|
On November 21 2013 20:54 Destructicon wrote:Do we know exactly how the MMR decay works? Say, is it possible to play 5 games per day and maintain or grow your current MMR? Do you need to play more then 5 games per day and if so, how may more?
Do you need to have a set number of games played by the end of a season?
I'm asking because its quite relevant to know how much activity do you really need per day or per week to stop the decay and start seeing results again.
check this thread out. decay seems to start at about two to three weeks of inactivity.
|
4713 Posts
On November 21 2013 21:06 CycoDude wrote:Show nested quote +On November 21 2013 20:54 Destructicon wrote:Do we know exactly how the MMR decay works? Say, is it possible to play 5 games per day and maintain or grow your current MMR? Do you need to play more then 5 games per day and if so, how may more?
Do you need to have a set number of games played by the end of a season?
I'm asking because its quite relevant to know how much activity do you really need per day or per week to stop the decay and start seeing results again. check this thread out. decay seems to start at about two to three weeks of inactivity.
Ok, if it starts 2-3 weeks after inactivity, then if you play even 1 game per day you shouldn't get demoted. Unless they also changed the definition of inactivity.
|
On November 21 2013 05:34 iamcaustic wrote:Here's my fancy attempt at graphs, based on AM region 1v1 data from nios.kr (June 1st, 2013 vs. November 18th, 2013). Basically demonstrates the changes in percentile thresholds at this point in time compared to the middle of this year. Basically, if you got demoted this season, don't feel too bad about it. Chances are your MMR didn't actually drop significantly.
That post should be in the OP. Best explanation.
|
Czech Republic12129 Posts
On November 21 2013 21:11 FaCE_1 wrote:Show nested quote +On November 21 2013 05:34 iamcaustic wrote:Here's my fancy attempt at graphs, based on AM region 1v1 data from nios.kr (June 1st, 2013 vs. November 18th, 2013). Basically demonstrates the changes in percentile thresholds at this point in time compared to the middle of this year. Basically, if you got demoted this season, don't feel too bad about it. Chances are your MMR didn't actually drop significantly. That post should be in the OP. Best explanation.
I don't get this explanation. How does this explain me being matched against opponents who are two or three levels above my skill? Back in Wings I wouldn't get those matches unless I won 10+ games in a row and even then I wouldn't get so hard opponents. Now I have the score 19:31 and getting roflstomped left and right... IMO this explains nothing about matchmaking being broken.
I can see some of my games being lost to my bad mechanics, but when I see in the replay how good my enemy was... damn.
This graph only explains why some friends of mine started their season in platinum instead of diamond :-)
|
So in Season 4, getting into Plat mean't you were top 33.9% and in Season 6 you are the top 16.3%.
|
Maybe not true for EU or KR, but the ladder is getting easier and easier on NA.
Most of the talented players that were looking to play at an amateur online cup level have decreased significantly.
The prize pools weren't getting higher (if not lower), Koreans would automatically take the top prizes 90% of the time.
Things are getting more cast and attention because things became more established, but the talent of the "masses" have decreased from many many many talented NA players leaving for console/moba.
Are people getting better steadily? yes, but are people getting better at the amateur level? No
Amateur- online tournament participants, minimum mid-masters.
|
On November 21 2013 08:19 DinoMight wrote:Show nested quote +On November 21 2013 08:15 dr.fahrenheit wrote:
Just get rid of one (or two) league(s) The problem with this, I feel, is that at every level, there is a huge disparity between players. A mid Silver would annihilate a low Bronze. A high Masters typically beats a low Masters pretty one-sidedly... The fewer leagues you have, the bigger the range of skill you have in each league, which is counter effective. The system is ideally trying to pit you against people of your level.
the game uses your MMR, not your league, to find you an opponent
your league is just a badge to vaguely reflect your MMR
lowering the number of leagues would make it harder for newbies to feel accomplishment though
|
On November 21 2013 22:30 TaishiCi wrote:
Maybe not true for EU or KR, but the ladder is getting easier and easier on NA.
