i always thought that deflation is obvious enouh to not even take into account other possibilities
'mmr gained' has no significance whatsoever because u cant differentiate between party and solo games; you got a sweet 'random sample' in your 'investigation' too
gained 550 in NS and lost 700 in HS and VHS
the net is +150
something is very wrong here
So much wrong shit that youre either a troll (in which case you wouldnt write so much) or a massive retard
i just see a lot of text that makes no sense
the inflation/delfation you could consider is accs getting boosted, and inflated smurf accounts. Although the numbers of these accs are so minor they have no effect you could call "mmr inflcation/delfation"
dont even understand how this can be a topic
so basically 3k is shitty now, was shitty last year and will be shitty next year too. good to know I will always be shitty.
I detailed how you can distinguish between solo MMR gained and party MMR lost in the case of Pyrion Flax. His solo MMR is in normal skill range. This is the only way it was possible for him to lose MMR in Very High and High Skill games, while gaining it in Normal.
But usually you can't distinguish party and solo que games, and even in PFlax example it is possible that some of his Normal Skill games were also party MMR. However, that is irrelevant. MMR gained or lost in both party and solo que comes from the same pool of MMR points. Because the pools of points are connected and freely interchange constantly the system doesn't care which one you are using at all. It makes no difference for inflation or deflation.
This sample is nowhere near big enough, nor is it intended to be representative according to statistical rules. Hence, if you can read, you would notice I made no statistical claims. It's just illustrative of what sort of things can and do happen to established players who are ordinary and in contrast to those at the top.
If there were an overarching inflationary tendency of some significance, ordinary players would drift upwards in MMR over time. If there was deflation ordinary players would drift down in MMR over time. Given the anecdotal evidence I've described one might argue further investigation is warranted. But in my opinion there is nothing to see here.
If you read this and concluded there was no firm conclusion to be reached (although we do see that pros are not accelerating off the top), then you understood what I wrote. If you additionally understand that data often does yield no certain results and yet still needs to be analyzed to establish nothing rather than something, then you may actually be a scientist.
OP, you know, people might take you more seriously if you boiled down your esoteric-sounding ideas to their bare essentials.
Of course i assume part of your goal of posting your epic monologues is to be taken seriously.
Unless you are in the middle of a whimsical soliloquy, in which case I will be sure to recommend you when some hotshot Hollywood honcho inevitably reimagines the gritty gamer reboot of Les Misérables featuring the rise and fall of Monsieur Rëlèntlēss Javért - the irresistable theorycrafting showman ;)
wassup guys . I just wanted to say that Relentless is exactly 3400 mmr and i loaded a game with him. unfortunatly the other team couldnt load and the game never happened. But i told Relentless : RELENTLESS ! FROM DOTABUFF? Hey ! and he never said any word. hes a 3400 mmr and im a 4.5k smurfing ( i suck too ) . but the attitude problem is real.
i play a game everyday on each account at least so i can feel better keeping my babys alive
Also, this part makes no sense at all and directly undermines the logic of your analysis:
"MMR gained or lost in both party and solo que comes from the same pool of MMR points. Because the pools of points are connected and freely interchange constantly the system doesn't care which one you are using at all."
Even though the system sometimes matches players queuing solo and partied-up players together, if you are investigating mmr inflation / deflation, solo and party games should definitely be considered separately - even 5-stacks playing under a team mmr.
The experimental operationalisations of the concept of "skill" in each of these different types of games have to be defined specifically and separate from one another, otherwise you just end up lumping all kinds of games together and whatever conclusions you draw will have weak / zero validity and generalizability.
But what am i saying?
You neither proposed an experimental method to test your hypotheses nor put forward any operationalized measures of "inflation" or "deflation". Until you do, you are no closer to reaching the actual answers than these ad hominem douches who credit/discredit anyone solely based on their mmr.
But it's an interesting topic, to be sure.
