Website Feedback
Closed Threads
IRC Web ChatTeamSpeak 3 (52 users)
Active: 5694 users | |
|
| DodgySmalls Canada. June 15 2012 15:51. Posts 102 | Profile # |
On June 15 2012 15:20 skipgamer wrote:
Keep it hidden, I haven't seen any compelling arguments for why it should be shown in this thread.
Most of the complaints are from people you can tell have an elitist attitude, I'd rather them go (back) to HoN/LoL anyway.
And if you can't tell if you have improved by simply judging how well you play and watching replays, you're doing it wrong, visible MMR isn't going to help you.
oh, also, if you completely don't know where you stand, just filter your recent matches by high/medium/low skill... at least that will give you a rough indication of what bracket you're in. That should be enough to drive you forward if you're not in high skill games.
I don't appreciate being called an elitist.
"just filter your recent matches by high/medium/low skill"
How do I do this?
Edit: As a sub note, I would like to add that this isn't a scientific discovery where there is a pre-verified solution to a problem in existence (where the burden of proof is on the new theory). In this case there is a dispute between two methods which are used by different video games.
By the same logic I could say I've never heard a compelling argument why it shouldn't be shown. Having a thick enough skin to realize that some idiot who trash talks you on the internet is irrelevant to your self worth is something I find easy.
Does that make me an elitist?Last edit: 2012-06-15 15:55:03 |
| | Please remove nyx assassin | |
|
|
| skipgamer Australia. June 15 2012 16:35. Posts 560 | Profile Blog # |
how to filter by skill level
I neither have the time nor inclination to argue about whether you are elitist... And yes, by the same logic you could, and I would not argue with it, just sharing my opinion, and no it does not.Last edit: 2012-06-15 16:35:39 |
|
|
| vanTuni June 15 2012 19:40. Posts 301 | Profile # |
To the people that responded to my original post:
I agree, that premades cause an issue when you want a match making sytstem that does not balance teams. We need to separate two issues here.
1) Lets have an example: if 1000 were the starting mmr for everybody, one way to group 100 random players would be to make ten games with an average of 1000 MMR on each side. But the better option would be to make ten games where one has an average MMR of say 1500, one with 1400, ..., one with 500. This can be achieved by putting a cap on MMR-Range in a single game.
Now 2) for the problem of premades. This is where the points you earn come into play. You only earn significant MMR rating point if you win against people on the same level as you or above. I proposed that the MMR of premades should always be the one of the player with the lowest MMR for match making purpose. You were suggesting that people will find groups of 4 with high and 1 with low MMR to get easy games. Well the answer to that lies in the above rule. Another example: If 4 people with 1500 MMR find a person with only 800, and they get queued against a group of 800s then the people with 1500 would get next to 0 MMR gains for wins. I would additionally propose that the 1 person witht he low MMR should also get significantly lower point for being "helped" by higher ratings. Now exactly because MMR is visible, having no gain at all will keep people from just stomping lower ranking players for easy wins.
To put it in a nutshell, have a very strict maximum diversity of MMR allowed into a game when people are solo queuing. When people queue in pre-made groups the lowest MMR of the group counts for all of them for match-maing purpose. The MMR gain for defeating people with lower MMR then your own is marginal at best. MMR gain is calculated individually. make it all visible, to prevent stomping via incentive not punishment. |
|
|
| SKC Brazil. June 15 2012 20:22. Posts 3433 | Profile # |
On June 15 2012 19:40 vanTuni wrote:
To the people that responded to my original post:
I agree, that premades cause an issue when you want a match making sytstem that does not balance teams. We need to separate two issues here.
1) Lets have an example: if 1000 were the starting mmr for everybody, one way to group 100 random players would be to make ten games with an average of 1000 MMR on each side. But the better option would be to make ten games where one has an average MMR of say 1500, one with 1400, ..., one with 500. This can be achieved by putting a cap on MMR-Range in a single game.
Now 2) for the problem of premades. This is where the points you earn come into play. You only earn significant MMR rating point if you win against people on the same level as you or above. I proposed that the MMR of premades should always be the one of the player with the lowest MMR for match making purpose. You were suggesting that people will find groups of 4 with high and 1 with low MMR to get easy games. Well the answer to that lies in the above rule. Another example: If 4 people with 1500 MMR find a person with only 800, and they get queued against a group of 800s then the people with 1500 would get next to 0 MMR gains for wins. I would additionally propose that the 1 person witht he low MMR should also get significantly lower point for being "helped" by higher ratings. Now exactly because MMR is visible, having no gain at all will keep people from just stomping lower ranking players for easy wins.
To put it in a nutshell, have a very strict maximum diversity of MMR allowed into a game when people are solo queuing. When people queue in pre-made groups the lowest MMR of the group counts for all of them for match-maing purpose. The MMR gain for defeating people with lower MMR then your own is marginal at best. MMR gain is calculated individually. make it all visible, to prevent stomping via incentive not punishment.
