|
Hello People,
it is time to for a new Blog because there are huge issues with the current MMR System if you decide to play Support only. My Goal was to reach 6000 MMR by playing only support but it does not really work. I will explain here why it is so hard to reach this goal by supporting.
Let us start with some numbers first:
![[image loading]](http://i.imgur.com/Kczj34g.png)
This graph represents my MMR trend. The Blue line is my solo MMR. As you can see, it went up to 5,8~K and stagnated around 5,7k for like one Month.
I could feel that the level of my play did not really increase over this month so my MMR stagnated. But my MMR also did not drop because I followed certain rules which I go through later.
As you can see, over the last week I literally lost more than 400 MMR. What could be the reason? Did I lose so many games because I got worse? Did I rage queue? Just bad luck with teammates? What I can tell you from my perspective all of these points are not true. There is an easier reason which is not that obvious.
Dota 2 MMR Distribution:
![[image loading]](http://i.imgur.com/eU9RLSu.png)
Source: Dota2toplist.com
The given data is not 100% accurate since it does not affect every player but still gives an idea how the MMR is actually distributed along players.
As you can see, the closer you get to 6k, the less players are available.
That means filling a game with players of the same MMR gets harder and harder the higher you get.
This will bring us to the next point.
The Skill Distribution along Players:
To understand how skilled people are without looking at their heroes and roles they play, I am going to use a logical model which seems maybe strange to you:
People in the top 1% are ten times better than people in the top 10%.
People in the top 0,01% are 100 times better than the people in the top 1%.
In reality this model is kind of accurate.
(Thanks to CecilSunkure who came up with the idea some time ago)
Dota 2 Solo Match Making:
To understand how points are won and lost we have to take a look how the current system works. There will be a team of 9 strangers plus you. So there are 10 players in the game which determine the average rating of the game.
Losing will decrease your rating by 25 and winning will increase it by 25. That means if you win more than 50% of your games your rating will increase while losing more than 50% of your games will result in a decrease.
The system will also try to determine teams that are balanced MMR wise. That means that the average MMR of team1 and team2 will match. This fact will result in several issues for support players.
An easy example:
10 players which each of 3,500 MMR will result in a 3,500 average MMR game.
That seems pretty easy. There will be no skill gap between the players since they are all in the same range and the same percentile (Top 42%).
An Other Example:
10 Players, two are 5,500, eight are 4,500 MMR. The avarage MMR of the game will be 4700 MMR.
This seems also balanced. Every team got one 5,500 player and four 4500. But there is an issue with the percentile. The 5500 players are in the top 1,75% while the 4,500 players are in the top 18% percentile. That means that the two high MMR players are more than 10 times better than their team mates.
The conclusion here is pretty simple. On paper teams are balanced, in reality, the skill gap between the individual players within the game is immense.
This fact on its on is no issue itself. But when we apply this fact to a real Dota 2 game, we will see that this is a really big issue.
Game Impact:
To simplify how games are won or lost we will just talk about the game impact. Let us assume Core players have over a full game the highest game impact. That is also pretty logic because they get distributed a higher farm percentile than supports. Supports create space for Cores. The cores need to use the space to get a high game impact as the game goes on.
4,500 MMR Player vs 5,500 MMR Player
Let us come back to what I said earlier. The 5500 core player is more than 10 times better than the 4500 core player.
The outcome of the game will be pretty one sided. Since the 5500 core is 10 times better he will as a logic result have a 10 times higher game impact than the 4500 core.
Now we take a look at the supports. The 5500 Support player would also have a 10 times higher game impact than the 4500 support player. That seems compared to the Core player balanced. But there is a piece we miss: The overall game impact percentiles.
Let us assume some values:
Core players game impact 25% each.
Support players game impact 12,5% each.
So cores base impact is already doubled compared to supports. Now imagine you want to impact the game as a support harder than your opponents cores when both the support and the core have equal MMR. If we take a look at the 4500 vs 5500 players that would mean you have to play 20 times better but you are only 10 times better on paper. It is highly unlikely that you impact a game harder than any core as a support when both players are equally skilled.
I am aware of the fact that we still play a team game and a lot can happen but I want to keep that aside and just talk about the model itself.
The reality of high level MMR
Since we went through the models I want to share my experience while I was playing.
Hours you are allowed to play:
The image above showed that there are not many players left in the same skill range when it comes to high MMR. That results in certain hours you can play high MMR publics. These hours are usually from 5 pm to 11 pm cet.
These times are not set in stone of course, but are still the safest to avoid low MMR games.