The best way we will know for sure is to pay Assange to get on the case. He can get some disgruntled dota 2 dev to eventually blow the whistle ;)
i bet 90% of dotabuffers dont know who Assange is. he is a pro 8k mmr living in a mountain
huge waste of time from all that "research"
at least, i knew sheever was awful, but man less that 50%, jesus
you really need to add a tl;dr.
is what you are asking would a currently x,000 MMR player beat an x,000 MMR player from 5yrs ago?
i'd say that a 4k player now would beat anyone from 5yrs ago. maybe 4.5k (exluding account buyers/smurfs of)
A 3k player now is about equal to a 3k player then and lower than that I'd say the 5yr old MMR depicted more skill than the current MMR. Basically the curve has stretched over time which is to be expected in a game where the knowledge and player base is increasing.
relentless is the best troll he waste energy and breath to type everyday and make sence in a post that doesnt make any sence. my english is 100% relentless comparted to his sotry. he need to wake up and realise he aint nothing but i grew up in the bottom
You don't have to write useless if you write relentless, like you don't say black nigs you just say nigs
@crudude: Every few weeks the shadow pool spits out confused, (prepubescent) boys who speak word salad and shit angst. But then again he's got nothing on the OP of this epic comedy:
@ OP - MMR is a ranking system that is supposed to be constantly updated and relative only to the current player pool.
Why do you care what MMR means between 2 unrelated time periods?
Lets take an extreme hypothetical example.
You play pee wee football at age 5. On a national ranking scale among your peers you have a score of 3000.
10 years later you and your pee wee peers are now 15 years old. On a national ranking scale among your peers you still have a score of 3000.
If 15 year old you played football vs 5 year old you, both 3000 ranking, does it matter that 15 year old should destroy 5 year old you?
is your 3000 ranking inflated/deflated because now you sprouted adult hairs?
Reese asks a legitimate question. "Then why now we have so many 7k people? They improved?"
I did already answer this several times in earlier discussions, but I will address this directly again. The MMR system takes a while to develop. You can't calibrate right to the very top, and winning enough games to max out your MMR takes a long time. This is why a few of the very best players in the world have still been rising to higher and higher MMRs over the past year.
Consider for example a couple of my friends who are over 7k MMR (Vroksnak and Beesa).
Two years ago these guys were only around 6k MMR. Did they get better since? They may have improved somewhat, but mostly they were able to rise because the system was slowly filling up with enough players who were close to their skill level for them to defeat. Until there were enough people at 6k MMR, the very best of the best had no one to defeat to distiguish themselves. Before Vroksnak and Beesa could only defeat 5k players because there were not enough people playing to provide a group of 6k players to beat.
Why didn't the pool of a few hundred 6k players exist before? It's because there were not enough 5k players. The MMR system is very large, 10s of millions of players... It takes time for it to fill out the tails of the distribution curve. Similarly when MMR began almost no one was in the 0 to 500 MMR bracket. But now there are lots of players there.
-----------------------------------------------------------------------------------------------------------
We don't have the data to prove it, but I suspect what Rocket describes is closer to the truth. The curve may be stretched, or flattened.
Unrequitted is correct. This is not a serious investigation. I'm just spending a little idile time pontificating while I wait for some codes to run.
I know some of you care very deeply about the precise MMR of forum posters, since you can't make rational judgements and so must believe arguments from authority. My current solo MMR is 3512, and party is 3522. Hopefully this information will allow you to properly calibrate the lack of respect for my posts you may wish to display. Unranked is slightly higher (near the low end of Very High) but no doubt that is of less importance to properly deriding forum posts.
@ B A C A E O G L W
Your analogy would be appropriate if the entire popluation "grew up" together. But instead it has come in waves. At first the players were nearly all experienced dota players from dota 1. By the time MMR was introduced there were a couple million players and almost all of them were experienced players. But in the last few waves, after each International more and more truly new players have joined the game. This past year many of those new players have been maturing and so there is some question as to how they impacted the distribution of MMR. When they started was the population heavy on lower MMRs? Has that bulge of new points spread out in both directions or have new players as a group all become better and deflated MMR as a whole?
