes in MMR for a single game after calibration ranging from +/- 0 to +/- 42. Usually its closer to +/- 25.
The real distribution of scores Valve has released http://blog.dota2.com/page/2/
This should be read as 5% are below 1100 MMR up to 99% are below 4100 MMR
5% 1100
10% 1500
25% 2000
50% 2250
75% 2731
90% 3200
95% 3900
99% 4100
The now hidden Normal, High, and Very High brackets are at
Normal = below 2250
High = 2250 to 2700
Very High = above 2700
Lots of people who post on dotabuff have scores in the top 1%. Many of them are so out of touch with ordinary dota players they don't believe the approximately 7 million people below 4k MMR even exist. Despite this, these same dotabuff posters often complain about those lower MMR ppl being placed in their games. It's a confused and contradictory mess.
The fact is its extremely difficult to put together a game of 5 v5 where all 10 have 5k MMR because such ppl are so rare. Sure I have a dozen like that on my friends list, but I recognize that this is not normal. When I was at NASA all my friends were aerospace engineers... nevertheless aerospace engineers are still very rare.
this MMR gain/loss is just terrible, man
http://dotabuff.com/matches/449057701
playing vs a very strong stack, game ends in a base race, I lose 27 points, wtf valve?
can some1 check the mmr of people i played with in the last 1-2 games just wanna know for fun (some of them have like 4.5k games and i have like 7 rofl)
Relentless 2700+ very high? I think that`s bugged alot. Worst player I have in friend list is 3000 mmr,and lowest player I ever seen is 2200..nobody can be lower than him 2200 even begginers 1 month players.
@Relentless, i noticed that high bracket starts at ~ 3400 and very high at ~ 3900, ur numbers are really wrong
No, my numbers are accurate. Originally Normal bracket was 85% of players. And Very High was the top 2.5%. But now Normal is 50% and Very High is the top 25%.
This change was made with the "First Blood" patch.
Bogi there are a huge number of players below 2k MMR. In fact there are as many below 2k as above 2.7k. How many Chinese citizens do you personally know? How many Indian citizens? Well, there are about 2 billion people in those countries. You not personally knowing them means nothing.
Your idea of what a new player really looks like is hopelessly flawed if you think its someone at 2.2k MMR.
https://www.youtube.com/watch?v=7cMSpmWpDFw This is what players with a few games of dota really look like.
A year ago when dotabuff used to show all the teams instead of only those with recent activity is was easy to see that there were millions of people playing dota below 2k MMR. These people did not magically disappear just because most of them have turned stats off.
http://dotabuff.com/teams/1086921 <- this team is still happily playing dota below 1k MMR as they have for a long time. But only their recent team games are still visible on dotabuff.
These millions of ordinary dota players are the ones buying the hats and the tickets and keeping dota 2 profitable so that it exists. You should be thankful for them rather than resentful or dismissive. Being skeptical of their existence is really quite absurd.
----------------------------------------------------------------------------
@Vaikiss Maybe 6 months ago I did a survey of about 5 thousand teams on dotabuff website to see who had stats on or off at what MMR levels. I found that 90% of those on teams with tMMR above 4.5k had stats on, dropping to about 50% with stats on for teams with tMMR around 3k and 10% with stats on for teams around 1.5k.
It is hard to be sure exactly what has happened with ranked games since so much more data is hidden now. But it does appear that a lot more people have stopped playing in the "team" games and moved to ranked games. Nevertheless dotabuff can still give us some rough idea of what the top MMR scores are probably like.
I will just estimated it from memory since calculation doesn't really make sense with so much data missing. I estimate that
~1 player in 2,000 is above 4500 MMR
~1 player in 3,500 is above 5000 MMR (top 2,000 MMR scores in the world)
~1 player in 15,000 is above 5500 MMR
~1 player in 70,000 is above 6000 MMR (top 100 MMR scores in the world)
@relentless sorry but ure very wrong. it IS possible to see what skill bracket ure playing in if u search the hero filter right after ur game. during calibration phase, my mmr started at 3,3 k and i was playing in normal skill, when i got it to 3,6 k after calibration i was in HIGH bracket and i only started to play VH bracket when my mmr got over 3,9 k. Test it yourself if u want to.
does anyone know what the metric is for how many points you gain/lose when you play solo ranked
is it based solely on game time or is it including the gold graph as well to determine how one sided a game was
in other words does prolonging the game actually get you more points are does it stay the same because the gold graph keeps heading in the same direction
Learning on your own takes forever, I don't even know if they had guides when i started so i stacked weird items and skills. Somehow there are people who played dota 1 and have lower mmr than me, I would like to know how many games these 5-6k mmr guys had in dota 1, because even though I feel I played a bunch, my mmr seems low even though I never played dota 1. And Relentless, what do you suspect the numbers for normal, high, and very high, (in terms of mmr) before the First Blood Patch?
