EPTRanked Ladder Activity Analysis - Page 2
| Forum Index > SC2 General |
|
WGT-Baal
France3503 Posts
| ||
|
frajen86
168 Posts
On August 26 2014 17:28 LoneYoShi wrote: Interesting read, I'm eager to see EU and KR ! EU is up in the OP - very similar percentages to the America servers. It's really surprising how small Bronze and Silver leagues are for active players. Running KR analysis now. | ||
|
frajen86
168 Posts
Not surprising, some key percentages are different, just like last season. Some thoughts * In general, the more you play 1v1 ranked, the more likely you are going to be promoted above bronze and silver league : ) * The KR 1v1 playerbase is ~5% more "active" than AM/EU * Blizzard-stated league distributions are still skewed for active players across all regions (bronze/silver undersized, gold-diamond oversized) * Protoss/Terran more popular in KR than AM/EU * Zerg less popular in KR than AM/EU * 1v1 ranked more popular in KR than ranked team games (compared to AM/EU) * 4v4 isn't as popular in KR compared to AM/EU * Ranked team ladder distributions still make no sense to me More conclusions to come later, I will try to run this again next week or two. At least one more time in the next month. Cheers | ||
|
purakushi
United States3302 Posts
| ||
|
nkr
Sweden5451 Posts
 | ||
|
Irri
Sweden41 Posts
| ||
|
Tyrhanius
France947 Posts
You can be placed in bronze league even if you master/diamand 1vs1. It's really long to get promoted, even if you crush your opponent, sometimes you can 1vs2 players because you're diamand and they're silver. And for this low league players, who don't want to hard try in 1vs1, it's not really funny to get matched vs way too strong player for them. | ||
|
frajen86
168 Posts
On August 30 2014 20:10 Tyrhanius wrote: I really think the matchmaking for teamgame should be rework and take count of the 1vs1 MMR (at the placement at least). You can be placed in bronze league even if you master/diamand 1vs1. It's really long to get promoted, even if you crush your opponent, sometimes you can 1vs2 players because you're diamand and they're silver. And for this low league players, who don't want to hard try in 1vs1, it's not really funny to get matched vs way too strong player for them. One thing I can try to do for the next analysis round is look at the players who are in both 1v1 ranked and team types and compare what leagues they are in. | ||
|
frajen86
168 Posts
In general, Terran is becoming more popular, Zerg is becoming less popular, and fewer people are playing the game compared to late August. Compare the # of people that played games last month in the previous analysis to the current one - it's about a 20% drop across all regions. Ladder is still skewed based on "active" players - bronze/silver greatly undersized, gold+ oversized. I am wondering if Blizzard is no longer using bonus pool remaining to develop MMR boundaries for 1v1 ladder. AM still plays the most team games, and KR still remains the most active server. Also, the % of players that have 1 week or less of bonus pool has shrunk across all regions as well, currently less than 10% for AM and EU servers. I am running the "league" comparison asked for in the above replies, but it is kind of slow ATM. I need to optimize the code more. | ||
|
frajen86
168 Posts
I added a new section, a cross-reference table between a player's 1v1 league and that same player's highest league for a team type. Not surprisingly, a healthy percentage of any team league will include players that are in a higher 1v1 league. Compared to the last reporting period (mid/late September): * In general, activity across all team games dropped, although for some team game types in EU and AM, the decrease is small (~10% or less, in some cases 1%) * In EU, the number of players that only play team games (but not 1v1) went up by 30%, compared to the number of 1v1 placed players, which went up by 22% * For AM and KR, the number of 1v1 and team-only players increased at roughly the same rate (~15% in KR, ~18% in AM) * In KR, the number of 1v1 games played in the last week/month dropped by about 50% (suggesting a relatively major drop in player activity for 1v1 games for the last month) * In AM/KR, the drop in number of 1v1 games played in the last week dropped about 6%, but the number of 1v1 games played in the last month increased by 13% in EU and 9% in AM (suggesting an overall uptick in player activity for the last month) Ladder distribution for active players (based on bonus pool) still appears to be skewed: Bronze and Silver very undersized, Gold/Platinum oversized, Diamond ~correct, Master a little oversized. When not considering player activity, the Blizzard targets appear to be closer for Bronze-Gold (oversizing Bronze and Silver league a bit), while Platinum-Master appear to be too small. However, taking into account player activity showed a much more accurate match to Blizzard targets at the end of season 3 last year. * In all regions, in Bronze league, Terran is the most popular race by far. * In AM and