Yes you are right as far as i am concerned. I lost 26, 24 won 25, 24.. about 24-26 +-. Even when the game is unfair and I performed best still lose 25...
if you are going 50% you are exactly at the MM you should be at................Do people think if they are a 3k rated player and win 50% of there matches vs 3k rated players that they should be rising? There is no forced 50% people will just plateau eventually unless they become better players
We are talking about solo ranked here. Realistically a solo player would get 50-55%. The thing is this current system gives you slow progress. If you win 55% of your solo games and lets say you played 100 games your mmr will only increase by a few. At least valve should include player performance
in ranked the rating will be +-25 if both teams have the same average MMR. If the game was hard it was due to factors not related to the MMR of either team, e.g. someone on your team underperforming by picking a hero he is bad at or someone on the opponent team playing his best hero. Getting +40 means the opponent team had a higher average rating and +10 means your team had a higher average rating.
Getting consistent 25s simply means that at your MMR, the player pool is large enough such that both teams are supposed to be evenly matched as far as MMR is concerned.
If you can win 55%, every 100 games you will be up +10 games or about 250 MMR.
If you can win 60%, every 100 games you will be up +20 games or about 500 MMR.
If you can win 70%, every 100 you will be up +40 games or about 1000 MMR.
If you don't cause a your team to be significantly more likely to win why should MMR go up? But also I don't know why people would be desperate to get their MMR to the max they can achieve. If you do that then you will be stuck playing only your best heroes or losing. I like playing lots of heroes and its nice to still be able to win some games whatever I pick. If I want to play a higher level game then I stack. Solo que is just for fun, why be desperate your max out your MMR?
http://dotabuff.com/players/103176226/matches?hero=zeus&game_mode=&match_type=real
Zues is at 78% win and 5.20 KDA on 31 games... In fact I'm 6-0 on solo que Ranked Zues games. I could play Zues every game and probably gain 500 MMR points. But then I would be stuck playing Zues or a handful of other heroes or losing. I can't play all the heroes anywhere near as well as Zues. I'm really best at solo mid heroes. If I wanted to max out MMR I would do that all the time. But I like playing support heroes better. It's more relaxing, less stressful.
@Wooolfgang
you see people around 50-55% at the point that they have already plateaued. If Dendi or other pro players played at a 2k MM they would not have a 50% winrate
^this
Your winrate should be higher until you reached your skill level. When your MMR reaches your skill level it should be 50%. Thats how the system works in theory.
No, you won't plateau at your supposed MMR because you can win a game on your teams back and I'm not talking about getting carried as a support. I mean some people will play horribly and do nothing but feed as a carry but will still gett MMR if their team simply deals with them and wins.
MMR is team dependant, ergo it's kind of useless to compare individual skill.
^ You have to win a game "on your teams back" in order to have a high MMR, you will still plateau with it once you reach your MMR (where enemies just play as good as you).
so what is this plateau stuff? Are you saying only 60% & 70% still haven't plateau yet?
What Sampson is saying only appllies for solo queue ranked, and if everyone is solo queueing. I would say that it only appllies if you are actually trying.
It will only plateau if you are playing a lot of solo queue ranked games.
@ NECROPHE[A]VY
in solo que you will not rise much past 50-55% (cept for god tier players) once you plateau. Group que is where the variation in winrates tend to come from. So to answer your question yes if they have a 70% winrate (solo) they have not reached their plateau.
@Dorkly
obviously If I join each game and go afk and lose over and over and end up with a 1k MM it does not mean I became a worse player.
I agree with the plateau part. What I'm trying to say is that isn't it a bit slow to increase your ranked mm with this current system? And it kinda inhibits the "trying to get better" attitude. Wouldn't it be better if for example you played well but you lost the game your mmr would still decrease but it would be less than (the normal is -25 so you get -20 or -15 ). It kinda sucks when you have a good game and you played really well and you just get -25. The same with winning the game but feeding hard. You know you don't deserve the +25. In solo ranked, you could only do so much because tbh some people will bring you down.