The skill bracket designation is merely the average MMR for the entire game, all ten players. When I established what these numbers were I used data for thousands of players not a quaint anecdote.
I have been in Normal, High, and Very High games all having the same MMR... as have you and pretty much everyone who bothers to check. That is because the bracket is assigned to a particular game after the match is made as an average of the MMR of the players on both teams. Your personal experience checking a handful of games doesn't show you anything.
The definitions used for the skill brackets was dramatically changed at the First Blood patch and then they hid most of the data. It is still possible check for your game personally, but gathering enough games to get a real sample and see the patterns has become prohibitively time consuming and tedious.
Before the First Blood patch it was more like Normal = below 3500 MMR, Very High = above 4000 MMR.
I have no doubt you did check and found your games changing skill bracket as you said. Its just not relevant. It doesn't mean what you thought it meant.
i wonder how does retards who created all this mmr system gets employed in such a companies i mean
shouldn't ppls who actually know what they are doing be working with this kind of stuff ?
so far everything looks like my 10 year old sister could code better algorythms for matchmakings
As for the metric for the amount of points gained/lost Valve left that pretty vauge. But they did say that relative MMR averages for each team matter and the integral of the gold graph from the last crossover point along with "other coeficients". So maybe they are measuring kills/deaths/xp/towers/raxees + who knows.
The two things that they admit to measuring for the change in MMR they give you for a game would work as follows.
Max MMR gain = you were expected to lose because the other team has a lot higher MMR but you won anyway
Max MMR gain = you maintained a large gold lead all game long so the graph looks like diagonal line going toward you winning.
Min MMR gain = you were expected to win because your team has a lot higher MMR and won, no surprise
Min MMR gain = you just barely came back after being down a lot of gold and won a single teamfight at the end and won while at total networth below the opposing team.
So far it looks like everybody on a party, possibly everyone on a team gets the same change in MMR after calibration. I'm not completely sure that's true yet but several people have at least reported that is their experience.
I have around 3100 now, and was stacking with a friend around 3800 now and some other between these, and played in high/very high, mostly high whenever we checked. Ill soloqueue one tomorrow and will check, but I dont have my solo MMR yet, but if it was the same as starting team mmr, it should be 3k too, couse I started there after first team calibration game.
Those below 2.2k players exists... BUT Bogi is right in that, I think a kinda better beginner player can be at 2.2k easily, I have one on my list (added from this forum, has 94 wins and 2200~ atm, doesnt seem to be a smurf, so its possible...), but most beginners dont ask questions here, dont read guides, dont watch l2p videos on youtube, etc. I watched everything on Dotacinema and XVRogues channel after like 300 games played without any idea about what I'm doing, and thinking supporting means buying the courier at start and upgrading before 15 mins... Thats what my friends with a lots of games in dota1 and above 500 games here thought me... I have above 3k now, none of them has visible MMR yet but I realised that sniper/drow pickers started to disappear from my games, and they were there while I played with them, so probable below me... Any time I stack with them for 1-2 games, Im the only one with stats on dotabuff, so they exists...
Ah and I checked my list again, some of the "he played soooo well I added him on friend list" ppl have around 3k too, I added them before finding dotabuff, and they were the pros in my games. :D
"added from this forum, has 94 wins and 2200~ atm, doesnt seem to be a smurf, so its possible..."
Wait, is that me? XD
Relentless, I don't think you're accounting for the fact that High/Very High players play more games than Normal players. Top 25% of players probably account for at least 50% of matches played.
When ranked started, my score after calibration was ~3300. I always checked my skill tier fairly regularly, with hopes of getting into High, and until relatively recently, I had never solo queued into a High game (and indeed, those games were played at odd times when there weren't many players online). I absolutely know I have never solo queued into Very High, and I think I the only time I've ever gotten into Very High was via stacking with three or four Very High solo-queuers (yes it's depressing). A Steam friend of mine with ~3800 MMR queues into High/Very High, with more games in Very High. I'm not sure what the situation in US servers are, but in Australia I think the 3500 High/4000 Very High benchmarks still apply.