Know the people:
There are not many people so you will meet the same people over and over again. That is very helpful. Since the "watch live games" list was changed some time ago, you can't check for the current MMR of the game. You will need to know the people and their MMRs inside the game to be sure it is no low MMR game.
Dodging low MMR games:
Well, I had this idea but the penalty of dodging games in general is pretty high so it is no valid option. And of course the fact that you can never be 100% sure that it is a low level game.
Supporting in low level games:
Not much to say about this. Play support in these game and thank everyone for -25 since whatever you do it does not really matter.
Low MMR vs high MMR Supporting:
I kept track of my loses and I literally hang around 20-30% win rate in low level MMR games while in high MMR games my win rate is around 70-80%.
Prove:
I could just made 10 screen shots from this nights public games... Literally 6 out of 10 games were 4k games. But here are some highlights:
![[image loading]](http://i.imgur.com/zRYj370.png)
![[image loading]](http://i.imgur.com/scHtX46.png)
Conclusion:
Laddering as a Support under the current conditions is like playing Russian roulette. You never know if people are able to carry. High MMR supporting is a pleasure, while low level supporting is exhausting and frustrating.
Do not get me wrong. When everyone in a game is within the same rating percentile, supporting is fine. Even in lower MMR games. The issue only occurs when games get heavily mixed up by the lack of high MMR players since the outcome of the game is already set.
Happy Laddering and good luck to my Support brothers and Sisters,
@Ler_GG
|
Kind of interested in knowing how accurate your model actually is in determining skill versus MMR, but regardless of that, I normally hover in the 5200 to 5400 range and I'm already noticing this. Seems like every game where I'm the lone 5k+ person supporting mid 4ks and the 5k person on the other team is playing a core role we almost always lose. Games where MMRs are closer, with everyone being 5-5.5k, tend to be much closer.
|
Lollllll streek getting flamed ahaha
|
i actually found 5->6 is easiest supporting
|
This sounds ridiculous. You are assuming far too much and you seem to be just basing your theory on personal experience.
The Skill Distribution along Players:
To understand how skilled people are without looking at their heroes and roles they play, I am going to use a logical model which seems maybe strange to you:
People in the top 1% are ten times better than people in the top 10%.
People in the top 0,01% are 100 times better than the people in the top 1%.
What? You say that in reality that this seems kind of accurate. I'm confused as to where and what this "logical model" was originally applied or based on. These numbers are completely arbitrary; it sounds like something that could be reasonable, but there isn't actually a way for you to know this. Then you go on to say -
Let us assume some values:
Core players game impact 25% each.
Support players game impact 12,5% each.
Again.... what? Where are you pulling this from? How can you possibly quantify any of this? I reject the assumption of the values you've associated with each role and you have no argument without the reader just going along with these made up numbers. Let alone the fact that "cores" and "supports" are extremely subjective terms and farming/level priority can and should be dynamic throughout the progression of a high level game.
Your screencaps of end game summary I find especially irrelevant... you just can't value someone's impact on a game by those stats alone...
|
On December 09 2014 10:56 trapani wrote:This sounds ridiculous. You are assuming far too much and you seem to be just basing your theory on personal experience. Show nested quote +The Skill Distribution along Players:
To understand how skilled people are without looking at their heroes and roles they play, I am going to use a logical model which seems maybe strange to you:
People in the top 1% are ten times better than people in the top 10%.
People in the top 0,01% are 100 times better than the people in the top 1%. What? You say that in reality that this seems kind of accurate. I'm confused as to where and what this "logical model" was originally applied or based on. These numbers are completely arbitrary; it sounds like something that could be reasonable, but there isn't actually a way for you to know this. Then you go on to say - Show nested quote +Let us assume some values:
Core players game impact 25% each.
Support players game impact 12,5% each. Again.... what? Where are you pulling this from? How can you possibly quantify any of this? I reject the assumption of the values you've associated with each role and you have no argument without the reader just going along with these made up numbers. Let alone the fact that "cores" and "supports" are extremely subjective terms and farming/level priority can and should be dynamic throughout the progression of a high level game. Your screencaps of end game summary I find especially irrelevant... you just can't value someone's impact on a game by those stats alone...
I just tried to build an understandable model why supporting is bad when you support in a low MMR game while you are high MMR yourself.
People in the top 1% are ten times better than people in the top 10%.
People in the top 0,01% are 100 times better than the people in the top 1%.[/quote]
This is reality - Deal with it.
Let us assume some values:
Core players game impact 25% each.
Support players game impact 12,5% each.
It is pretty obvious that cores impact public games way more than supports.
