|
Please keep the QQ to a minimum if you do not like this update. We are happy to hear your reasoning for not liking a ranked system, but no "OMG VOLVO WHY" posts. |
On December 26 2013 23:55 Kupon3ss wrote:Show nested quote +On December 26 2013 23:02 juracule wrote:On December 26 2013 08:03 cecek wrote:
If you're really stomping games hard, which you should be if you're playing games 1k points below your level, I think the smurf detection will kick in and make you progress much faster. I imagine if a 5k player started playing on a 4k account, he would be back at 5k in under 30 games, easily. Of course I don't have a souce for that. Basic math disproves that already. On average you lose or gain about 26-27 MMR points. 30x27 = 810 and this is with 100% winrate. So much for easily in under 30 games. Lets assume a 60% winrate (which is already p good seeing as the worst player matters more than the good player on a team). 60% of your 30 games is 18. 18-12 = 6. 6x27 = 162. Congrats after 30 games with 60% winrate you gained +162 MMR. you gain/lose 25 or so points when you have a high "certainty" due to the your average game being around 50% and the game believing that's where you belong. Once you actually start going on winning/losing streaks that "certainty" increases and you're able to gain/lose much more points as long as you're ACTUALLY not the rating you deserve.
??? I assumed a 60% winrate, that means you wont be streaking much in those games. Unless you contest the 60% winrate?
Also I think that you mean decreases.
|
I have won 9 of my calibration games and overall I have won 89 of my 152 games. Why I have only a rating of 3146?
|
Don't really know if I like this. Since the arrival of ranked, everybody I want to play with is like: let's play ranked, it MEANS something.
And I'm like: I had a hard day working my ass off and now just want to play a bit of good old doto. After 5 min or so I give up and we queue for ranked:
If we win: Yeah points!
If we lose: WTF XY OP! fuck this game, hey Blackfish why u feed?!!!!?????
Normally we just queued, sometimes lost sometimes won, didn't matter, just a game, just for fun. But it's always the same.
As soon as something is on the line, even if it's just some numbers between 1000-6000 everybody goes nuts. 
|
On December 27 2013 00:20 Skomski wrote:
I have won 9 of my calibration games and overall I have won 89 of my 152 games. Why I have only a rating of 3146?
I would assume you'd still have low certainty after so few games played so going up should be fairly easy. Also might be that you have stacked with friends before you were playing ranked which does not count for your solo rating (I think).
|
On December 27 2013 00:08 juracule wrote:Show nested quote +On December 26 2013 23:55 Kupon3ss wrote:On December 26 2013 23:02 juracule wrote:On December 26 2013 08:03 cecek wrote:
If you're really stomping games hard, which you should be if you're playing games 1k points below your level, I think the smurf detection will kick in and make you progress much faster. I imagine if a 5k player started playing on a 4k account, he would be back at 5k in under 30 games, easily. Of course I don't have a souce for that. Basic math disproves that already. On average you lose or gain about 26-27 MMR points. 30x27 = 810 and this is with 100% winrate. So much for easily in under 30 games. Lets assume a 60% winrate (which is already p good seeing as the worst player matters more than the good player on a team). 60% of your 30 games is 18. 18-12 = 6. 6x27 = 162. Congrats after 30 games with 60% winrate you gained +162 MMR. you gain/lose 25 or so points when you have a high "certainty" due to the your average game being around 50% and the game believing that's where you belong. Once you actually start going on winning/losing streaks that "certainty" increases and you're able to gain/lose much more points as long as you're ACTUALLY not the rating you deserve. ??? I assumed a 60% winrate, that means you wont be streaking much in those games. Unless you contest the 60% winrate? Also I think that you mean decreases.
I don't think winning or losing changes uncertainty much. From the blog
However, we do try to ensure that each team has a 50% chance of winning in any given match. (This is criteria #1 in the listed above.) We do not examine individual win / loss streaks or try to end them. However, if you are on a winning streak, in general your MMR is probably rising, which will tend to cause you to be matched with higher skilled opponents and teammates. Win rate is not a meaningful measure of player skill.
Win count is also not useful as indicator of skill, and the matchmaker does not use it for that purpose.
