Help Support Our Growing Community

DOTAFire is a community that lives to help every Dota 2 player take their game to the next level by having open access to all our tools and resources. Please consider supporting us by whitelisting us in your ad blocker!

Want to support DOTAFire with an ad-free experience? You can support us ad-free for less than $1 a month!

Go Ad-Free
Smitefire logo

Join the leading DOTA 2 community.
Create and share Hero Guides and Builds.

Create an MFN Account






Or

I'm a sad person.

Please review our General Rules & Guidelines before posting or commenting anywhere on DOTAFire.

Thread Locked

This thread has been locked by the moderators, you cannot reply to it.

Forum » General Discussion » I'm a sad person. 26 posts - page 2 of 3
Permalink | Quote | PM | +Rep by Hamstertamer » October 27, 2014 4:36am | Report
There IS such a thing as MMR hell. Since MMR evolves so slowly by just -25 or +25, it takes a looong time to grind out matches before you get matched with people of your skill level.

I'm going to prove it with math.

Suppose you're Arteezy, but because of some bug in Volvo's code you've been calibrated at 1K MMR. You are now playing with 4 russians who feed couriers against 5 peruvians who feed couriers. It would make sense to assume that you get a 60% winrate in a situation like this : if you suppose that each player has a similar contribution to his team, if 5 scrubs against 5 scrubs gives you 50% winrate and 5 Arteezys against 5 scrubs gives you 100% winrate, 4 scrubs + 1 Arteezy against 5 scrubs gives you 60% winrate if you take the average.

If you assume +/- 25 MMR each game, in average you gain (0.6-0.4)*25 = 5 MMR each game.

So to gain 1000 MMR, you need to play...200 games.

So in order to reach 5K MMR, Arteezy would have to grind out 800 games.

But you're not Arteezy. You're just probably some good-but-not-perfect player who sits at around 54-55% winrate like lots of people I see around.

So if you're a 5K player who gets calibrated at 1K MMR, you will have to grind out 1600 games. Assuming 1h/game (6.82 obliges), that's 1600 hours of doto which assuming you play 12 hours per day and have no life is 133 days (4 months and a half) of doing nothing but play doto while not going to school or getting fired from your job just to get the MMR that matches your skill level.

This is MMR hell. It's real. If you get calibrated way below your actual skill then there is no comeback possible besides making another account.
Strategy guide : Anti-pubstomper guide.
Hero guides : Spectre , Windranger and Clinkz
== Broodmother guide out! ==

Hamstertamer
<Editor>

Awards Showcase
Show more awards

Memorable (89)
Posts: 2620
View My Blog
Permalink | Quote | PM | +Rep by Unscathed » October 27, 2014 5:09am | Report
Hamstertamer wrote:

There IS such a thing as MMR hell. Since MMR evolves so slowly by just -25 or +25, it takes a looong time to grind out matches before you get matched with people of your skill level.

I'm going to prove it with math.

Suppose you're Arteezy, but because of some bug in Volvo's code you've been calibrated at 1K MMR. You are now playing with 4 russians who feed couriers against 5 peruvians who feed couriers. It would make sense to assume that you get a 60% winrate in a situation like this : if you suppose that each player has a similar contribution to his team, if 5 scrubs against 5 scrubs gives you 50% winrate and 5 Arteezys against 5 scrubs gives you 100% winrate, 4 scrubs + 1 Arteezy against 5 scrubs gives you 60% winrate if you take the average.

If you assume +/- 25 MMR each game, in average you gain (0.6-0.4)*25 = 5 MMR each game.

So to gain 1000 MMR, you need to play...200 games.

So in order to reach 5K MMR, Arteezy would have to grind out 800 games.

But you're not Arteezy. You're just probably some good-but-not-perfect player who sits at around 54-55% winrate like lots of people I see around.