Lets say you play Skywrath Mage and you end up 15-2-5 and your opponents got a Void or Slark with equal stats.
Your game impact will be equal but yours will probably be not high enough to win the game because you literally get outcarried.
And when it comes to dota numbers and stats do not apply for some reason. Every time people come up with stats some people show up and go all the way "Hey boys its dota you cant measure stats blablabla...".
That is what a model is for.
|
I've had two streaks where I'd lost ~500 mmr within 1-2 weeks of playing without changing anything at all about my playstyle, playtimes, or heroes / roles played so your initial assumption really could well be down to bad luck / temporarily slightly worse personal form / whatever - I'm sure my bad streaks were. Also, it's pretty dumb to claim that cores have ten times more impact on a game than supports do. In fact, 'creating space' is probably MORE important in mid-level mmr games (let's be honest, 4-5k is not 'low') because virtually everyone has the basic skillset to lasthit, build proper items, and rightclick the right targets when the time comes for that. What most players at that rating struggle with is knowing when to transition out of laning, when to gank or use tps, when and where to ward well (as opposed to your usual river wards 24/7), stacking camps efficiently and so on and so on. Given your example of 4+1 vs 4+1 teams where only one player has much higher mmr, I am a hundred percent certain having the high mmr guy playing as carry would not always be better than putting him on a support.Of course there are games where its decided by whose carry is 'better', but there are just as many if not more games that are decided by whose carry had the better babysitter, whose mid had more space to roam, or whose side had smarter wards and smokes.
I just tried to build an understandable model why supporting is bad when you support in a low MMR game while you are high MMR yourself.
People in the top 1% are ten times better than people in the top 10%.
People in the top 0,01% are 100 times better than the people in the top 1%.
That's a pretty dumb assumption to make, by the way
|
On December 09 2014 11:02 Ler wrote:
People in the top 1% are ten times better than people in the top 10%.
People in the top 0,01% are 100 times better than the people in the top 1%.
This is reality - Deal with it.
That's a questionable metric of skill. Is the best player in the world twice as good as the second-best player in the world? (Is the world record marathon twice as fast as the previous world record?)
Leaving that aside, by that reasoning players in the top 1% are 1-10 times better than players in the top 10% (excluding top 1%) since they could be in the top 1.01%, top 3%, etc.
|
On December 09 2014 11:25 Salazarz wrote:I've had two streaks where I'd lost ~500 mmr within 1-2 weeks of playing without changing anything at all about my playstyle, playtimes, or heroes / roles played so your initial assumption really could well be down to bad luck / temporarily slightly worse personal form / whatever - I'm sure my bad streaks were. Also, it's pretty dumb to claim that cores have ten times more impact on a game than supports do. In fact, 'creating space' is probably MORE important in mid-level mmr games (let's be honest, 4-5k is not 'low') because virtually everyone has the basic skillset to lasthit, build proper items, and rightclick the right targets when the time comes for that. What most players at that rating struggle with is knowing when to transition out of laning, when to gank or use tps, when and where to ward well (as opposed to your usual river wards 24/7), stacking camps efficiently and so on and so on. Given your example of 4+1 vs 4+1 teams where only one player has much higher mmr, I am a hundred percent certain having the high mmr guy playing as carry would not always be better than putting him on a support.Of course there are games where its decided by whose carry is 'better', but there are just as many if not more games that are decided by whose carry had the better babysitter, whose mid had more space to roam, or whose side had smarter wards and smokes. Show nested quote +I just tried to build an understandable model why supporting is bad when you support in a low MMR game while you are high MMR yourself.
People in the top 1% are ten times better than people in the top 10%.
People in the top 0,01% are 100 times better than the people in the top 1%.
That's a pretty dumb assumption to make, by the way
Dumb? Its simple logic.
Oh and since you are 100% sure I am wrong just tune in when I stream in a few hours and you gonna see yourself :/
|
On December 09 2014 11:35 Ler wrote:Show nested quote +On December 09 2014 11:25 Salazarz wrote:I've had two streaks where I'd lost ~500 mmr within 1-2 weeks of playing without changing anything at all about my playstyle, playtimes, or heroes / roles played so your initial assumption really could well be down to bad luck / temporarily slightly worse personal form / whatever - I'm sure my bad streaks were. Also, it's pretty dumb to claim that cores have ten times more impact on a game than supports do. In fact, 'creating space' is probably MORE important in mid-level mmr games (let's be honest, 4-5k is not 'low') because virtually everyone has the basic skillset to lasthit, build proper items, and rightclick the right targets when the time comes for that. What most players at that rating struggle with is knowing when to transition out of laning, when to gank or use tps, when and where to ward well (as opposed to your usual river wards 24/7), stacking camps efficiently and so on and so on. Given your example of 4+1 vs 4+1 teams where only one player has much higher mmr, I am a hundred percent certain having the high mmr guy playing as carry would not always be better than putting him on a support.Of course there are games where its decided by whose carry is 'better', but there are just as many if not more games that are decided by whose carry had the better babysitter, whose mid had more space to roam, or whose side had smarter wards and smokes. I just tried to build an understandable model why supporting is bad when you support in a low MMR game while you are high MMR yourself.