That would assume people are playing for the sake of raising their MMR, when most people playing AP pubs, specially with low level teammates, are probally doing it for fun. Every single game like that would be horribly unbalanced, useless for MMR purposes and not fun at all. People that stomp low level players, like smurfs, don't do it for the points. They do it because they like stomping low level players. Your system not only makes that easier, it would make queing by your self even worse, since queing with other people gives you a huge advantage. I don't see how that would work. You can't say it's fine since that game won't count for points because it's not a competitive ladder, points are not your goal and shouldn't matter at all. Pubs are not meant to be competitive.Last edit: 2012-06-15 20:24:17 |
|
|
| BurningSera United Kingdom. June 15 2012 22:21. Posts 5279 | Profile Blog # |
the biggest problem of showing mmr or not is due to there is almost no credibility in mmr. People can just use the exact same fcking heroes for 10k games and whoring wins with mates or abusing some dual lane combo to get high k/d/a or mmr but still dont have clue about the game. I hate those players so much because they are usually the worst kind of team mate you can get. I had some very unpleasant experience from 3 months of HoN and i really think that is the real reason killing HoN.
performance bar and battlepoint are enough for epeen stroking imo. i don't want any kind of approvals from the game. Just play well you will meet some good players and these people will be the ones confirm that you are a good player, if you really needed that kind of approval. |
| | Spawn more Overlords - Zerg player at heart. 2009, 820, Yaphets, YamateH <3 | |
|
|
| Count9 China. June 15 2012 22:40. Posts 3699 | Profile Blog # |
On June 15 2012 16:35 skipgamer wrote:how to filter by skill levelI neither have the time nor inclination to argue about whether you are elitist... And yes, by the same logic you could, and I would not argue with it, just sharing my opinion, and no it does not.
Doesn't work at all, unless every game I've ever played (including what appears to be one of my first 5 games) are all high level lol |
| |
|
| epoc Finland. June 15 2012 22:59. Posts 841 | Profile # |
| Stats system made Hon's community pretty bad. |
|
|
| paralleluniverse Australia. June 15 2012 23:41. Posts 2972 | Profile # |
On June 15 2012 22:21 BurningSera wrote:
the biggest problem of showing mmr or not is due to there is almost no credibility in mmr. People can just use the exact same fcking heroes for 10k games and whoring wins with mates or abusing some dual lane combo to get high k/d/a or mmr but still dont have clue about the game. I hate those players so much because they are usually the worst kind of team mate you can get. I had some very unpleasant experience from 3 months of HoN and i really think that is the real reason killing HoN.
performance bar and battlepoint are enough for epeen stroking imo. i don't want any kind of approvals from the game. Just play well you will meet some good players and these people will be the ones confirm that you are a good player, if you really needed that kind of approval.
You have no idea how MMR works, so you should stop talking.
MMR has nothing to do with K/D/A, it's simply got to do with winning. If you use the same hero for 10k games, abusing some duel lane combo and you're winning a lot then you deserve high MMR.
If you only know how to cannon rush in SC2, and you can use cannon rush to beat 95% of all players by cannon rushing in.every.single.game, then guess what -- you're better than 95% of all players, and deserve nothing less than an MMR that's in the top 5%.Last edit: 2012-06-16 00:36:21 |
| | Check out "Shadow of the Eternals" on Kickstarter |
|
|
| paralleluniverse Australia. June 16 2012 02:22. Posts 2972 | Profile # |
On June 15 2012 15:20 skipgamer wrote:
Keep it hidden, I haven't seen any compelling arguments for why it should be shown in this thread.
Most of the complaints are from people you can tell have an elitist attitude, I'd rather them go (back) to HoN/LoL anyway.
And if you can't tell if you have improved by simply judging how well you play and watching replays, you're doing it wrong, visible MMR isn't going to help you.
oh, also, if you completely don't know where you stand, just filter your recent matches by high/medium/low skill... at least that will give you a rough indication of what bracket you're in. That should be enough to drive you forward if you're not in high skill games.
You are being completely ridiculous. It is virtually impossible to judge how good someone is based on watching their replays, because how good someone isn't absolute, it's relative. To know how good someone is you need to first know how good EVERYONE is.
2039287. That's a random game ID. Since these players have been matched together, we can assume they all have approximately equal skill. Now tell me how good is this group of players. Top 5%? Top 10%? Top 15%? How is it possible to know?
How good am I? Am I in the top 20% of players? The bottom 20%? The top 50%? The bottom 50%? The 34.7th percentile? The truth is I honestly have no fucking idea? Not even the slightest clue, not even within a 50% wide ballpark.Last edit: 2012-06-16 02:28:27 |
| | Check out "Shadow of the Eternals" on Kickstarter |
|
|
| SKC Brazil. June 16 2012 02:38. Posts 3433 | Profile # |
On June 16 2012 02:22 paralleluniverse wrote:
Show nested quote +On June 15 2012 15:20 skipgamer wrote:
Keep it hidden, I haven't seen any compelling arguments for why it should be shown in this thread.
Most of the complaints are from people you can tell have an elitist attitude, I'd rather them go (back) to HoN/LoL anyway.
And if you can't tell if you have improved by simply judging how well you play and watching replays, you're doing it wrong, visible MMR isn't going to help you.
oh, also, if you completely don't know where you stand, just filter your recent matches by high/medium/low skill... at least that will give you a rough indication of what bracket you're in. That should be enough to drive you forward if you're not in high skill games.
You are being completely ridiculous. It is virtually impossible to judge how good someone is based on watching their replays, because how good someone isn't absolute, it's relative. To know how good someone is you need to first know how good EVERYONE is. 2039287. That's a random game ID. Since these players have been matched together, we can assume they all have approximately equal skill. Now tell me how good is this group of players. Top 5%? Top 10%? Top 15%? How is it possible to know?