Just one person will actually make you lose the game
*This axe in this game intentionally fed the other team http://dotabuff.com/matches/464519658
*After getting 9-1 early with ES (kinda imba hero so who cares) and giving space this alche built basher first and could'nt even farm shit http://dotabuff.com/matches/464441679
I just think that rating is really dependent on the other 4 randoms. So why not give due credit to the people who played well? And give a little punishment to those who didn't? With this people would actually try to play better. The point is that what you get isn't static (+25,-25) and you could actually review and re-assess on how you played. I'd say giving (+10,-10) isn't that much of a difference right?
TLDR: Getting +25, -25 sucks because it doesn't actually measure how well you played; ranked solo is hard with 4 other randoms; It is slow to increase mmr with this system
The fact is people don't get better quickly. 100 games for an incremental improvement from maybe 2500 MMR to 2750 MMR makes sense. If you let it change more quickly you just rush people into games they are not ready to handle. Expectations of players at different levels are very different. Also winning 4 o 5 games in a row is just luck. Getting a high KDA in a certain game really doesn't mean anything. Its better if the movement is slower.
The real problem with their system is it matches a huge range of MMR into the same game. It is a consequence of stacks having very large ranges of MMR... but it still should not happen if people do not want it. I should be allowed to set a maximum range on MMR that they can put on my team and against me. If that means a match can't be found for 10 minutes because I wanted only +/- 200 MMR... well tough for me. If I don't like it I can change my setting. Valve just forces everyone accept the match they get and won't even show what they are. If they showed the MMRs and how far apart they are. People would be angry.
+10/-10 isn't that huge of a difference imo. And moving from 2500 to 2750 isn't that big of an improvement really. Games will still feel the same. About +/- 500 mmr players will feel the difference in games and skill level but 200 difference, I doubt it. That's about 2 months of playing time for me. And it certainly is a slow progress for 100 games.
Even adding only +5/-5 is ok with me as long as there is something that you could look up to after playing a game. Seriously sucks to see that stale +25/-25 after every game. It is more rewarding that way
a system like that just does not work. Having you teammates worrying about their own stats rather than winning the game only leads to bad decision making.
an example of this might be a WR going a force staff when your team needs a orchid/sheep stick to deal with the earth spirit. The WR player just force staffs and windruns away and says 'gg team feed' AND gets rewarded for doing this by losing less points
And also, if you are the best player on your team, its expected that you play better than your teammates, AND it would be your job to carry your team to victory, so if you lose and your stats are slightly better than your team, why would you deserve a reward, when its still your fail as much as theirs? Those games when you were mid and had 20-4-10 but your team had 0-13-4 all of them and you ganked their lanes and helped them as much as you can happen but as long as its one of 100 games, its fine, there will be a one of 100 games when your opposite has no chance. Most loses means that with a better draft or if everyone played a hero they can, it would still be winnable.
Progress is not really slow, IF you belong to a lot higher tier. If you should be 1000 above, you will go up 500-600 fast, then slowly get to your max possible rating if you play tryhard mode and do your best every game. If you only belong 200 higher, why would you need a faster progress? Ive seen ppl going up 800 in weeks, so its not slow. Also, those bad games mostly happen, couse someone is a lot lower on mmr than he should be, so he can single-handedly win the game. But it changes fast, and you cant do anything against it, theres no real solution.
@Sampson
That's unless you create the system properly.
It's possible to create an algorithm that looks at your items and grades your choices on a skill range. The more your item build looks like a pro build ( in any given situation - taking in account your opponents picks, your team picks and your opponents items ) - the more points you get for it. It can be used to calculate your skill rating.
Elo is a flawed metric because it needs to be zero-sum. It will continue sucking until Valve ditches it for a non-zero sum system.
It would let people get points even when they lost but played good. It's the ultimate holy grail of matchmaking and could make this game 3 times as fun :)
@Vaeldiithia
While it may be true that you can get 500-600 rating quickly, it doesn't mean the system is good. It might have been good for you to win those 20-25 games in a row but the matchmaking algorithm ruined the fun to all of your opponents in those games. It failed to recognize a high skilled player is owning. So basically, because of this one player, 20-25 games turned to shit.