So if you're a 5K player who gets calibrated at 1K MMR, you will have to grind out 1600 games. Assuming 1h/game (6.82 obliges), that's 1600 hours of doto which assuming you play 12 hours per day and have no life is 133 days (4 months and a half) of doing nothing but play doto while not going to school or getting fired from your job just to get the MMR that matches your skill level.

This is MMR hell. It's real. If you get calibrated way below your actual skill then there is no comeback possible besides making another account.


Agh... You dont know how that scared the **** out of me... Not even kidding
Don't Worry, Be Happy

Late credits to Janitsu for the sig

Unscathed
<Editor>

Awards Showcase
Show more awards

Remarkable (47)
Posts: 3432
View My Blog
Permalink | Quote | PM | +Rep by Dimonychan » October 27, 2014 6:10am | Report
Hamstertamer wrote:

There IS such a thing as MMR hell. Since MMR evolves so slowly by just -25 or +25, it takes a looong time to grind out matches before you get matched with people of your skill level.

I'm going to prove it with math.

Suppose you're Arteezy, but because of some bug in Volvo's code you've been calibrated at 1K MMR. You are now playing with 4 russians who feed couriers against 5 peruvians who feed couriers. It would make sense to assume that you get a 60% winrate in a situation like this : if you suppose that each player has a similar contribution to his team, if 5 scrubs against 5 scrubs gives you 50% winrate and 5 Arteezys against 5 scrubs gives you 100% winrate, 4 scrubs + 1 Arteezy against 5 scrubs gives you 60% winrate if you take the average.

If you assume +/- 25 MMR each game, in average you gain (0.6-0.4)*25 = 5 MMR each game.

So to gain 1000 MMR, you need to play...200 games.

So in order to reach 5K MMR, Arteezy would have to grind out 800 games.

But you're not Arteezy. You're just probably some good-but-not-perfect player who sits at around 54-55% winrate like lots of people I see around.

So if you're a 5K player who gets calibrated at 1K MMR, you will have to grind out 1600 games. Assuming 1h/game (6.82 obliges), that's 1600 hours of doto which assuming you play 12 hours per day and have no life is 133 days (4 months and a half) of doing nothing but play doto while not going to school or getting fired from your job just to get the MMR that matches your skill level.

This is MMR hell. It's real. If you get calibrated way below your actual skill then there is no comeback possible besides making another account.


Yes it's working slowly, and I'm not satisfied with that too, but, no matter how slowly, you RISE. MMR Hell is when you are surrounded by scrubs, but can't reach that 55% winrate and you are like - win-lose-win-lose, or win-win-win/lose-lose-lose, you are victim of streaks. This is not possible unless you take a small amount of games(like if I took my max losestreak of 7 and said Dota is broken), or you are of the same skill level and you are where you belong to.

And your calcualtion are... no, I won't say they're completely wrong, but they are very innacurate. Like if 5 rtz's vs 5 scrubs = 100% win, this doesn't mean rtz+4 vs 5 will yield 60% winrate which seems logical but is not true on practice. Thereby, all of the next calculations are impressive, but not quite close to reality.


Let's see legendary Juice, who mamanged to get 2500 rating during 144 games with 85% wr(while judging by your calculation, rtz(who I think is more skilled than Juice) would have to grind trough 500 games for that result). Not to mention 1000->2000 is way easier than 4000->5000, and Juice's experiment proves that.

My experience proves that too: ironically enough, a half of an hour ago I have played my 100th ranked match in this month. With a winrate of 61%(61-39), I now officially declare myself Arteezy II! 8{D

But seriously, that means (61-39)*25 = 22*25 = 550 MMR gain in a month, and due to me having the winrate similar to that of one was used in your calculation, yes, in 200 games I would gain 1100(but I wouldn't, as 3k is harder) MMR. And I don't even play on 4k level(well at least I rate myself near 3200-3500). Arteezy, being ridiculously better than me would just rape this MMR bracket with much greater winrate. And as for a player that has true MMR 1000 pts bigger(if I rated myself correctly), reaching it in 200 games is okay.