How is that relevant though? He said "if you can't tell if you have improved by seeing how well you play or watching replays". Give an example with an old replay and a recent replay of a player and it is reasonable to expect being able to see if they improve. Being in the "top 5%" is a completelly diferent subject. You can improve your skill level but drop down on an overall ranking if the other players improve faster or simply if the worst players stop playing, like what happens in the SC2 leagues.
His post is adressing the issue that an MMR is important for your own self evaluation. He says there are other ways to do that. You are concerned about how well you play compared to the other players. For me that is far less important. If you are close enough to the top for it to matter, you can check the watch games tab.
On June 15 2012 23:41 paralleluniverse wrote:
Show nested quote +On June 15 2012 22:21 BurningSera wrote:
the biggest problem of showing mmr or not is due to there is almost no credibility in mmr. People can just use the exact same fcking heroes for 10k games and whoring wins with mates or abusing some dual lane combo to get high k/d/a or mmr but still dont have clue about the game. I hate those players so much because they are usually the worst kind of team mate you can get. I had some very unpleasant experience from 3 months of HoN and i really think that is the real reason killing HoN.
performance bar and battlepoint are enough for epeen stroking imo. i don't want any kind of approvals from the game. Just play well you will meet some good players and these people will be the ones confirm that you are a good player, if you really needed that kind of approval.
You have no idea how MMR works, so you should stop talking. MMR has nothing to do with K/D/A, it's simply got to do with winning. If you use the same hero for 10k games, abusing some duel lane combo and you're winning a lot then you deserve high MMR. If you only know how to cannon rush in SC2, and you can use cannon rush to beat 95% of all players by cannon rushing in.every.single.game, then guess what -- you're better than 95% of all players, and deserve nothing less than an MMR that's in the top 5%.
The issue is complaining about is the fact that people believe MMR correlates to skill, when that isn't exactly true. People below your MMR can be better than you and people above your MMR can be worse then you, it all depends on things like which kind of heroes you play, if you are often carried or carry friends, how hard you "try hard", etc. People act like morons often because they think MMR is a definitive measure of skill, when it is not.
About your example, I would say a guy that plays random and beat 90% of the players is a overall better Starcraft player than a guy that cannon rushes only and beats 95% of the players. That guy is definatelly better at cannon rushing though, which is the only meaningful thing you could conclude from that information. It's fine cannon rushing every single game, but don't believe you are better than someone just because of a stupid ladder system, specially on a non-competitive enviroment. SingSing has a better MMR than Dendi, but I think very few people would argue that he is a better player simply because of that.Last edit: 2012-06-16 02:47:56 |
|
|
| paralleluniverse Australia. June 16 2012 02:48. Posts 2972 | Profile # |
On June 16 2012 02:38 SKC wrote:
Show nested quote +On June 16 2012 02:22 paralleluniverse wrote:
On June 15 2012 15:20 skipgamer wrote:
Keep it hidden, I haven't seen any compelling arguments for why it should be shown in this thread.
Most of the complaints are from people you can tell have an elitist attitude, I'd rather them go (back) to HoN/LoL anyway.
And if you can't tell if you have improved by simply judging how well you play and watching replays, you're doing it wrong, visible MMR isn't going to help you.
oh, also, if you completely don't know where you stand, just filter your recent matches by high/medium/low skill... at least that will give you a rough indication of what bracket you're in. That should be enough to drive you forward if you're not in high skill games.
You are being completely ridiculous. It is virtually impossible to judge how good someone is based on watching their replays, because how good someone isn't absolute, it's relative. To know how good someone is you need to first know how good EVERYONE is. 2039287. That's a random game ID. Since these players have been matched together, we can assume they all have approximately equal skill. Now tell me how good is this group of players. Top 5%? Top 10%? Top 15%? How is it possible to know?
How is that relevant though? He said "if you can't tell if you have improved by seeing how well you play or watching replays". Give an example with an old replay and a recent replay of a player and it is reasonable to expect being able to see if they improve. Being in the "top 5%" is a completelly diferent subject. You can improve your skill level but drop down on an overall ranking if the other players improve faster or simply if the worst players stop playing, like what happens in the SC2 leagues. His post is adressing the issue that an MMR is important for your own self evaluation. He says there are other ways to do that. You are concerned about how well you play compared to the other players. For me that is far less important. If you are close enough to the top for it to matter, you can check the watch games tab.
Again, improvement is relative to others. You can be improving in an absolute sense, yet still be falling behind. So this kind of improvement is not useful information. Yet you'll never know. Top 5%? How can anyone know whether they are in the top 5%? You can't even tell whether or not your in the top 50% or bottom 50%, let alone 5%. The tacky workaround of the games tabs isn't accurate because it considers all players in the game, not just yourself, and it's in 33% levels.Last edit: 2012-06-16 02:49:37 |
| | Check out "Shadow of the Eternals" on Kickstarter |
|
|
| SKC Brazil. June 16 2012 02:53. Posts 3433 | Profile # |
On June 16 2012 02:48 paralleluniverse wrote:
Show nested quote +On June 16 2012 02:38 SKC wrote:
On June 16 2012 02:22 paralleluniverse wrote:
On June 15 2012 15:20 skipgamer wrote:
Keep it hidden, I haven't seen any compelling arguments for why it should be shown in this thread.
Most of the complaints are from people you can tell have an elitist attitude, I'd rather them go (back) to HoN/LoL anyway.