20 x 9 people x 30 minutes per game = 90 hours of poor games caused by matchmaking flaw - and it's just for ONE miscalibrated player
Nah, I cant win I only improved like 200 from my original, but its couse I belong there. Ive seen others on my list going up that much, and yes, it ruined the game for enemies, but what other solotuion you can come up with? How can you place those players to their real mm faster? When you play casually, you get a rating and its a lot below your try-hard mmr. What can a system do with ppl just switching into tryhard mode once? Its not the systems fault, its the players attitude change. If those players rating would change more rapidly, then it will go too high and ruin the games for the teammates, so you just CANT solve this. Its not the systems or valves fault, you cant do anything with it.
Edit: I just sent this before reading your answer to Sampson. Pro builds dont work better in pubs than pub builds. Sometimes troll builds win you games couse your oppontent wont think of that, sometimes troll builds lost you the games couse even thou you are fed, you still dont have the neccessary late-game survavibility/damage. But in other games, the troll build wins you the game in mid-game... How can a system decide that? Any other system is abusable, or if its not, ppl will still try, play for their stats, build the most-points items, and ruin games by that.
@Vaeldiithia
It's not black and white - there's a spectrum of possible solutions that could alleviate the whole problem more or less.
Tryharding or not tryharding - for sure doesn't account for 100% of imbalanced games - just a fraction of them. It doesn't mean that the other causes can't be eliminated.
Additionally, when it comes to tryharding, you can create a model that predicts WHEN you will tryhard. Sounds like black magic and mind reading BUT IT IS ABSOLUTELY POSSIBLE to train a computer to recognize if a player will tryhard this game. You could solve this problem partially by introducing such solution.
On the spot I can think of several factors that could be used for this (together with every other factor):
- The Win/Loss pattern of your last games.
- How much you spent online so far.
- The day of the week ( duh - in one of the day of the week you can be happy because it's for example friday and you are always happy in fridays and play better because of this during this day ). Or maybe you hate mondays and to fix your mood you tryhard with your best hero to win some games to feel better.
- When did you play your tryhard heroes last.
- During match finding phase - extracting features out of your teammates and opponents ( nationalities, flaming rating, trolliness etc. that can be extracted with different models ) then use this data to predict whether you will be forced to tryhard ( by badmouthing opponents or by seeing cyryllic in the allies chat ).
Matter of fact, given access to this data, I could create a prototype for this in 2-3 days. As of now, Valve's solution is a pure Elo system.
Im studying computer learning. I know its good.
But this game is situational, this game is good couse everything is situational. You can do perfectly on a game and throw it by one big mistake that was truly yours, so you deserve a cookie or lose more rating than your team for that?
Also, if you have any kind of AI giving points, others WILL make their models and put it on the internet. They will NOT be the same, but people will try to use it to abuse. May or may not be successful, they will try and ruin with it. Either getting too high advantage if its good or throwing games if its bad. So it means lot more work for nothing.
Anything is abusable and players will try...
I tend to agree that if you are much better than the other 9 players currently in your game then you should be easily able to snag another 25 points and get to a higher matched game. This sometimes is actually impossible (not that I'm complaining).
I had a game around 10 games back or so playing as an offlane clockwerk. I think I had got around 7 - 1 - 2 by minute 10 and completely shut down the lich and PA I was against.
I was originally going to go mid but some guy first picked necro and was adamant he was going mid. He gave up first blood as soon as the creep waves met in mid lane and I kind of knew it was going to happen because before first blood was announced I heard "CYKA CYKA BLYAT" seconds before over the in game mic.
Mid steamrolled over our mid and basically there was nothing we could really do as they 5 man doto'd all day all night.
Its hard sometimes when you depend on other random people to win a lane and not get totally owned. Ah well!
King Goose, I believe here is an example of such game, based on your account:
http://dotabuff.com/matches/447329298
When you do half of the teams damage and two carries go:
2:9
1:8
There's not much more you can do. While a pro player would be able to win this game, it doesn't mean that matchmaking is good.
You should get increase in matchmaking rating while meepo and LC should lose points.
Right now, for that match, you've lost as many points as Legion and Meepo - which is pretty flawed to me. Given the bad conditions you had to play in, you should gain rating.
Please sign in to post comments.
So ranked basically only factors your winrate? Since every win and every lose I get +25/-25 (based on my matches/ i might be wrong).
I think it kinda sucks since playing solo ranked you basically get 50-55% winrate. I hope valve puts other factors as well ie. hero performance