|
On March 09 2015 08:04 ZackAttack wrote: Very interesting. It seems like either blizzard has really not done anything or this year is a very different problem from last year. Like others have noted, it could be similar to the January update, where the actual ladder changes didn't happen until a couple days after the "we've changed things" announcement.
|
Indeed there has not been any trend changes since the supposed update in nios.kr data. If I have time in the coming week I can provide you nios.kr data since January 22nd for you to complete your graph (thought the logger has not been continuously on and there are some gaps here and there). Will be a busy week though so no promises. And if Blizzard does more noticeable changes before that, then there will not be need.
One interesting possibility regarding the offset/threshold changes is what if they actually increased the offset/threshold values for many leagues (not decreased like everyone expected). And as there are no mid-season demotions the changes are not visible in nios.kr data. This speculation is based on this: http://www.teamliquid.net/forum/viewpost.php?post_id=23869343
Edit: I presume you would like them with EST time stamps (GMT -0500)?
|
This came too late for me, omfg. I played like a beast #1 Diamond with a good win/lose ratio but then I had to go back to studying. Thnx for another coincidence blizzard...
|
Been playing this game for 5 years and BW before it. Stopped playing for 3 months and went from playing Masters in Diamond to ~Gold.
#Blizzard logic.
The fun part is I get to grind MMR streaming Random race now
|
HEY!?!?
I got promoted! ITS WORKING :O IM DIAMOND NOW
After 74 wins in row im finaly dia (38-0 this season)
|
On March 10 2015 04:39 TheBloodyDwarf wrote:HEY!?!? I got promoted! ITS WORKING :O IM DIAMOND NOW After 74 wins in row im finaly dia (38-0 this season) That's just luck, you'll start next season in Gold League. That's where we all belong!
|
United States12212 Posts
So apparently the early data shows that the players who are playing post-adjustment are showing up in their expected leagues. I think frajen's post on viewing the distribution through the activity filter is probably pretty close. What's I think most disheartening to all the Diamond hopefuls out there is that it's pretty late in the season. So, if you were to take a look at who the active post-adjustment players are right now, the distribution is probably pretty accurate. That's not going to be reflected on Nios until day 1 of next season. Prediction: all those same Diamond players are going to be pissed when they're still in Diamond next season.
|
On March 10 2015 08:58 Excalibur_Z wrote: So apparently the early data shows that the players who are playing post-adjustment are showing up in their expected leagues. I think frajen's post on viewing the distribution through the activity filter is probably pretty close. What's I think most disheartening to all the Diamond hopefuls out there is that it's pretty late in the season. So, if you were to take a look at who the active post-adjustment players are right now, the distribution is probably pretty accurate. That's not going to be reflected on Nios until day 1 of next season. Prediction: all those same Diamond players are going to be pissed when they're still in Diamond next season.
I'm confused - so you're saying masters is supposed to be 0.86%? I am currently in masters so I really don't care (for myself), but my friends are working their asses off for the promo and are at the top of diamond league, without promos. What gives?
|
United States12212 Posts
Short answer yes*, long answer read frajen's post: http://www.teamliquid.net/forum/starcraft-2/459913-ranked-ladder-activity-analysis?page=2#34
Obviously it's not supposed to be 0.86%, but that's because it's not 0.86%. What Nios sees and what Blizzard sees are completely different, and because Blizzard defines the boundaries that's what matters. Therefore, you need to have the capability to look at it through their lens, and that's what frajen's post aspires to do. Here's the difference:
Nios: "Here are all the accounts in each league." Players: "But that's way off from what it's supposed to be, this isn't right at all!" Blizzard: "Here are all the players in each league who are actively playing the game (have less than X bonus pool)."
If Blizzard sees Master at 2%, then that's what it is. We can say it's wrong all we want, but we're doing so with incomplete information and it's important to keep that in perspective.
|
On March 10 2015 10:21 Excalibur_Z wrote:Short answer yes*, long answer read frajen's post: http://www.teamliquid.net/forum/starcraft-2/459913-ranked-ladder-activity-analysis?page=2#34Obviously it's not supposed to be 0.86%, but that's because it's not 0.86%. What Nios sees and what Blizzard sees are completely different, and because Blizzard defines the boundaries that's what matters. Therefore, you need to have the capability to look at it through their lens, and that's what frajen's post aspires to do. Here's the difference: Nios: "Here are all the accounts in each league." Players: "But that's way off from what it's supposed to be, this isn't right at all!" Blizzard: "Here are all the players in each league who are actively playing the game (have less than X bonus pool)." If Blizzard sees Master at 2%, then that's what it is. We can say it's wrong all we want, but we're doing so with incomplete information and it's important to keep that in perspective. Thank you for the explanation.
|
That all makes a lot more sense now. I think Blizzard didn't do an optimal job of communicating it that way, though.