It is possible that Valve has actually changed it back to how it was originally. But I have not put in the large effort required to examine lots of games again. It's possible that Very High did return to the top 2 or 3%. I can't be certain since Valve made it so difficult to check for even one game.
I'm not disputing that some beginning players could go up to 2.2k. I believe starting MMR is 2k, but plenty of them go down from that point as well. Even among players who are into dota enough to play team ranked captain's mode games together 10% are below 2.2k. Nearly half of the ocean of casual players is below that mark.
There are about 1.75 million players between 2k and 2.2k MMR and 1.75 Million below 2k.
Some of you struggle to accept that you are top 10% or top 1% of dota players because you know plenty of people are better than you. But it is easy for both things to be true because of the large numbers of dota players. Being in the top 10% or even the top 1% is just not that amazing.
http://freedom.mysdhc.org/guidance/information/From%20High%20School%20to%20Pro%20Statistics.pdf
Dota is similar to real sports. The mark at which professional level talent in sport begins is usually close to 1 in 10,000. If you are in the top 1% that's nothing. The top 1% of 1%, that is pro level. Nevertheless being top 1% still really and truly means there are 99% of people who you can beat.
I've had games in which i got around 50 mmr, but then again I usually lose 30+ mmr every game I lose.
@Relentless
Assuming the top 2% play 10 games per day, while the bottom 70% plays an average of 2 games per week, how would your calculations look like? I don't think valve changed anything, except adding more players to "bottom 70%".
- 23-30 (27 as a mean)
- Hhell no
- Ur MMR increases
- Same as me
- He can increase it around 500 points
- U can't queue ranked for 5 games, about how much points, around 10 I think
A long time ago it was possible to see what brackets everyone was in. At that time you could just check the brackets lots of players from lots of teams played at lots of different MMR scores and see where the brackets tended to be. But now its just too hard to do that.
You could try to guess games played per day, etc. But that is so much harder. And you used to be able to see all the games played for the whole day and all the games played for all brackets. But now you can't because they only show a tiny fraction of the games from each day at a given time and only the very top MMR games for each hero and overall. There is just far to much data hidden now to make any reasonable guess. If they changed it back to the original 3500 MMR/4000 MMR for High/Very High, which maybe they did, there is no way to be sure. You can only check for your games personally to see what the bracket is and because of how they put very large ranges of MMR together in the same game its a very poor indication of the bracket definition.
When I last checked that I would be in Very High one game, Normal the next game, Very High the next game while solo queing and winning all 3 games. It just does not tell you anything about the definition of the brackets to check that way. You can only say "well I personally was put in Normal/High/Very High x%/y%/z% of games." The trouble is people see what they think is a pattern on 3 or 6 or 10 games and don't realize how easy it is for that to be coincidence and don't check enough games to see that there are exceptions and contraindications. As I said, maybe they switched back to 3.5k/4k or something new. I can't really know. The last time I was able to really know it was at 2.25k/2.7k.
Gotta love the auto-wins. You get like 20-23+.. but when ur stacked to win and lose.. -27 to -30? Gosh so stoopid.
I have only gone up since I started I think I have lost like 1 solo que match so far
So i just completed my solo ranked 10matches and i seem kinda lower than i expected. 9 wins and only 1 loss, yet i get 2911 rating. Is this normal to start so low? even though 9 wins 1 loss in ranked so far...
@BjC u dont start nessecarily low ..but what did u expect ..all the players u queue with or are in ur games are <50% and have low number of games
that doesnt mean that every specific player is bad but it s an indicator
@Iso: ye, you said it on the forum in a topic too, so I guess its public information, just didnt want to say your name couse Idk if you are okey with that :)
@BjC
they take all prior data into consideration of your MM not just the calibration. The calibration is just double checking. If you are lower that you think you should be it is fairly easy to rise(that is IF you are lower that what you should be)
Please sign in to post comments.
hey guys ,
- how much mmr do u guys get in general?
- does mmr gain decrease if i pick a hero alot ?
- how do winning streaks affect my mmr gain ?
- i always get +25-28 mmr per game
-can a good player get much faster to the top ( withing 20 games ) when the acc is ranked already ?
- what happens with my mmr if i leave a game ?
thx in advance