EU, as players advance in leagues, Zerg becomes more popular while Terran becomes less popular. Protoss stays about the same, with one exception. In Master league, for the least active players, Protoss actually jumps about 5% (and Random drops 5%). * In KR, as players advance in leagues, Zerg becomes more popular while Terran becomes less popular. That being said, Zerg remains the least popular race in KR for active players at all leagues. * In KR, for active players, around Diamond league, Terran begins to grow more popular, and it's interesting to note that in Master league, Terran is by far the most popular race in KR (whereas Zerg is the most popular race in AM/EU). * In KR, for inactive players, Zerg actually remains the most popular race in Diamond and Master. edit: correction on activity for team games | ||
|
frajen86
168 Posts
Compared to the last reporting period (beginning of November/week before Blizzcon): * As this season is relatively young (only ~6 weeks have passed), the number of inactive players is not high compared to last report. One can conclude that the longer a season goes, the more likely players will have more bonus pool accumulated. Conversely the % of players with less than 1 week of bonus pool is relatively high (~15%) * In general, activity across team games stayed the same * For all regions, the number of 1v1 games played in the last week increased. In KR the increase was huge, by 196% percent (almost double). Again this includes 2 days of season 3 ladder. * In EU there was a 7% increase and in AM there was a 21% increase. This was the last week prior to the "dreampool" being reversed. * In KR, there was a 245% increase in number of 1v1 games played in the last month. In EU there was a 10% decrease and in AM there was a 2% decrease (negligible). The "last month" timeframe is essentially all of season 4 with the dreampool. * Overall I believe player activity greatly increased in KR this season, while it stayed roughly the same in AM and EU. It's hard to say whether the dreampool being reverted greatly impacted the huge KR rise in activity, because at the time of the analysis there was only 2/3 days of season 3 ladder maps. 1v1 ladder distribution for active players (based on bonus pool) still appears to be skewed: Bronze and Silver undersized, Gold very oversized, Platinum/Diamond/Master close to target. In AM and KR regions, Diamond is a undersized. Ignoring bonus pool as activity and using the total playerbase, Bronze appears to be on target while Silver and Gold are oversized. Meanwhile, Platinum/Diamond/Master are all undersized. * In all regions, in Bronze league, Terran is the most popular race by far. Terran also appears to be more popular in higher leagues for the AM region. * In AM and EU, as players advance in leagues, in general Zerg becomes more popular while Terran becomes less popular. * On AM, Terran actually jumps up in percentage from Platinum to Diamond league for active players, and jumps up in percentage from Platinum to Master for the least active players. * In EU, the transition from Zerg being less popular than Terran happens from Gold to Platinum league. * In AM and EU, Protoss stays about the same across all leagues/activity levels, with one exception. In EU region Master league, for the least active players, Protoss actually jumps about 5% (and Random drops 5%). * In KR, Terran is the most popular race for active players in all leagues, except for Gold league where it is slightly edged out by Protoss. Zerg grows more popular as active players reach higher leagues. Protoss drops after Gold league from about 32% to 27% and stays around there into Master. * In KR, for inactive players, Terran is by far the most popular race in Bronze, Diamond, and Master league. If you're interested in the race distribution and player activity please see the original post with the updated charts. | ||
|
RedHoodlum
United States9 Posts
| ||
|
frajen86
168 Posts
| ||
|
frajen86
168 Posts
Since there's a lot of discussion I wanted to put in my 2 cents. This is just for AM (America) region. Bonus pool remaining shows roughly the same story. Bronze is undersized, Silver is very undersized, Gold is very oversized, Platinum is a little oversized, Diamond is a little oversized, Master is undersized. ![[image loading]](http://19hz.info/Starcraft/Historical%20Data/AM/2015-03-08%20-%20Bonus%20Pool%20Remaining.png) About 11% of players have 1 or less bonus pool week remaining. ~20% with 3 or less. Majority of 1v1 placed players (~60%) have 7+ weeks remaining. As mentioned elsewhere, adjusting the boundaries doesn't fix the problem long-term because of MMR decay. This was mentioned in Psione's post so that's good. Anyways, even if 50-100 players were promoted from Diamond to Master (to match the 2% target), Diamond would still need to "lose" about 300-350 more players to hit its 18% target. That means in-season demotions (from Diamond to Plat). But Platinum is already correctly sized, so the resultant cascading effect would mean a lot of in-season demotions across all leagues (except Bronze), something that Blizzard is probably not too interested in having happen. edit: http://www.teamliquid.net/forum/viewpost.php?post_id=23870327 korona says: "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." which essentially means at the start of next season a lot of people are going to get demoted | ||