It seems only suprising match outcomes raises uncertainty, MMR rising=/= uncertainty rising. If valve constantly tries to match you a 50% chance of winning and ignore streaks then how can any match be suprising? Hopefully these games are often enough. I do see screenshots of them on reddit ( gain 44 points) but then the MMR gain goes straight back down to normal afterwards. Now if we can find out what exactly reactivates calibration... most likely something dependent on win rate anyway...
Anyway didn't valve wanted to find alternatives to the current MMR system because they're thought it was bad? No need to defend it to the death.
In other news I think I'm out of ELO hell (note I call it ELO hell cause the gaming experience is just shit, seems like the conventinoal meaning is something you can't get out of). Only took 24/7 for a few days of my life...
|
The biggest problem imo at the moment is, actually, the fact they enforce a 50% winrate. Normally you would want to find people who are all of equal skill rating and as a result get perfectly even teams so the chance of either team winning becomes 50%. Obviously this is not possible. So there are two things you can do; you can enforce the 50% winrate but by doing this you will handicap the best player in a particular game; he will get matched with people who have a lower average winrate/rating than that of the enemy team. Or you can make the teams randomized within a certain poule which might make games less uneven, but at least they are fair for everyone, instead of punishing the good players and coddling to the worst players.
Currently the better you get the harder you are gonna get dragged down (talking about solo queuing).
|
On December 26 2013 23:17 ChunderBoy wrote:Show nested quote +On December 26 2013 23:08 Stancel wrote:On December 26 2013 22:02 juracule wrote:On December 26 2013 00:44 Treefrog42 wrote:
My solo MMR is 3316 I think. Is that good? Apparently it's top 10%. I don't get this? Why give two cahoots about what your MMR is? Are you having fun, close games? Or are you having games where you feel like you're getting paired with 4 shitters? If you have the first why is it relevant what your MMR is? On December 25 2013 22:14 w.s wrote:
How surprised I am to see that the majority of people opposing the system are people low ranked that believe they are much better than their current rank and call it "elo hell". Why don't you just face it? The system works and you got placed exactly where you belong and if you weren't then winning games should be easy right since you're so much better? I was always on one of the first few pages pre ranked and that's the case now as well. How surprised I am that the majority of the people who support the system are people who have a high MMR that re-enforces their belief that they aren't as shit as they actually are. Because some people like an idea of how well they're doing or whatever with a visible rating and beesa pls stop deleting me from friendlist ;_; stop changing ur name, its like some ppl not only change their names but also their avatars. ive finished enough jigsaw puzzles when i was a kid my brain cant handle anymore. readd me
say you??? I almost deleted you if it wasnt for valve's tag function.
|
On December 26 2013 19:19 Andre wrote:Show nested quote +On December 26 2013 10:28 Kupon3ss wrote:On December 26 2013 09:54 Laserist wrote:
There is no way to climb up the MMR rating by playing 5th role supports. My supporting is better than my mid or carry but I lose a good percentage of my support games. Fails, feeds, blames etc.. 4K MMR is nowhere near what valve has stated. this has been proven false time and time again. Plenty of players have gotten to the top of ladders playing exclusively support. Even players who are not necessarily support players like fear have done it on great success. at the top of the ladder, individual mechanics actually play less of a role while understanding and game sense are even more important But it is EASIER to climb the ladder as a mid player, I would say. The amount of impact you can have in any given game is usually dictated by the role you play, supports can do a shitload(especially chen/ench players), but generally speaking as a mid player you're gonna have(at least statistically) more success in this regard. The point is that as a mid role, or maybe even carry/suicide to some extent massing games will reward you more mechanically, as a support you will generally require a lot of game sense, game knowledge, decision making - things that are quite hard to improve upon and which usually result only in minor victories that are important at a higher level of play. I think even 2009 said in one of his VODs that ladder system of any kind favors a mid role.