And if you can't tell if you have improved by simply judging how well you play and watching replays, you're doing it wrong, visible MMR isn't going to help you.
oh, also, if you completely don't know where you stand, just filter your recent matches by high/medium/low skill... at least that will give you a rough indication of what bracket you're in. That should be enough to drive you forward if you're not in high skill games.
You are being completely ridiculous. It is virtually impossible to judge how good someone is based on watching their replays, because how good someone isn't absolute, it's relative. To know how good someone is you need to first know how good EVERYONE is. 2039287. That's a random game ID. Since these players have been matched together, we can assume they all have approximately equal skill. Now tell me how good is this group of players. Top 5%? Top 10%? Top 15%? How is it possible to know?
How is that relevant though? He said "if you can't tell if you have improved by seeing how well you play or watching replays". Give an example with an old replay and a recent replay of a player and it is reasonable to expect being able to see if they improve. Being in the "top 5%" is a completelly diferent subject. You can improve your skill level but drop down on an overall ranking if the other players improve faster or simply if the worst players stop playing, like what happens in the SC2 leagues. His post is adressing the issue that an MMR is important for your own self evaluation. He says there are other ways to do that. You are concerned about how well you play compared to the other players. For me that is far less important. If you are close enough to the top for it to matter, you can check the watch games tab.
Again, improvement is relative to others. You can be improving in an absolute sense, yet still be falling behind. So this kind of improvement is not useful information. Yet you'll never know. Top 5%? How can anyone know whether they are in the top 5%? You can't even tell whether or not your in the top 50% or bottom 50%, let alone 5%. The tacky workaround of the games tabs isn't accurate because it considers all players in the game, not just yourself, and it's in 33% levels.
Well, if it's in 33% "levels", why can't you tell wether you are in the top 50% or bottom 50%? The biggest question is, why does it matter? It's AP pubs. Wanna see if you are actually good? Go play some inhouses, find some cf's, etc. If you are around mid level, improving is very clear. You can easily tell when you start playing better. If you are starting to get good, starting to understand the game, getting to the point where it's not as easy to tell if you are improving, you should probally realize that these kind of games will tell you nothing anyway and you should find somewhere else if you really want to play serious Dota.
Btw, why doesn't SC2 show you your true MMR? Because it's not the best way to gauge skill or improvement. It's too volatile. So they created a diferent way to give you points. Bonus pool makes it too weird, but the reasoning is actually not as crazy as some people believe.Last edit: 2012-06-16 02:59:49 |
|
|
| paralleluniverse Australia. June 16 2012 02:58. Posts 2972 | Profile # |
The issue is complaining about is the fact that people believe MMR correlates to skill, when that isn't exactly true. People below your MMR can be better than you and people above your MMR can be worse then you, it all depends on things like which kind of heroes you play, if you are often carried or carry friends, how hard you "try hard", etc. People act like morons often because they think MMR is a definitive measure of skill, when it is not.
About your example, I would say a guy that plays random and beat 90% of the players is a overall better Starcraft player than a guy that cannon rushes only and beats 95% of the players. That guy is definatelly better at cannon rushing though, which is the only meaningful thing you could conclude from that information. It's fine cannon rushing every single game, but don't believe you are better than someone just because of a stupid ladder system, specially on a non-competitive enviroment. SingSing has a better MMR than Dendi, but I think very few people would argue that he is a better player simply because of that.
Except MMR *does* accurately measure skill. And there's mountains of empirical evidence for it. The fact that Blizzard is able to achieve 50% win ratio matchmaking by using MMR in SC2 and WoW. That TrueSkill is able to do the same, while it also maximizes the number of draws relative to existing rating systems like ELO. (http://research.microsoft.com/pubs/67956/NIPS2006_0688.pdf) Humans have never invented a better measure of skill than MMR.
MMR takes everything into account, everything you've listed and more, because it only cares about whether you win. So any factor that increases or decreases you chance of winning is already implicitly accounted for in MMR.
A random player who beats 90% of players despite having greater game knowledge is still objectively worse than a cannon rusher that beats 95% of players. What matters is not who knows more, or who "plays better". The only thing that matters, and the only thing that should matter in terms of being better, is who WINS mores. 95% > 90%.Last edit: 2012-06-16 03:00:36 |
| | Check out "Shadow of the Eternals" on Kickstarter |
|
|
| BurningSera United Kingdom. June 16 2012 02:59. Posts 5279 | Profile Blog # |
On June 15 2012 23:41 paralleluniverse wrote:
Show nested quote +On June 15 2012 22:21 BurningSera wrote:
the biggest problem of showing mmr or not is due to there is almost no credibility in mmr. People can just use the exact same fcking heroes for 10k games and whoring wins with mates or abusing some dual lane combo to get high k/d/a or mmr but still dont have clue about the game. I hate those players so much because they are usually the worst kind of team mate you can get. I had some very unpleasant experience from 3 months of HoN and i really think that is the real reason killing HoN.
performance bar and battlepoint are enough for epeen stroking imo. i don't want any kind of approvals from the game. Just play well you will meet some good players and these people will be the ones confirm that you are a good player, if you really needed that kind of approval.