People in the top 1% are ten times better than people in the top 10%.
People in the top 0,01% are 100 times better than the people in the top 1%.
That's a pretty dumb assumption to make, by the way Dumb? Its simple logic. Oh and since you are 100% sure I am wrong just tune in when I stream in a few hours and you gonna see yourself :/
If top 0.01% were in fact 100 times better than the people in top 1%, you would literally never see someone like Dendi lose in pubs. Like, he should be able to 1v5 at that point or something?
|
On December 09 2014 12:23 Salazarz wrote:Show nested quote +On December 09 2014 11:35 Ler wrote:On December 09 2014 11:25 Salazarz wrote:I've had two streaks where I'd lost ~500 mmr within 1-2 weeks of playing without changing anything at all about my playstyle, playtimes, or heroes / roles played so your initial assumption really could well be down to bad luck / temporarily slightly worse personal form / whatever - I'm sure my bad streaks were. Also, it's pretty dumb to claim that cores have ten times more impact on a game than supports do. In fact, 'creating space' is probably MORE important in mid-level mmr games (let's be honest, 4-5k is not 'low') because virtually everyone has the basic skillset to lasthit, build proper items, and rightclick the right targets when the time comes for that. What most players at that rating struggle with is knowing when to transition out of laning, when to gank or use tps, when and where to ward well (as opposed to your usual river wards 24/7), stacking camps efficiently and so on and so on. Given your example of 4+1 vs 4+1 teams where only one player has much higher mmr, I am a hundred percent certain having the high mmr guy playing as carry would not always be better than putting him on a support.Of course there are games where its decided by whose carry is 'better', but there are just as many if not more games that are decided by whose carry had the better babysitter, whose mid had more space to roam, or whose side had smarter wards and smokes. I just tried to build an understandable model why supporting is bad when you support in a low MMR game while you are high MMR yourself.
People in the top 1% are ten times better than people in the top 10%.
People in the top 0,01% are 100 times better than the people in the top 1%.
That's a pretty dumb assumption to make, by the way Dumb? Its simple logic. Oh and since you are 100% sure I am wrong just tune in when I stream in a few hours and you gonna see yourself :/ If top 0.01% were in fact 100 times better than the people in top 1%, you would literally never see someone like Dendi lose in pubs. Like, he should be able to 1v5 at that point or something?
If Dendi would play Mid in a 4,500 game on smt like Tinker he would probably solo most of the games, yes.
I once played on a 3,5k account to 4k and I literally went with 90% win rate and an avarge KDA of 15~ through this bracket so I believe my numbers are kinda accurate.
Dota is pretty simple. But you can just take a look at T1 and T2 Teams.. the skill gap is immense.
|
On December 09 2014 11:02 Ler wrote:Show nested quote +On December 09 2014 10:56 trapani wrote:This sounds ridiculous. You are assuming far too much and you seem to be just basing your theory on personal experience. The Skill Distribution along Players:
To understand how skilled people are without looking at their heroes and roles they play, I am going to use a logical model which seems maybe strange to you:
People in the top 1% are ten times better than people in the top 10%.
People in the top 0,01% are 100 times better than the people in the top 1%. What? You say that in reality that this seems kind of accurate. I'm confused as to where and what this "logical model" was originally applied or based on. These numbers are completely arbitrary; it sounds like something that could be reasonable, but there isn't actually a way for you to know this. Then you go on to say - Let us assume some values:
Core players game impact 25% each.
Support players game impact 12,5% each. Again.... what? Where are you pulling this from? How can you possibly quantify any of this? I reject the assumption of the values you've associated with each role and you have no argument without the reader just going along with these made up numbers. Let alone the fact that "cores" and "supports" are extremely subjective terms and farming/level priority can and should be dynamic throughout the progression of a high level game. Your screencaps of end game summary I find especially irrelevant... you just can't value someone's impact on a game by those stats alone... I just tried to build an understandable model why supporting is bad when you support in a low MMR game while you are high MMR yourself. Show nested quote +People in the top 1% are ten times better than people in the top 10%.