We just don't have the information to know for sure. Why do I care? I'm just curious. We have been absorbing a lot more former LoL players this year.
I think it is possible using dotabuff data to track a significant sample of players from lots of MMRs over a year of play and see how people are typically moving up and down after they calibrate. But it's probably a lot more work than I am willing to do just for the sake of curiosity. I think it would be interesting to know what path a player takes, what heroes are the best to learn for MMR growth, what diffrences are there between someone who becomes a 4k + player and someone who drops or stays below 2k MMR?
Even though party and solo ranked scores are separate, the pool of points is the same and all the games exchange the same MMR points. So when someone raises their solo MMR it takes points from the collective pool of solo and party MMR points. And when someone lowers their party MMR it gives points to the collective pool of party and solo MMR points. It has the same effect on the overal system whether you play solo or party games.
@ OP - First off, don't let a number define you bro.
Secondly it sounds like you understand the change in Dota 2 population makes comparing MMR unreasonable, but you still want to track it. Also you want to analyze heroes and MMR even though they are constantly nerfed and buffed with each patch making that exercise unreasonable.
The purpose of Match Making Rank and here I have spelled it out for effect, is to Match you with similar players who are queuing with you for a competitive game. It is not supposed to do anything else.
Ofc I didn't read whole fucking topic, but I find it really interesting how always people with less skill/MMR find to analise and spend their time trying to seek for some sort of problem.
Instead, just get better at this game, it's much easier.
"I think it would be interesting to know what path a player takes, what heroes are the best to learn for MMR growth, what diffrences are there between someone who becomes a 4k + player and someone who drops or stays below 2k MMR?"
The answers to these questions are probably highly contextualized. Some factors to be taken into consideration could include:
1.Popularity of specific hero picks by region (and their win rates)
2. Players' behavior at 2k, 3k, 4k mmr, etc. (one can probably arrive at a set of highly generalizable behaviors provided the relevant hypotheses are rigorously tested).
With that said, the two potential problems about trying to quantify whether a player's skill if assumed constant, corresponds to the same mmr year-on-year are from my perspective these:
1. Short of analysing only 1v1 matches, how do you isolate individual player skill in the vast majority of games, i.e 5v5? Measures like gpm, kda, even win rate do not sufficiently account for outcomes (win or loss) which are primarily caused by team rather than individual actions. There is no feasible way to control for variances in team performance.
2. Therein lies the underlying question: does mmr measure a player's "skill"? The accuracy of any mmr-related study would be contingent upon establishing the validity of this number (i.e to what degree does mmr measure a player's skill)
Would be cool if someone from the dotabuff team weighed in - they may have a perspective on this, given how much of these data sets they deal with.
@ B A C A E O G L W
All great points, and entirely correct. I often find that the conclusions we would really like to be able to draw from a data set are insufficiently supported. Thank you for pointing this out.
However, I do think that we could still gain insight into the typical experience a dota player, or just a smurf could expect by tracking the experience of a sufficient number of individuals.
@Unrequited-
1. This is precisely why it is going to be so useful that the new Meta tab enables sorting by region.
2. see Meta tab again
------------------------------------------------------
I don't think 1 v 1 matches are ever relevant to understanding dota playing skill because in a real game you can't focus on the 1 v 1 matchup but for a few seconds at a time. If you do focus on a 1 v 1 situation, in a real game it kills you because you forget about the what the other heroes can do and likely will do.
MMR measures how difficult the pub dota games you won and lost recently were. It's not very good at measuring the skill of a player... but it is helpful for you to know the average skill of you opponents. It's not really for measuring individuals, but populations. So you know that as your MMR goes up the games are harder to win, require more skill to win. Is your skill greater in those harder games? Maybe yes, Maybe no. But the games themselves are harder because the average skill of everyone in the game is higher at higher MMRs.
stop 
Чтобы оставить комментарий, пожалуйста, войдите в систему.