Most of the talented players that were looking to play at an amateur online cup level have decreased significantly.
The prize pools weren't getting higher (if not lower), Koreans would automatically take the top prizes 90% of the time.
Things are getting more cast and attention because things became more established, but the talent of the "masses" have decreased from many many many talented NA players leaving for console/moba.
Are people getting better steadily? yes, but are people getting better at the amateur level? No
Amateur- online tournament participants, minimum mid-masters.
How so? Any proof for such a claim? Most people in the thread are saying they are getting demoted for no apparent reason, myself included. We are trying to identify why the ladder behaves the way it does, not say one server is better than another.
|
I'm still waiting for someone to explain how a system which begins to reduce your MMR after two weeks of inactivity, and can only drop you a full league after nearly four weeks of inactivity, is chiefly responsible for the majority of ladder players dropping a league. Moreover, I'm interested in hearing how this system is expected to have a larger effect than a drop in the player base of >50% from 2013 Season 4 to 2013 Season 6.
|
Obviously, fewer players play and with the MMR-decay masters/dimaonds get demoted to gold/platinum and when they decide to play again they play the real gold/plat players...
|
Was in Diamond beginning of HOTS and end of WOL. I then got demoted to Platinum and was struggling to even win 50% of my matches.
Now they have put me in Gold and I am currently winning a little more than 50%. I was mad to being put so low but now I get it.
I now understand why with the charts! Tnx for the input guys!
|
I think this kind of behavior is pretty typical with competitive games. Saw the same thing happen in WC3 about a year after TFT was out.
|
|
|
On November 21 2013 23:49 c0ldfusion wrote:
I think this kind of behavior is pretty typical with competitive games. Saw the same thing happen in WC3 about a year after TFT was out.
Yeah. Like i said earlier... look at bw ladder today. It's incredibly hard after people played the game for 10years. There are litterly NO newb  (except me :p)
|
On November 21 2013 23:46 CrankOut wrote:
Obviously, fewer players play and with the MMR-decay masters/dimaonds get demoted to gold/platinum and when they decide to play again they play the real gold/plat players...
That is not my point; I understand the mechanism, and I do not dispute that MMR decay will cause some ladder deflation, as well as increased variance in skill at the same MMR. I'm asking people to think about and defend the scale that they are proposing. Given no change in the average skill across the ladder, how many games do people have to lose to decayed players for a two league shift to occur? What percentage of players are going inactive for 2+ weeks, and then playing regularly enough following that break to affect the behavior of the entire ladder?
According to the graph above (and replacing the Season 6 numbers with the current ones from the same site), ~85,000 players have dropped from ranked since Season 4. This is more than half the players, and this is larger than Season 4 bronze and silver combined. If these players were evenly distributed across leagues, then this would have no effect. If these players were disproportionately low MMR players, the ladder would rebalance around it, and players would be demoted. This has a large and obvious effect, while a global effect of MMR requires a substantial population of players with month-long inactivity cycles.
People are going to be sorely disappointed when Blizzard eliminates MMR decay and they don't get promoted.
|
On November 21 2013 23:31 Ctone23 wrote:Show nested quote +On November 21 2013 22:30 TaishiCi wrote:
Maybe not true for EU or KR, but the ladder is getting easier and easier on NA.
Most of the talented players that were looking to play at an amateur online cup level have decreased significantly.
The prize pools weren't getting higher (if not lower), Koreans would automatically take the top prizes 90% of the time.
Things are getting more cast and attention because things became more established, but the talent of the "masses" have decreased from many many many talented NA players leaving for console/moba.
Are people getting better steadily? yes, but are people getting better at the amateur level? No
Amateur- online tournament participants, minimum mid-masters. How so? Any proof for such a claim? Most people in the thread are saying they are getting demoted for no apparent reason, myself included. We are trying to identify why the ladder behaves the way it does, not say one server is better than another.