You have no idea how MMR works, so you should stop talking. MMR has nothing to do with K/D/A, it's simply got to do with winning. If you use the same hero for 10k games, abusing some duel lane combo and you're winning a lot then you deserve high MMR. If you only know how to cannon rush in SC2, and you can use cannon rush to beat 95% of all players by cannon rushing in.every.single.game, then guess what -- you're better than 95% of all players, and deserve nothing less than an MMR that's in the top 5%.
ye mmr without calculating k/d/a factors in it is even worse, it is basically just w/l rate? We already have w/l rate right now.
sc2 has 3 races and people only care about 1v1. dota is 5v5 and has 100+ heroes. i dont see how mmr of a more emphasizing personal skill game (aka sc2) can be related to a team work orientated game.
and no you dont deserve to be put in high mmr 'If you use the same hero for 10k games, abusing some duel lane combo and you're winning a lot' because you would have very little practical knowledge of all other heroes. It is a 5v5 game which involved 10 heroes in any game. The high mmr would give the player the bragging right to flame the team but in reality he is just a whiner (and not necessary a good player). I have seen too many player like that in HoN and i know ALOT players only play certain heroes throughout their many years of dota and i dont want the whiners get their bragging right to grief/flame in my games. Obviously not all of them are like that, but there are always some bad ones out there ruining people's game.
Hero performance bar is a very fair 'mmr' imo. and obviously we have our mmr hidden atm by Valve, like tobi said, there is no need to show mmr as soon as people get good games (from my personal experience, some good and some bad atm but is acceptable overall). Just keep playing until you are placed in a game vs dendi then you will know you are super good. Otherwise go play scrim/cw/tourney whatever for real higher level games.
We have no idea how valve calculate mmr ATM. You should probably stop talking too since you blindly assume that mmr has nothing to do with k/d/a. |
| | Spawn more Overlords - Zerg player at heart. 2009, 820, Yaphets, YamateH <3 | |
|
|
| paralleluniverse Australia. June 16 2012 03:03. Posts 2972 | Profile # |
On June 16 2012 02:53 SKC wrote:
Show nested quote +On June 16 2012 02:48 paralleluniverse wrote:
On June 16 2012 02:38 SKC wrote:
On June 16 2012 02:22 paralleluniverse wrote:
On June 15 2012 15:20 skipgamer wrote:
Keep it hidden, I haven't seen any compelling arguments for why it should be shown in this thread.
Most of the complaints are from people you can tell have an elitist attitude, I'd rather them go (back) to HoN/LoL anyway.
And if you can't tell if you have improved by simply judging how well you play and watching replays, you're doing it wrong, visible MMR isn't going to help you.
oh, also, if you completely don't know where you stand, just filter your recent matches by high/medium/low skill... at least that will give you a rough indication of what bracket you're in. That should be enough to drive you forward if you're not in high skill games.
You are being completely ridiculous. It is virtually impossible to judge how good someone is based on watching their replays, because how good someone isn't absolute, it's relative. To know how good someone is you need to first know how good EVERYONE is. 2039287. That's a random game ID. Since these players have been matched together, we can assume they all have approximately equal skill. Now tell me how good is this group of players. Top 5%? Top 10%? Top 15%? How is it possible to know?
How is that relevant though? He said "if you can't tell if you have improved by seeing how well you play or watching replays". Give an example with an old replay and a recent replay of a player and it is reasonable to expect being able to see if they improve. Being in the "top 5%" is a completelly diferent subject. You can improve your skill level but drop down on an overall ranking if the other players improve faster or simply if the worst players stop playing, like what happens in the SC2 leagues. His post is adressing the issue that an MMR is important for your own self evaluation. He says there are other ways to do that. You are concerned about how well you play compared to the other players. For me that is far less important. If you are close enough to the top for it to matter, you can check the watch games tab.
Again, improvement is relative to others. You can be improving in an absolute sense, yet still be falling behind. So this kind of improvement is not useful information. Yet you'll never know. Top 5%? How can anyone know whether they are in the top 5%? You can't even tell whether or not your in the top 50% or bottom 50%, let alone 5%. The tacky workaround of the games tabs isn't accurate because it considers all players in the game, not just yourself, and it's in 33% levels.
Well, if it's in 33% "levels", why can't you tell wether you are in the top 50% or bottom 50%? The biggest question is, why does it matter? It's AP pubs. Wanna see if you are actually good? Go play some inhouses, find some cf's, etc. If you are around mid level, improving is very clear. You can easily tell when you start playing better. If you are starting to get good, starting to understand the game, getting to the point where it's not as easy to tell if you are improving, you should probally realize that these kind of games will tell you nothing anyway and you should find somewhere else if you really want to play serious Dota. Btw, why doesn't SC2 show you your true MMR? Because it's not the best way to gauge skill or improvement. It's too volatile. So they created a diferent way to give you points. Bonus pool makes it too weird, but the reasoning is actually not as crazy as some people believe.
Please stop being serially wrong. The reason why Blizzard hides MMR because it's NOT volatile. They said it at Blizzcon, it's stabilizes quickly and hardly ever changes. It's boring. That's why it's hidden.
And because it's a pure skill measure. They want to rank by skill plus time, not pure skill.
In fact, if you've ever played WoW, where MMR is not hidden, you'll know that after you hit the MMR that corresponds to your true skill, it stays there and it's very hard for it to go up or down my much, because it's hard to improve your skills. You can even easily simulate MMR on a computer, to see that it's NOT volatile, it hardly changes after several games (equations are in the paper I linked in the previous post).