MMR Inflation means that the MMR corresponding to a certain skill level in dota is going up. People who have not really got any better have higher MMRs over time. Some people think that experienced players are accumulating extra MMR from all the new players, who join and then lose it. Or some think that account selling is inflating the top scores.
The opposing view is that the community of dota players becomes more experienced over time and therefore the same MMR score of a year ago is actually a higher skill level than it used to be. Some think that smurfs are deflating MMR since a lot of players who should be ranked higher are actually playing in lower MMR games taking MMR, but then they stop playing once they reach games of their level and make a new smurf rather than losing the points back to the system. Effectively smurfs are removing points from play deflating MMR.
A third perspective is that MMR is a nearly zero-sum game and these effects are trivial. MMR means the same thing this year as a year ago.
To investigate these effects. Let's consider the experience of some well known dota players.
The Commentators:
_______________________________________________________________________________
Pyrion Flax
https://www.dotabuff.com/players/9332911
in the last year has played 510 ranked games and gained an estimated 150 MMR. He gained 550 MMR in Normal skill games and lost 700 MMR in High and Very High skill games. So it appears he did improve, winning a lot of solo pub ranked games, but still caused his friends to lose a lot of party games. Net +150 MMR.
Sheever
https://www.dotabuff.com/players/37147003
Played 1,239 ranked games in the last year, gaining an estimated (+175 MMR). +375 MMR in Normal. +75 MMR in High, -275 in Very High. Sheever also makes good players lose party games.
Nahaz
https://www.dotabuff.com/players/61214244
Has actually lost in all level of ranked games over the past year on 266 ranked games losing an estimated -1250 MMR.
Tobiwan
https://www.dotabuff.com/players/33366656
Played 220 ranked games gaining an estimated 500 MMR in a year.
SirActionSlacks
https://www.dotabuff.com/players/68186278
Played 720 games gaining an estimated 550 MMR
The Pros:
-----------------------------------------------------------------------------------------------------------------------
The pros MMR is already so high it's unlikely their wins are worth the full 25 points ordinary players recieve for a win.
Admiral Bulldog
https://www.dotabuff.com/players/76482434
Played 752 ranked games this year gaining +184 wins, which would be +4600 MMR if they were counted as +25 per game.
Overall Admiral Bulldog would be 5325 MMR Above his calibration point if each game was worth +25.
Miracle
https://www.dotabuff.com/players/105248644
Played 1848 ranked games this year gaining +276 wins, which would be +6900 MMR if they were counted as +25 per game.
Overall Miracle would be 9950 MMR Above his calibration point if each game was worth +25.
XBOCT
https://www.dotabuff.com/players/89625472
Played 470 ranked games this year gaining +94 wins, which would be +2350 MMR if they were counted as +25 per game.
Overall Miracle would be 6825 MMR Above his calibration point if each game was worth +25.
Wagamama
https://www.dotabuff.com/players/32995405
Played 1148 ranked games this year gaining +70 wins, which would be +1750 MMR if they were counted as +25 per game.
Overall Waga would be 5875 MMR Above his calibration point if each game was worth +25.
Illidan Stormrage
https://www.dotabuff.com/players/26316691
Played 596 ranked games this year gaining +74 wins, which would be +1850 MMR if they were counted as +25 per game.
Overall Waga would be 7675 MMR Above his calibration point if each game was worth +25.
------------------------------------------------------------------------------------------------------------------
So we see that the commentators are only a few hundred MMR points from where they were a year ago, and they may have legitimately improved that much over a year ago. Meanwhile the system will not let the pros take anywhere near as many MMR points as they might like or pros would be from 10k to 15k MMR already.
It is clear that the system is bounded, limiting the pros from aquiring inflationary amounts of MMR from lower level players. Similarly 6ks are limited in taking MMR from 5ks who in turn are limited in taking it from 4ks and so on down. Because there have to be people losing at each MMR level for anyone to move up to a higher level there is stability.
We also see that long established players of no particular talent are not losing or gaining much MMR. They have only moved up slightly approaching the 50% winrate mark. This suggests that despite the pro's filling out the top tail of the curve the curve overall is not moving.