Otherwise, it also says a lot about the actual state of the ladder.
|
On March 10 2015 10:21 Excalibur_Z wrote:Short answer yes*, long answer read frajen's post: http://www.teamliquid.net/forum/starcraft-2/459913-ranked-ladder-activity-analysis?page=2#34Obviously it's not supposed to be 0.86%, but that's because it's not 0.86%. What Nios sees and what Blizzard sees are completely different, and because Blizzard defines the boundaries that's what matters. Therefore, you need to have the capability to look at it through their lens, and that's what frajen's post aspires to do. Here's the difference: Nios: "Here are all the accounts in each league." Players: "But that's way off from what it's supposed to be, this isn't right at all!" Blizzard: "Here are all the players in each league who are actively playing the game (have less than X bonus pool)." If Blizzard sees Master at 2%, then that's what it is. We can say it's wrong all we want, but we're doing so with incomplete information and it's important to keep that in perspective.
As others have said, they've just done a terrible, terrible job of explaining... anything.
If percentages ARE on target, then they should have simply said so, giving at least some indication as to why they think so. Still even in the data presented, this does not seem to be the case. Instead they have led us to believe in their posts that targets are indeed off and the ladder distribution itself is in need of adjustment.
In addition, initial matchmaking for new players looks to have not been fixed, which they also said would be fixed in this patch (and which is perhaps even a larger issue than the distributions).
|
That "active players" thing is bullshit, or they explained their modifications so bad it hurts my soul. "The modifications are live and everything is going as expected" ? Why did they acknowledge the need of some kind of change to the league boundaries if those changes do nothing ? It makes absolutely zero sense, a change that does nothing is not a needed change and they wouldn't have rushed to implement it in the current season. So it's just not working, and they're trying to buy time with answers devoid of meaning, which they always do.
|
Well, the problem is that if your MMR dropped due to decay, they're not going to re-increase it for you.
So I'd been near top of Diamond for a while. Even made it to Masters once. Didn't play for 2 months and got put into Gold. Changing the range limits and whatever isn't going to get me back to Diamond. Because MMR DECAY has already taken place.
I think the system is now ready for you to get back into whatever league you belong in, but keep in mind there are still a lot of guys like me in Gold/Plat who've been playing for 5 years, took a couple of months off, and got demoted 1-2 ranks.
So our MMR is Gold and we're fighting each other in Gold. The system has no way of telling that we should both be Diamond.
|
On March 09 2015 08:28 korona wrote:Indeed there has not been any trend changes since the supposed update in nios.kr data. If I have time in the coming week I can provide you nios.kr data since January 22nd for you to complete your graph (thought the logger has not been continuously on and there are some gaps here and there). Will be a busy week though so no promises. And if Blizzard does more noticeable changes before that, then there will not be need. One interesting possibility regarding the offset/threshold changes is what if they actually increased the offset/threshold values for many leagues (not decreased like everyone expected). And as there are no mid-season demotions the changes are not visible in nios.kr data. This speculation is based on this: http://www.teamliquid.net/forum/viewpost.php?post_id=23869343Edit: I presume you would like them with EST time stamps (GMT -0500)? Here are nios.kr stats for NA server since January 22nd: http://www.mediafire.com/download/e3vaic64a2w9awr/2015-03-10_nios_kr_data_americas.zip
I formatted them into same format as in Mintograde's sheet (tab separates each column). Some data is missing such as race percentages and people in other servers (have the numbers for other servers in separate files but did not bother to combine). This data also includes the raw numbers of players in each league that are missing from Mintograde's data. As there seems to have been daylight saving time change in Mintograde's sheet from EST to EDT I added that there as well.
Observations: Now there are visible trend changes since Blizzard made the change. 1) Relative size of bronze has started increasing, 2) Relative size of gold has started to decrease very slowly.
|
United States12212 Posts
On March 10 2015 21:12 [PkF] Wire wrote: That "active players" thing is bullshit, or they explained their modifications so bad it hurts my soul. "The modifications are live and everything is going as expected" ? Why did they acknowledge the need of some kind of change to the league boundaries if those changes do nothing ? It makes absolutely zero sense, a change that does nothing is not a needed change and they wouldn't have rushed to implement it in the current season. So it's just not working, and they're trying to buy time with answers devoid of meaning, which they always do.