To dig further, my winrate with my signature hero is 67%, and I think anyone who wants to climb rapidly will pick his signature heroes for that case, further increasing their winrate.

A pic to prove and illustrate the things stated above:
Spoiler: Click to view


off topic

Dimonychan

Awards Showcase
Show more awards

Remarkable (43)
Posts: 1595
View My Blog
Permalink | Quote | PM | +Rep by Hamstertamer » October 27, 2014 6:49am | Report
MMR hell is not the point where you are surrounded with scrubs but you can't rise. This doesn't exist. If you're a good player but you're surrounded with scrubs you DO rise, but it's too slow to notice, and it's frustrating as hell. If you play 5 games a day you'll end up winning 25 MMR each day in average which is really little so even if you rise slowly it takes a million years to do so. MMR hell is when you rise so slowly and it takes you so much patience to deal with bad teammates that you get discouraged and give up. Which is something lots of people went through and is perfectly understandable.

Of course the 60% winrate thing is just an approximation to get an idea. Bur it's a realistic winrate of a good player who's playing way below his skill level.

I was making this post to counter the argument that "your MMR is where your skill level is". This is false, because your MMR takes like 1K - 2K games to actually get where your skill level is. And this gap is what people call MMR hell. Essentially I'm pretty annoyed when people reply something like "your MMR is where your skill is" to someone complaining about ELO hell because it's objectively not true.

Winrates with signature heroes are a completely different thing. I have 80-85% winrates with my most successful heroes IIRC Venomancer, Necrophos and Spectre, and Meepo not far behind (yes I was playing Spectre before it was cool ^^). Maybe I could farm MMR by picking them every game. But I don't think that would work, I think it's just the natural spread. Point is, the fact that you have extremely successful heroes comes mainly from randomness and not from skill.

To my knowledge nobody can reliably achieve high winrates. Just look at the profiles of professional players if you're not convinced. When you see Arteezy streaming a game where his whole team is feeding he's losing this game just like the next scrub. There are games you can't salvage no matter how hard you try. The only thing a professional player in scrub tier can do is salvage SOME 1 vs 5 feed fest games. No matter you look at it you're only 20% of your team's impact so my model is not that stupid.
Strategy guide : Anti-pubstomper guide.
Hero guides : Spectre , Windranger and Clinkz
== Broodmother guide out! ==

Hamstertamer
<Editor>

Awards Showcase
Show more awards

Memorable (89)
Posts: 2620
View My Blog
Permalink | Quote | PM | +Rep by Dimonychan » October 27, 2014 7:50am | Report
25 MMR each day(or rather each 5 games, that's actually the exact speed I'm rising with and I'm completely okay with that so far, no dem feelings), it's normal if your skill level doesn't exceed +1-1.5k MMR difference. If you make it faster, the whole point of RMM would be missed as it should match people as close as it can. And when I read Elo-hell complaints, people usually also mention "forced 50%" meaning the are most likely not rising at all, and if thats the case, thiir MMR is exactly as it should be. If they rise extremely slowly, like if they have 52-53% winrate, they are most likely somewhere near their level as there are less and less games player can influence enough to win.

So MMR system works, slowly, but works. How do you expect it to determine if you are pro player who deserves over 9k MMR or just a lucky scrub? If someone complains about his slow rise, he is either close to his final destination, or just expects too much from the system.

Quoted:
Point is, the fact that you have extremely successful heroes comes mainly from randomness and not from skill.

How so? Maybe we understand "signature heroes" term differently? While in the picture I provided Lich and Abaddon are by no means my signature heroes(though Aba is my 3rd most played hero actually), Kunkka surely is. How could I be lucky enough to achieve 67% winrate and 6.7 KDA over a course of almost 50 games? Signature heroes are that you perform well with constantly, due to experience with them and their style of play that suits you.

Examples of signature heroes:

Juice perform extremely well with Spirits and WK - while his stats are already good on most heroes he played, these 3 are just ridiculous.