People in the top 0,01% are 100 times better than the people in the top 1%. This is reality - Deal with it. Show nested quote +Let us assume some values:
Core players game impact 25% each.
Support players game impact 12,5% each. It is pretty obvious that cores impact public games way more than supports. Lets say you play Skywrath Mage and you end up 15-2-5 and your opponents got a Void or Slark with equal stats. Your game impact will be equal but yours will probably be not high enough to win the game because you literally get outcarried. And when it comes to dota numbers and stats do not apply for some reason. Every time people come up with stats some people show up and go all the way "Hey boys its dota you cant measure stats blablabla...". That is what a model is for.
Well you're not providing stats, just assumptions. And those assumptions hide more assumptions behind them (the premise that impact of team members is additive for example). It's entirely possible you're right that it's hard to gain MMR as a support after a certain point, but you're not helping your point by coming up with this very dodgy model.
|
By looking at your screens and dotabuff it seems that you're playing the wrong heroes. You are trying to play too much like a proper 5 and honestly a pub is still a pub no matter the MMR. Look the chinese ladder, the support picks are always cancer stuff that still have the impact late game with farm (ogre wd silencer aa sk shaker visage enigma etc).
|
Northern Ireland22212 Posts
People in the top 0,01% are 100 times better than the people in the top 1%
this sounds so ridiculous given how at very high levels, it's only very small things that will separate players. to say that, for example a 7000mmr player is 100 times better than a 6000mmr player is insanity
|
On December 09 2014 18:36 ahswtini wrote:this sounds so ridiculous given how at very high levels, it's only very small things that will separate players. to say that, for example a 7000mmr player is 100 times better than a 6000mmr player is insanity
Insanity? Interesting ..
|
Hi Ler,
Thanks for your contribution, it is very insightful even though I find some metrics and definitions questionable, you gave a good idea about how MMR works etc..
But I couldn't fully understand why you said it is impossible to reach 6k mmr by playing solely support heroes. Is it because your impact is lower than a core for a given play?
Also Dota is a game which not all roles and heroes have similar impact, rather it is more viable for higher MMR player in the team to play more core roles(essentially mid or having farm priority sort of).
This is because, contributing the game by being better shows itself more when you play 1v1(mid) or having better farm to create an edge and ability to use it.
By the way, in the first screenshot, it seems your team doesn't seem like outplayed by the high MMR guy rather it seems like a collective effort.
I had a similar game last night where the teams are balanced according to average mmr but the other team consist of 4 ~3k mmr + one 4.8k mmr while ours is widely distributed between 4.1k to 3k. Other 4 guys have no impact in the game rather than providing meat shield while 4.8k storm got 38 kills and snowball into 1v5 and we lost. I imagine if he played a 5th role support, probably we'd roll them but whatever, who cares if average mmr is the same. we lost ~30 points maybe have a slight edge
http://www.dotabuff.com/matches/1079168482
|
One make shift solution I thought of is playing support who can transition ala WK, Alchemist, Silencer, etc. In their own right they are good cores and you don't need to play off your allies. I've had games where I was alchemist supporting and just go "fuck it" and just 6 slot in the next 20 minutes because my cores were too dumb to rely upon to close out the game themselves.
Obviously not the most ideal solution since the drawback to most supports who can transition is that they aren't as good at ganking/zoning out enemy heroes so it can detract from your early game to play them.
|
I'm a pretty mediocre support. I peaked at 5.2k MMR a few months ago (due to inflation that should be higher if I tried again). So it seems to value me correctly, I am not a good player.
I also buy the basic argument that support doesn't work once you are the highest rated player in your game. I started to feel that effect at my level already when playing at off hours.
|
|
|
5003 Posts
>> The Skill Distribution along Players:
>> To understand how skilled people are without looking at their heroes and roles they play, I am going to use a logical >> model which seems maybe strange to you:
>> People in the top 1% are ten times better than people in the top 10%.
>> People in the top 0,01% are 100 times better than the people in the top 1%.
I can see where your intuition is coming from but the bell curve is simply a distribution of where the players are -- it makes no statements regarding the relative (or absolute) differentials of skill (however the hell you measure it) because it is an ordinal ranking.
There could be many *other* supporting reasons why what you said could be true (for example, the average 0.01% player probably put in that much more effort and time and get more out of those effort and time than the average 1% player) but those are the reasons you need to appeal to, not the bell curve or some ordinal ranking.
|
|
|
|
|
|
EPT