And again, if you want to see if you're actually good, there is no way to do it. You can play whatever you want, with whoever you want, but it still tells you nothing about how good you are compared to the playerbase,Last edit: 2012-06-16 03:08:53 |
| | Check out "Shadow of the Eternals" on Kickstarter |
|
|
| SKC Brazil. June 16 2012 03:04. Posts 3433 | Profile # |
On June 16 2012 02:58 paralleluniverse wrote:
Show nested quote +
The issue is complaining about is the fact that people believe MMR correlates to skill, when that isn't exactly true. People below your MMR can be better than you and people above your MMR can be worse then you, it all depends on things like which kind of heroes you play, if you are often carried or carry friends, how hard you "try hard", etc. People act like morons often because they think MMR is a definitive measure of skill, when it is not.
About your example, I would say a guy that plays random and beat 90% of the players is a overall better Starcraft player than a guy that cannon rushes only and beats 95% of the players. That guy is definatelly better at cannon rushing though, which is the only meaningful thing you could conclude from that information. It's fine cannon rushing every single game, but don't believe you are better than someone just because of a stupid ladder system, specially on a non-competitive enviroment. SingSing has a better MMR than Dendi, but I think very few people would argue that he is a better player simply because of that.
Except MMR *does* accurately measure skill. And there's mountains of empirical evidence for it. The fact that Blizzard is able to achieve 50% win ratio matchmaking by using MMR in SC2 and WoW. That TrueSkill is able to do the same, while it also maximizes the number of draws relative to existing rating systems like ELO. (http://research.microsoft.com/pubs/67956/NIPS2006_0688.pdf) Humans have never invented a better measure of skill than MMR. MMR takes everything into account, everything you've listed and more, because it only cares about whether you win. So any factor that increases or decreases you chance of winning is already implicitly accounted for in MMR. A random player who beats 90% of players despite having greater game knowledge is still objectively worse than a cannon rusher that beats 95% of players. What matters is not who knows more, or who "plays better". The only thing that matters, and the only thing that should matter in terms of being better, is who WINS mores. 95% > 90%.
Just because there is no better way doesn't mean it's good, there is no good way. MMR works in a system with fixed teams or 1v1. In a 5v5 game with random teammates it's much, much worse since you cannot say the responsability of a win or lose is yours or from one of your teammates. Unless you can analyze ingame data, it is impossible to create a good MMR system for a game like Dota with the technology we have now.
On June 16 2012 03:03 paralleluniverse wrote:
Show nested quote +On June 16 2012 02:53 SKC wrote:
On June 16 2012 02:48 paralleluniverse wrote:
On June 16 2012 02:38 SKC wrote:
On June 16 2012 02:22 paralleluniverse wrote:
On June 15 2012 15:20 skipgamer wrote:
Keep it hidden, I haven't seen any compelling arguments for why it should be shown in this thread.
Most of the complaints are from people you can tell have an elitist attitude, I'd rather them go (back) to HoN/LoL anyway.
And if you can't tell if you have improved by simply judging how well you play and watching replays, you're doing it wrong, visible MMR isn't going to help you.
oh, also, if you completely don't know where you stand, just filter your recent matches by high/medium/low skill... at least that will give you a rough indication of what bracket you're in. That should be enough to drive you forward if you're not in high skill games.
You are being completely ridiculous. It is virtually impossible to judge how good someone is based on watching their replays, because how good someone isn't absolute, it's relative. To know how good someone is you need to first know how good EVERYONE is. 2039287. That's a random game ID. Since these players have been matched together, we can assume they all have approximately equal skill. Now tell me how good is this group of players. Top 5%? Top 10%? Top 15%? How is it possible to know?
How is that relevant though? He said "if you can't tell if you have improved by seeing how well you play or watching replays". Give an example with an old replay and a recent replay of a player and it is reasonable to expect being able to see if they improve. Being in the "top 5%" is a completelly diferent subject. You can improve your skill level but drop down on an overall ranking if the other players improve faster or simply if the worst players stop playing, like what happens in the SC2 leagues. His post is adressing the issue that an MMR is important for your own self evaluation. He says there are other ways to do that. You are concerned about how well you play compared to the other players. For me that is far less important. If you are close enough to the top for it to matter, you can check the watch games tab.
Again, improvement is relative to others. You can be improving in an absolute sense, yet still be falling behind. So this kind of improvement is not useful information. Yet you'll never know. Top 5%? How can anyone know whether they are in the top 5%? You can't even tell whether or not your in the top 50% or bottom 50%, let alone 5%. The tacky workaround of the games tabs isn't accurate because it considers all players in the game, not just yourself, and it's in 33% levels.
Well, if it's in 33% "levels", why can't you tell wether you are in the top 50% or bottom 50%? The biggest question is, why does it matter? It's AP pubs. Wanna see if you are actually good? Go play some inhouses, find some cf's, etc. If you are around mid level, improving is very clear. You can easily tell when you start playing better. If you are starting to get good, starting to understand the game, getting to the point where it's not as easy to tell if you are improving, you should probally realize that these kind of games will tell you nothing anyway and you should find somewhere else if you really want to play serious Dota. Btw, why doesn't SC2 show you your true MMR? Because it's not the best way to gauge skill or improvement. It's too volatile. So they created a diferent way to give you points. Bonus pool makes it too weird, but the reasoning is actually not as crazy as some people believe.