The reason why it is easier is that a mid player plays a hero that typically snowballs, is relevant in fights longer than a support, and has the opportunity to completely outplay the opposing mid player and straight up win. Although it is very difficult to outright lose and feed in middle, most people do not know how to play damage control nor do they know how to play from behind very well, so most of the time the mid player ends up feeding 2-3 kills and the game is simply over at that point (mid player that won is now like 4-5 levels ahead of everyone and just ganks the shit out of everyone else).
|
On December 26 2013 19:19 Andre wrote:
I think even 2009 said in one of his VODs that ladder system of any kind favors a mid role.
To be fair, 11 system uses KDA stats much more heavily than Valve does, which is part of why mid is a better role to raise rating there. To gain rating as a support there you basically have to do so well that you end up killing you way into outfarming your solos.
It's not as pronounced here, but most of the other factors that make mid better for raising rating do apply though.
|
United States12240 Posts
On December 27 2013 01:25 juracule wrote:
The biggest problem imo at the moment is, actually, the fact they enforce a 50% winrate. Normally you would want to find people who are all of equal skill rating and as a result get perfectly even teams so the chance of either team winning becomes 50%. Obviously this is not possible. So there are two things you can do; you can enforce the 50% winrate but by doing this you will handicap the best player in a particular game; he will get matched with people who have a lower average winrate/rating than that of the enemy team. Or you can make the teams randomized within a certain poule which might make games less uneven, but at least they are fair for everyone, instead of punishing the good players and coddling to the worst players.
Currently the better you get the harder you are gonna get dragged down (talking about solo queuing).
Everybody knows this is not true and it's been discussed to death in this thread already.
|
Judging the players in my games(not necessarily good or bad) and the information Valve gives, I am not sure if everybody agree on "this is not true". Not saying it forces, but not sure it doesn't either. I have too much skill variance in my games.
Sometimes I have a fantastic support, insta-counter gank other lanes, does fantastic harass and controls the early game well. A few games later, boom last pick sniper into hard lane. Maybe system couldn't measure players skill adequately. I mean you can be considerably better or worse than players within 300+ MMR span.
|
On December 27 2013 02:29 Laserist wrote:
Judging the players in my games(not necessarily good or bad) and the information Valve gives, I am not sure if everybody agree on "this is not true". Not saying it forces, but not sure it doesn't either. I have too much skill variance in my games.
Sometimes I have a fantastic support, insta-counter gank other lanes, does fantastic harass and controls the early game well. A few games later, boom last pick sniper into hard lane. Maybe system couldn't measure players skill adequately. I mean you can be considerably better or worse than players within 300+ MMR span.
??
My question is, what are you saying then? They sorta enforce a 50% winrate, but they sorta don't? Valve has stated previously they do not force a 50% winrate. How do you even go about "forcing" a win or loss? Say you put Dendi with 4 people who have never played dota in their lives versus 5 people who have never played dota in their lives. He could just as easily be dragged down by his teammates as he simply carries his team 1v5 every teamfight. Both lines of reasoning (I'm a good player I just get dragged down by my teammates when I lose/ I'm a good player I can carry my team easily given good support) have shown up multiple times in this thread. Not mentioning the fact that pub matches themselves are insanely hard to predict because the system doesn't even know what heroes the players are going to pick.
Skill variance doesn't really mean all that much anyhow, for various reasons players are going to exhibit a wide range of skills in any given match they play, even the same player. Some of those reasons include being pidgeon-holed into an unfamiliar role or hero, opponent lineup, or even something as random as their mood on that day.
EDIT: Regarding the notion that Valve sometimes forces a win or loss so players tend towards a 50% win-rate - forcing a win or loss is not the only way that a 50% win-rate can be achieved. You don't force a coin to land tails up if you get 10 heads in a row, you trust in probability and the law of big numbers that if you make another 500 flips, such blips become irrelevant. I'm pretty sure that's how Valve's system works too, by improving the algorithms and estimation of player skills matches become more even which naturally causes winrates to tend towards 50%.
|
On December 27 2013 01:58 Excalibur_Z wrote:Show nested quote +On December 27 2013 01:25 juracule wrote:
The biggest problem imo at the moment is, actually, the fact they enforce a 50% winrate. Normally you would want to find people who are all of equal skill rating and as a result get perfectly even teams so the chance of either team winning becomes 50%. Obviously this is not possible. So there are two things you can do; you can enforce the 50% winrate but by doing this you will handicap the best player in a particular game; he will get matched with people who have a lower average winrate/rating than that of the enemy team. Or you can make the teams randomized within a certain poule which might make games less uneven, but at least they are fair for everyone, instead of punishing the good players and coddling to the worst players.