EvilArthas just rapes people with WK(http://www.dotabuff.com/players/163075690) over 200 games - and he played much more, and used to have better results.

http://www.dotabuff.com/players/94338967 <-look at this guys and his SS - 120+ games are enough to exclude any luck.

So usage of signature heroes really makes difference.

Dimonychan

Awards Showcase
Show more awards

Remarkable (43)
Posts: 1595
View My Blog
Permalink | Quote | PM | +Rep by Timminatorr » October 27, 2014 8:28am | Report
i think valve should really try to make it so that you can gain points when losing.
atleast then you get properly rewarded for playing well.

Timminatorr
<Editor>

Awards Showcase
Show more awards

Memorable (57)
Posts: 2376
Permalink | Quote | PM | +Rep by KoDyAbAbA » October 27, 2014 8:33am | Report
^needed but very difficult to implement in real game. like, very difficult

KoDyAbAbA


Memorable (65)
Posts: 2366
Steam: Sir Rat-A-Lot
View My Blog
Permalink | Quote | PM | +Rep by Xyrus » October 27, 2014 8:37am | Report
Hamstertamer wrote:

MMR hell is not the point where you are surrounded with scrubs but you can't rise. This doesn't exist.

I calibrated at around 2.1k MMR, so far I've bothered to raise that to 2.25k. That was a few months ago now, I haven't bothered playing Solo Ranked again since. I've watched various streams, as well a few 4k+ players on my friedns list. I can tell you now, there are scrubs at every level of MMR.

...MMR Hell is everywhere! The only way to escape it is to stop playing...

Xyrus
<Moderator>

Awards Showcase
Show more awards

Established (104)
Posts: 2429
Steam: Xyrus
View My Blog
Permalink | Quote | PM | +Rep by CapnDaft » October 27, 2014 10:27am | Report
Dimonychan wrote:

So yeah, attitude is huge. Change it, and you'll see your MMR grow.

I was generalizing my past 5 games, not in general� I mean Im canadian, I'm kind with people


EDIT: Also not sure why you built a daedalus on magnus in your last match. The guy is a piss poor carry, but even if you were playing him as carry you should still get a refresher orb first. But I'm not even sure why they drafted magnus when you had an ES and a Cent.


Alright,
We had enough follow up, with the es and centaur, this is why I decided to get assaults since it'll help me and nayx a lot, since he infested me a lot in this game. Also, Magnus has mana problems, big ones, and since I wanted assault cuirass for my team, I didn't have a lot of regen, which means the combo with refresher would be just a pain, 'cause once again, we had the follow ups. I decided to get daedlus because of the attack speed that gave me AC, which I think would be way a way better combo since it did got me some kills. I was playing him more as a utility carry, like a necro would or someone in that kind.
Spoiler: Click to view

CapnDaft

Awards Showcase
Show more awards

Notable (8)
Posts: 368
Steam: Captain Daft
Permalink | Quote | PM | +Rep by DonnyDonnowitz » October 28, 2014 4:50am | Report
CapnDaft wrote:


Alright,
We had enough follow up, with the es and centaur, this is why I decided to get assaults since it'll help me and nayx a lot, since he infested me a lot in this game. Also, Magnus has mana problems, big ones, and since I wanted assault cuirass for my team, I didn't have a lot of regen, which means the combo with refresher would be just a pain, 'cause once again, we had the follow ups. I decided to get daedlus because of the attack speed that gave me AC, which I think would be way a way better combo since it did got me some kills. I was playing him more as a utility carry, like a necro would or someone in that kind.



Again it goes back to my original post, play to a heroes strengths and not their weaknesses. That wasn't the only game where you built an AC/Daed on Magnus. Also judging by the GPM/CS in some of your games I think you might need to work on your last hitting.

DonnyDonnowitz


Notable (3)
Posts: 46

DOTAFire is the place to find the perfect build guide to take your game to the next level. Learn how to play a new hero, or fine tune your favorite DotA hero’s build and strategy.

Copyright © 2019 DOTAFire | All Rights Reserved