Please stop being serially wrong. The reason why Blizzard hides MMR it's because it NOT volatile. They said it at Blizzcon, it's stabilizes quickly and hardly ever changes. It's boring. That's why it's hidden. And because it's a pure skill measure. They want to rank by skill plus time, not pure skill. In fact, if you've ever played WoW, where MMR is not hidden, you'll know it after you hit the MMR that corresponding to your true skill, it stays there and it's very hard for it go up or down my much, because it's hard to improve your skills. You can even easily simulate MMR on a computer, to see that it NOT volatile, it hardly ever changes (equations are in the the paper I linked in the previous post). And again, if you want to see if you're actually good, there is no way to do it. You can play whatever you want, with the whoever you want, but it still tells you nothing about how good you are compared to all the players.
"Overview
At the heart of the system is a hidden value known as the matchmaking rating, or MMR for short. Matchmaking rating helps to ensure you play against players around your skill level and influences how many points you stand to gain or lose per match. Your points will drift toward your hidden MMR over a period of time, but because MMR is more volatile than points, your MMR is never cemented at a fixed value. For this reason, it is extraordinarily difficult to reverse engineer MMR from points."
Taken from http://www.teamliquid.net/forum/viewmessage.php?topic_id=195273.
It's a well respected guide in TL. Show me Blizzard saying the opposite first, because I'm also pretty sure they said MMR is volatile.
Last edit: 2012-06-16 03:15:07 |
|
|
| paralleluniverse Australia. June 16 2012 03:14. Posts 2972 | Profile # |
On June 16 2012 03:04 SKC wrote:
Show nested quote +On June 16 2012 02:58 paralleluniverse wrote:
The issue is complaining about is the fact that people believe MMR correlates to skill, when that isn't exactly true. People below your MMR can be better than you and people above your MMR can be worse then you, it all depends on things like which kind of heroes you play, if you are often carried or carry friends, how hard you "try hard", etc. People act like morons often because they think MMR is a definitive measure of skill, when it is not.
About your example, I would say a guy that plays random and beat 90% of the players is a overall better Starcraft player than a guy that cannon rushes only and beats 95% of the players. That guy is definatelly better at cannon rushing though, which is the only meaningful thing you could conclude from that information. It's fine cannon rushing every single game, but don't believe you are better than someone just because of a stupid ladder system, specially on a non-competitive enviroment. SingSing has a better MMR than Dendi, but I think very few people would argue that he is a better player simply because of that.
Except MMR *does* accurately measure skill. And there's mountains of empirical evidence for it. The fact that Blizzard is able to achieve 50% win ratio matchmaking by using MMR in SC2 and WoW. That TrueSkill is able to do the same, while it also maximizes the number of draws relative to existing rating systems like ELO. (http://research.microsoft.com/pubs/67956/NIPS2006_0688.pdf) Humans have never invented a better measure of skill than MMR. MMR takes everything into account, everything you've listed and more, because it only cares about whether you win. So any factor that increases or decreases you chance of winning is already implicitly accounted for in MMR. A random player who beats 90% of players despite having greater game knowledge is still objectively worse than a cannon rusher that beats 95% of players. What matters is not who knows more, or who "plays better". The only thing that matters, and the only thing that should matter in terms of being better, is who WINS mores. 95% > 90%.
Just because there is no better way doesn't mean it's good, there is no good way. MMR works in a system with fixed teams or 1v1. In a 5v5 game with random teammates it's much, much worse since you cannot say the responsability of a win or lose is yours or from one of your teammates. Unless you can analyze ingame data, it is impossible to create a good MMR system for a game like Dota with the technology we have now.
This is an argument from ignorance. Just because you don't understand how MMR can measure skill accurately for a team doesn't mean it can't.
In fact, TrueSkill was designed for team games, and again, there are mountains of empirical evidence that it works correctly, some of which are presented in the paper I linked.
The "team dynamics" is accounted for in an aggregation formula (t_j in the paper linked) which computes the team's performance based on the performance of the individual players. The function that is used depends on the nature of the game. DotA is the type of game where a weak player can make you lose, so the function would probably be something like a weighted combination of the minimum performance plus the average performance. The point is this function can be tuned, to maximize the accuracy of MMR. In the paper t_j is simply the sum, and yet TrueSkill is still remarkably successful.
And the mountains of empirical evidence suggests that MMR is a great measure of skill. Again, how do you explain near 50/50 matchmaking for games like WoW and team SC2? Moreover, Tom Chilton laid out the workings of WoW's MMR system on the old forums when MMR was first introduced to WoW, which says that it measures skill. How do you think people like ExcaliburZ got all the conceptual details of the SC2 MMR system, without Blizzard having said a word about SC2's system and without any numerical analysis of real data? Chilton spilled the beans on WoW's MMR system, that's how.Last edit: 2012-06-16 03:23:40 |
| | Check out "Shadow of the Eternals" on Kickstarter |
|
|
| SKC Brazil. June 16 2012 03:25. Posts 3433 | Profile # |
On June 16 2012 03:14 paralleluniverse wrote:
Show nested quote +On June 16 2012 03:04 SKC wrote:
On June 16 2012 02:58 paralleluniverse wrote:
The issue is complaining about is the fact that people believe MMR correlates to skill, when that isn't exactly true. People below your MMR can be better than you and people above your MMR can be worse then you, it all depends on things like which kind of heroes you play, if you are often carried or carry friends, how hard you "try hard", etc. People act like morons often because they think MMR is a definitive measure of skill, when it is not.