He likely means the very top end of the MMR range, which is likely true. If lots of top end players quit then it affects the competition level at the top of the ladder. For example this is how season start GM MMR entry thresholds have shifted in last 3 seasons (likely no offset changes during these seasons, so the numbers are comparable):
NA:
S14: 1900 (or slightly lower), S15: 1830, S16: ~1800
EU:
S14: 1950, S15: 2130, S16: ~2070
|
On November 22 2013 00:09 iggym wrote:Show nested quote +On November 21 2013 23:46 CrankOut wrote:
Obviously, fewer players play and with the MMR-decay masters/dimaonds get demoted to gold/platinum and when they decide to play again they play the real gold/plat players... That is not my point; I understand the mechanism, and I do not dispute that MMR decay will cause some ladder deflation, as well as increased variance in skill at the same MMR. I'm asking people to think about and defend the scale that they are proposing. Given no change in the average skill across the ladder, how many games do people have to lose to decayed players for a two league shift to occur? What percentage of players are going inactive for 2+ weeks, and then playing regularly enough following that break to affect the behavior of the entire ladder? According to the graph above (and replacing the Season 6 numbers with the current ones from the same site), ~85,000 players have dropped from ranked since Season 4. This is more than half the players, and this is larger than Season 4 bronze and silver combined. If these players were evenly distributed across leagues, then this would have no effect. If these players were disproportionately low MMR players, the ladder would rebalance around it, and players would be demoted. This has a large and obvious effect, while a global effect of MMR requires a substantial population of players with month-long inactivity cycles. People are going to be sorely disappointed when Blizzard eliminates MMR decay and they don't get promoted.
First the graph you are referring to is incorrect. The first numbers are not from start of S14, but from the end of S13. In the beginning of S14 Blizzard changed the offsets and tried to fix the distribution. After that Blizzard has not changed the offsets (or if they have, the change has been very small. In a week or two I can check this season's offsets more carefully).
Also in the end of the season naturally more people have played their 1v1 placements. Thus the player numbers are not comparable if you compare numbers from start of the season to numbers from end of the season.
But if you look at table I showed earlier:
http://www.teamliquid.net/forum/viewpost.php?post_id=20264300
It compares numbers from last season (S15) and this season (S16). First it compares numbers from 1 week from the season start. You can notice major shift towards the lower leagues. Then you can check the numbers from the end of last season. Between 8 k to 15 k people entered both EU and NA 1v1 ladders each week after the first week. You can easily deduce that major percentage of the ladder population goes inactive for 2 weeks or more during one season. MMR decay is logically the main reason for the population shift, like I answered to you already earlier: http://www.teamliquid.net/forum/viewmessage.php?topic_id=436056¤tpage=5#100
And yes. If the decay mechanism would now be removed, nothing visible would happen for a while. Ladder is so messed up, that it takes lots of time for people to fight back to the MMR levels they were typically before. The distribution would start healing up, but it would take time. In this case it might be beneficial to do a full MMR reset to all players. The MMR changes much more rapidly when starting from blank MMR. People can get to master range with less than 10 games when starting from blank MMR (but master promotion at earliest at 25th game). But after the MMR change rate has stabilized, it takes easily 20 straight wins or more to go over one league MMR range.
On November 22 2013 00:09 iggym wrote:
Given no change in the average skill across the ladder, how many games do people have to lose to decayed players for a two league shift to occur?
For each player ~ 16 to 20 games per league. Thus if you lose 40 games in a row, then your MMR is little over 2 leagues lower than before (if no decay happens in between).
|
One thing that seems to be forgotten is the difference in Ranked and unranked play that factors in quite a bit.
In Ranked im giving it my best and hitting high masters consistantly.
In unranked my MMR makes me face Gold - Diamond -- rarely masters players - because i actually start to lose a lot of games offracing or playing weird stuff.
This is why you are facing masters players, though you arent getting an increase in your own points according to winning a masters guy. Worst case you gain points for beating the gold level player he is in unranked even if you are diamond and his portrait is masters.
|
So basically silver and bronze are about 60% of the ladder right now, which is way off from Blizzard's intended distribution.
Does anyone think Blizzard will try to remedy this at all? Anyone with insight please let us know!