Currently the better you get the harder you are gonna get dragged down (talking about solo queuing). Everybody knows this is not true and it's been discussed to death in this thread already.
Elaborate on that what is not true? Valve has said themselves they enforce a 50% winrate for either team. Everything what on wrote on that topic is just logical reasoning going from there.
EDIT:
You're confusing personal winrate with the winrate of a team. From valve's blog post:
"The teams are balanced. (Each team has a 50% chance to win.)"
|
On December 27 2013 02:40 sgfightmaster wrote:Show nested quote +On December 27 2013 02:29 Laserist wrote:
Judging the players in my games(not necessarily good or bad) and the information Valve gives, I am not sure if everybody agree on "this is not true". Not saying it forces, but not sure it doesn't either. I have too much skill variance in my games.
Sometimes I have a fantastic support, insta-counter gank other lanes, does fantastic harass and controls the early game well. A few games later, boom last pick sniper into hard lane. Maybe system couldn't measure players skill adequately. I mean you can be considerably better or worse than players within 300+ MMR span. ?? My question is, what are you saying then? They sorta enforce a 50% winrate, but they sorta don't? Valve has stated previously they do not force a 50% winrate. How do you even go about "forcing" a win or loss? Say you put Dendi with 4 people who have never played dota in their lives versus 5 people who have never played dota in their lives. He could just as easily be dragged down by his teammates as he simply carries his team 1v5 every teamfight. Both lines of reasoning (I'm a good player I just get dragged down by my teammates when I lose/ I'm a good player I can carry my team easily given good support) have shown up multiple times in this thread. Not mentioning the fact that pub matches themselves are insanely hard to predict because the system doesn't even know what heroes the players are going to pick. Skill variance doesn't really mean all that much anyhow, for various reasons players are going to exhibit a wide range of skills in any given match they play, even the same player. Some of those reasons include being pidgeon-holed into an unfamiliar role or hero, opponent lineup, or even something as random as their mood on that day.
sure people skills arent always the same but there are certain basics you can rely on at any time. like not skippingm eld on ta or rushing 2 naked bfs with any carry hero. laserist pretty much nails whats bugging me...sometimes the experience overall is good (neglecting obvious trolls etc) teammates seem to get whats up but there are too many games were its not like that
|
Initially I am trying to say, everybody doesn't agree on %50 enforcement is a lie. We don't know how MMR is calculated exactly. We don't know which exact parameters they use. What we know is, win/loss is a key parameter and some other parameters are also used.
Everyone also expects to be matched(either ally or opponent) with better opponents when he/she feels improved in time. When you win more than usual, you expect to be matched with better players, better game experience right?
It is frustrating to know after a 3-4 match win streak, you'll be matched with clueless players and stomp/stomped. This is the perception of %50 wr enforcement. An improved, maybe imaginary, MMR system should avoid this more than what currently is.
|
United States12240 Posts
On December 27 2013 02:47 juracule wrote:Show nested quote +On December 27 2013 01:58 Excalibur_Z wrote:On December 27 2013 01:25 juracule wrote:
The biggest problem imo at the moment is, actually, the fact they enforce a 50% winrate. Normally you would want to find people who are all of equal skill rating and as a result get perfectly even teams so the chance of either team winning becomes 50%. Obviously this is not possible. So there are two things you can do; you can enforce the 50% winrate but by doing this you will handicap the best player in a particular game; he will get matched with people who have a lower average winrate/rating than that of the enemy team. Or you can make the teams randomized within a certain poule which might make games less uneven, but at least they are fair for everyone, instead of punishing the good players and coddling to the worst players.