That's part of the risk with doing it so late in the season I think. I spoke to Psione yesterday trying to wrap my head around it. He said that it's very important that everyone plays normally so they get the most accurate data possible, and that it was a change that will only apply moving forward. He's not a dev so there's always the possibility of "dev translation" issues, but what this means to me is that they're probably looking at two separate population parses: pre- and post-update. If the pre-update league boundaries were messed up and needed to be fixed, and now the post-update boundaries are showing 2%/18%/20%/32%/20%/8% then they're not going to see any problems because hey, those numbers look correct now. The post-update population parses are also necessarily ongoing. However, that means that all the people looking at nios.kr for movement shifts may not even see anything.
In order to see what Blizzard is currently looking at, we would probably have to apply some filter where we only look at accounts that have played a game within the past 5 days, after the push happened. That would mean we would see relatively few data points (like, I dunno, 200 master players, 1800 diamond players, 2000 plat players, 3200 gold players, etc.) but that would be enough to draw a crude chart which would gradually get smoother as more players logged in to play games. I actually don't see an ability to do that on Nios' site (I know SC2Ranks used to have custom filtering but I think since he switched over to copying Nios' method he may have lost that).
And right as I say that, korona posts exactly that parse. Awesome. (E: Guess not, this looks like a total players parse again, but done from end of Jan+).
|
On March 11 2015 02:33 Excalibur_Z wrote: And right as I say that, korona posts exactly that parse. Awesome. (E: Guess not, this looks like a total players parse again, but done from end of Jan+). Indeed. I just added missing data for Mintograde. Have been logging the total numbers to recognize if there are potential unannounced offset threshold changes during the seasons (for larger changes the trend changes are quickly visible in nios data). I don't have such data as frajen86 to make activity filter analysis
But there are now those two trend changes (relative size of bronze has started to increase, relative size of gold has turned from increasing to slightly decreasing)
|
I was recently promoted to Platinum, but I'm not sure if it had anything to do with the change. I was #1 gold most of the season and was expecting a promotion soon.
|
Thanks korona, I added your data to the spreadsheet.
|
United States12212 Posts
Crossposted from Bnet forums: http://us.battle.net/sc2/en/forum/topic/16646674622
I've been saying since the ladder adjustment that it is important to consider the update a "clean slate" in terms of forward development. That is, they made the adjustment to the league boundaries, and now they're looking at the data to see who is playing the game from which league. A major flaw with using the "all accounts" parse from Nios is that it included pre-update data which is no longer relevant. If some Master player last played a game 51 days ago then we don't need to track him.
I ran a parse today through all tracked Nios.kr accounts and broke out only the accounts which had games played from 9 days ago or less (March 5 being 8 days ago but in the event time zones or date rounding played a role I expanded the coverage to 9 days). This is what it looks like:
NA Server All All % 3/5/2015+ 3/5/2015+% GM 198 0.22% 198 0.51% Master 796 0.86% 440 1.14% Diamond 7601 8.26% 4414 11.39% Plat 12165 13.22% 6252 16.13% Gold 35642 38.72% 16133 41.63% Silver 27588 29.97% 8478 21.88% Bronze 8049 8.75% 2834 7.31% Total 92039 38749
And here's the spreadsheet (man did it take a lot of work to whittle down the raw data to less than 2,000,000 cells, which is the limit for Google Sheets): https://docs.google.com/spreadsheets/d/1UzFJbOj5aE6ljRsWJ251t41tzaJrmM1q0kCR_FGCzaQ/edit?usp=sharing
Now, we already know that Nios doesn't track every account in existence, there are some holes here and there, but for the most part this is probably a decent parse. What we see here is that Master, Diamond, and Platinum are probably still a little underpopulated after the adjustment, some of the bloat from Silver transferred into Gold, and Bronze seems to have freed up a bit.
EDIT: Minor caveat, we don't know if the "x bonus weeks behind" activity filter also applies, which would change the distribution a little bit. I sent a PM to frajen86 asking for him to duplicate my parse with his own scraper (both to verify my findings and to apply that filter) since Nios doesn't include bonus pool information.
|
|
|
|