I think the biggest problem that this creates isn't "OMG I'm Plat now I used to be Masters" but rather that being in a specific league is no longer really an indication of one's skill, so it's hard to tell how good someone is without knowing their MMR (which is hidden).
Anyone trying to run a team league or cup or any other sort of amateur tournament for fun will run into issues :/
|
On November 22 2013 00:57 korona wrote:Show nested quote +On November 22 2013 00:09 iggym wrote:On November 21 2013 23:46 CrankOut wrote:
Obviously, fewer players play and with the MMR-decay masters/dimaonds get demoted to gold/platinum and when they decide to play again they play the real gold/plat players... That is not my point; I understand the mechanism, and I do not dispute that MMR decay will cause some ladder deflation, as well as increased variance in skill at the same MMR. I'm asking people to think about and defend the scale that they are proposing. Given no change in the average skill across the ladder, how many games do people have to lose to decayed players for a two league shift to occur? What percentage of players are going inactive for 2+ weeks, and then playing regularly enough following that break to affect the behavior of the entire ladder? According to the graph above (and replacing the Season 6 numbers with the current ones from the same site), ~85,000 players have dropped from ranked since Season 4. This is more than half the players, and this is larger than Season 4 bronze and silver combined. If these players were evenly distributed across leagues, then this would have no effect. If these players were disproportionately low MMR players, the ladder would rebalance around it, and players would be demoted. This has a large and obvious effect, while a global effect of MMR requires a substantial population of players with month-long inactivity cycles. People are going to be sorely disappointed when Blizzard eliminates MMR decay and they don't get promoted. First the graph you are referring to is incorrect. The first numbers are not from start of S14, but from the end of S13. In the beginning of S14 Blizzard changed the offsets and tried to fix the distribution. After that Blizzard has not changed the offsets (or if they have, the change has been very small. In a week or two I can check this season's offsets more carefully). Also in the end of the season naturally more people have played their 1v1 placements. Thus the player numbers are not comparable if you compare numbers from start of the season to numbers from end of the season. But if you look at table I showed earlier: http://www.teamliquid.net/forum/viewpost.php?post_id=20264300It compares numbers from last season (S15) and this season (S16). First it compares numbers from 1 week from the season start. You can notice major shift towards the lower leagues. Then you can check the numbers from the end of last season. Between 8 k to 15 k people entered both EU and NA 1v1 ladders each week after the first week. You can easily deduce that major percentage of the ladder population goes inactive for 2 weeks or more during one season. MMR decay is logically the main reason for the population shift, like I answered to you already earlier: http://www.teamliquid.net/forum/viewmessage.php?topic_id=436056¤tpage=5#100And yes. If the decay mechanism would now be removed, nothing visible would happen for a while. Ladder is so messed up, that it takes lots of time for people to fight back to the MMR levels they were typically before. The distribution would start healing up, but it would take time. In this case it might be beneficial to do a full MMR reset to all players. The MMR changes much more rapidly when starting from blank MMR. People can get to master range with less than 10 games when starting from blank MMR (but master promotion at earliest at 25th game). But after the MMR change rate has stabilized, it takes easily 20 straight wins or more to go over one league MMR range. Show nested quote +On November 22 2013 00:09 iggym wrote:
Given no change in the average skill across the ladder, how many games do people have to lose to decayed players for a two league shift to occur?
For each player ~ 16 to 20 games per league. Thus if you lose 40 games in a row, then your MMR is little over 2 leagues lower than before (if no decay happens in between).
Thank you for such a good response!
I missed those numbers you posted before, sorry about that. I was not aware that the ladder grew by that rate (~10K/week) each season. That is far, far, far more than I intuited, and you're right, that means a substantial portion of the player base goes inactive then comes back. It not only means that I was wrong about the number of half-season inactive players, but that I was also wrong about the number of players who have dropped out entirely... so MMR decay has a bigger effect than I realized and the base drop-off has a smaller effect than I realized.
Whoops. Sorry folks.
|
|
|
|
|
|
EPT![[image loading]](https://pbs.twimg.com/media/BZaXMeoCMAEd2k3.png)