About your example, I would say a guy that plays random and beat 90% of the players is a overall better Starcraft player than a guy that cannon rushes only and beats 95% of the players. That guy is definatelly better at cannon rushing though, which is the only meaningful thing you could conclude from that information. It's fine cannon rushing every single game, but don't believe you are better than someone just because of a stupid ladder system, specially on a non-competitive enviroment. SingSing has a better MMR than Dendi, but I think very few people would argue that he is a better player simply because of that.
Except MMR *does* accurately measure skill. And there's mountains of empirical evidence for it. The fact that Blizzard is able to achieve 50% win ratio matchmaking by using MMR in SC2 and WoW. That TrueSkill is able to do the same, while it also maximizes the number of draws relative to existing rating systems like ELO. (http://research.microsoft.com/pubs/67956/NIPS2006_0688.pdf) Humans have never invented a better measure of skill than MMR. MMR takes everything into account, everything you've listed and more, because it only cares about whether you win. So any factor that increases or decreases you chance of winning is already implicitly accounted for in MMR. A random player who beats 90% of players despite having greater game knowledge is still objectively worse than a cannon rusher that beats 95% of players. What matters is not who knows more, or who "plays better". The only thing that matters, and the only thing that should matter in terms of being better, is who WINS mores. 95% > 90%.
Just because there is no better way doesn't mean it's good, there is no good way. MMR works in a system with fixed teams or 1v1. In a 5v5 game with random teammates it's much, much worse since you cannot say the responsability of a win or lose is yours or from one of your teammates. Unless you can analyze ingame data, it is impossible to create a good MMR system for a game like Dota with the technology we have now.
This is an argument from ignorance. Just because you don't understand how MMR can measure skill accurately for a team doesn't mean it can't. In fact, TrueSkill was designed for team games, and again, there are mountains of empirical evidence that it works correctly, some of which are presented in the paper I linked. The "team dynamics" is accounted for in an aggregation formula which computes the team's MMR based on the MMR on the invidual players. The function that is use can depends on nature of the game. DotA is the type of game where a weak player can make you lose, so the function would likely be something like a weighted combination of the minimum MMR plus the average MMR. The point is this function can be tuned, to maximize the accuracy of MMR. And the mountains empirical evidence suggests that MMR is a great measure of skill. Again, how do you explain near 50/50 matchmaking?
There is no 50/50 matchmaking? I just said MMR works great in 1v1, so it obviously work fine in SC. Show me an example where there is a 50/50 winrate in a random teams game like Dota in balanced conditions. There aren't any, so you can't possibly use that as evidence.
http://research.microsoft.com/en-us/projects/trueskill/
4 players with 2 teams gives you 46 games per gamer. 8 playes would mean 91, so with 5 playes we can easily assume over 50 games. Now, that is in IDEAL conditions, made for FPS's, that are a lot less variable than Dota. It says it can take up to 3 times that number, depending on things like variablility of play, amount of players, etc. Dota also adds other difficulties, such as a big number of arranged team, picking influencing the game hugely, not a big issue in an FPS for example, diferent roles, etc. It shouldn't be a surprise if you need 3-4 times as many games per gamer to judge someone's skill level. That's around 200 games per gamer. That's a lot. That also assumes you don't improve between 200 games.
It is not a great system for a complicated, random team game like Dota, where your own ability ussually varies a lot just because you play diferent roles and heroes and issues outside the game, like picks, have a huge influence. Games like SC2 are completelly static, you alway play the same race and ussually the same way. It would be way worse if people switched races constantly, for example. In WoW, you also ussually have the same team and the same team composition. Saying your team is better than the other team because you won makes a lot more sense than saying all 5 of you are worse than all 5 of the other team, but that's all a system like this can do.Last edit: 2012-06-16 03:31:53 |
|
|
| paralleluniverse Australia. June 16 2012 03:38. Posts 2972 | Profile # |
"Overview At the heart of the system is a hidden value known as the matchmaking rating, or MMR for short. Matchmaking rating helps to ensure you play against players around your skill level and influences how many points you stand to gain or lose per match. Your points will drift toward your hidden MMR over a period of time, but because MMR is more volatile than points, your MMR is never cemented at a fixed value. For this reason, it is extraordinarily difficult to reverse engineer MMR from points." Taken from http://www.teamliquid.net/forum/viewmessage.php?topic_id=195273. It's a well respected guide in TL. Show me Blizzard saying the opposite first, because I'm also pretty sure they said MMR is volatile.
He's saying that MMR isn't fixed, but fluctuates a bit around your true skill, which is pretty much what I said. And you would already know this if you play WoW arena's where you're essentially mostly fluctuating +/- 100 points (on a scale from 0-3000, so 100 is very small). How does he know this given that he admits that "it is extraordinarily difficult [impossible] to reverse engineer MMR", where is his source? The source is information about WoW's system. |
| | Check out "Shadow of the Eternals" on Kickstarter |
|
|
| sereniity Sweden. June 16 2012 03:42. Posts 1132 | Profile Blog # |
On June 15 2012 22:59 epoc wrote:
Stats system made Hon's community pretty bad.
The DoTA community is equally as bad and doesn't have a stat system, besides that there's no need for KDR to be shown etc, I just want a damn in-game ladder... |
| |
|
| Prev 1 2 3 4 5 6 7 8 9 10 Next All | |
|
|
|
| |
|
|