Currently the better you get the harder you are gonna get dragged down (talking about solo queuing). Everybody knows this is not true and it's been discussed to death in this thread already. Elaborate on that what is not true? Valve has said themselves they enforce a 50% winrate for either team. Everything what on wrote on that topic is just logical reasoning going from there. EDIT: You're confusing personal winrate with the winrate of a team. From valve's blog post: "The teams are balanced. (Each team has a 50% chance to win.)"
"Enforce" or "force" is the key functional word. "Forcing a 50% win rate" means that you won your last game, now you're going to lose your next game so you stay at 50%. People rationalize this by thinking they are grouped with teammates with much lower MMRs in order to influence the outcome of the game. In fact, you said this yourself. That is not the case. If you read through their matchmaking post, it's extremely clear.
They target as small a skill variance as possible. The one time you may have a wider variance is when player stacks are involved. I have friends who are better than me and friends that are worse, and sometimes to no small degree. Pretty much the only time you're going to have players with a significant MMR difference in the same game is if they queued up as a group that way. That doesn't mean the outcome is predetermined though, because statistically the averages for both teams will make a competitive match. Also, in case you haven't noticed, you have no control over this anymore (the option to queue only against other solo-queuers was removed). Take a look at the example game that Valve provided so you can get an idea.
|
On December 27 2013 02:54 Laserist wrote:
Initially I am trying to say, everybody doesn't agree on %50 enforcement is a lie. We don't know how MMR is calculated exactly. We don't know which exact parameters they use. What we know is, win/loss is a key parameter and some other parameters are also used.
Everyone also expects to be matched(either ally or opponent) with better opponents when he/she feels improved in time. When you win more than usual, you expect to be matched with better players, better game experience right?
It is frustrating to know after a 3-4 match win streak, you'll be matched with clueless players and stomp/stomped. This is the perception of %50 wr enforcement. An improved, maybe imaginary, MMR system should avoid this more than what currently is.
It's equally likely that you're below the average MMR of your games. There is no 50% win ratio enforcement because valve doesn't know if your current MMR is the one you deserve to be at. It will keep matching you at whatever MMR you're at (which will naturally put you closer and closer to 50% assuming you've hit the MMR that matches your skill and don't improve). Sometimes you get unlucky and your teammates screw up, sometimes you get lucky and your teammates carry you. I have a feeling that people are simply more perceptive of teammates' mistakes when they lose so they come up with shit like "valve teams me up with retards to drag me back down after I win a lot."
|
Tbh I think Valve should group games into chunks, of say 20 games, and then move your rating when you've completed those 20 games. Because this thread makes it obvious no one understands that the rating is only accurate based on a pretty large sample, and things that happen in individual games hardly matter. I could go play 5 games right now where my mid feeds constantly, and it wouldn't really bother me because I know in the long run the enemy mid will feed just as much. Some psychologist would have a field day with Dota players I feel.
|
Yeah, I don't care "look at noob xxx" attitude of players since most of the time they try to cover their own mess.
Only things disturbs me the skill variety between the games and players. It is beyond the explanation of bad day/poor performance/unfavored heroes etc.. It is like riding a bike, you should have some basics at a certain MMR level. In 4K range, it is very easy to see some guys perform super good and others are fail and matched around the same level. I put myself out of the topic.
|
My personal problem with MMR is that it doesn't suit my particular play style. I hate playing high-risk, high-reward heroes. I like to play hero's that I can always help my team out consistently and not heavily impact anything in my game whether it be for good or for bad. I'm not a play-maker, and I don't like to lead or make plays or play aggressive by myself, and I don't like to "pub stomp" with a shadow blade and stupid shit like that... I find that stuff retarded, I want to try-hard and play like a professional team. I'm always in position to help, map is always warded/de-warded, item choices are always to help my team/give the other team headaches, etc. The problem is that I don't like to fail or be the reason why my team loses, and so then it follows that I also cannot be the reason they win either. So even though I am a very good technical player, I will never be anywhere near the top of MMR because of the way I like to play the game. I always like to play smart, low-risk DOTA. The way MMR is designed it will never be able to accurately place me due to this fact. I believe it assumes the kind of DOTA I play is wrong DOTA, so since I play like that I must be bad... But that just simply isn't true.
|
|
|
|
|
|
EPT