|
ZackAttack
United States884 Posts
On March 09 2015 05:59 frajen86 wrote: After this last update was announced, I decided to just get more data and hold off on the February analysis. Since there's a lot of discussion I wanted to put in my 2 cents. This is just for AM (America) region. Bonus pool remaining shows roughly the same story. Bronze is undersized, Silver is very undersized, Gold is very oversized, Platinum is a little oversized, Diamond is a little oversized, Master is undersized. About 11% of players have 1 or less bonus pool week remaining. ~20% with 3 or less. Majority of 1v1 placed players (~60%) have 7+ weeks remaining. As mentioned elsewhere, adjusting the boundaries doesn't fix the problem long-term because of MMR decay. This was mentioned in Psione's post so that's good. Anyways, even if 50-100 players were promoted from Diamond to Master (to match the 2% target), Diamond would still need to "lose" about 300-350 more players to hit its 18% target. That means in-season demotions (from Diamond to Plat). But Platinum is already correctly sized, so the resultant cascading effect would mean a lot of in-season demotions across all leagues (except Bronze), something that Blizzard is probably not too interested in having happen. This would suggest that maybe the patch did work and maybe people aren't going to get all their promotions like they thought. My question would be how is this different from the league changes last year. Apparently there was a huge amount of promotion the last time they did this and things on nois.kr changed quickly. Is this year's problem because of inactivity? | ||
|
frajen86
168 Posts
On March 09 2015 08:01 ZackAttack wrote: This would suggest that maybe the patch did work and maybe people aren't going to get all their promotions like they thought. My question would be how is this different from the league changes last year. Apparently there was a huge amount of promotion the last time they did this and things on nois.kr changed quickly. Is this year's problem because of inactivity? Well... notice that these distributions look similar to how they looked towards the end of last year. I don't see much change from last December to now. Also, I don't know exactly how this "patch" was implemented. Personally I don't have any idea how they planned on fixing things; it seemed impossible to me w/out forced demotions. As far as I can tell nothing really changed. IMO massive promotions wouldn't fix anything; I think that's the wrong kind of perspective to have given this data. Of course I am assuming the activity metric for players is still closely tied to bonus pool (I see no reason why it would have changed over the last 2 years though). When I get done with this batch I'll run the February data (a month ago) and do that comparison; maybe there was a change from last month to now. About "the problem" - promoting people in Gold+ doesn't actually fix the distribution. 1) There's a small pool of Diamond league players that might belong in Master (do-able but seems like it hasn't happened yet). 2) Other than that, people need to get demoted to fix Bronze/Silver (do-able, but hundreds/thousands of people would need to be demoted, however even if they changed the MMR boundaries the player base wouldn't really notice until next season's initial placements) 3) Or there needs to be an influx of new players with MMR skill levels in Bronze/Silver (unlikely to happen at this point) 4) MMR decay is being adjusted next season. korona/Excalibur_Z can explain the effects of that way better than me more thoughts: ******* As far as the changes from January 2014. Unfortunately the first time I did this kind of analysis was in June last year, so I don't know what happened between those months. But essentially, distributions used to be relatively accurate in August 2013, with Master being a bit large. Shifting the MMR boundaries for leagues might have "fixed" things in the short-term. However nowadays, I would guess that new players are ending up in Gold league "incorrectly" - due to the MMR range for Bronze/Silver being too low. Also, I bet MMR decay has caused some higher league players to drop into Gold as well. So over time, you end up with Gold league being grossly inflated while Bronze/Silver stays small. In theory, I think this would have been avoided if a lot of new players kept coming in - to fill out the lower leagues - but maybe there haven't been enough new players to do this. http://www.teamliquid.net/forum/starcraft-2/423477-ladder-analysis-activity-metric August 2013 (Season 4 lock): Leagues seem to be closer to target distributions for active players. Master is actually a bit large, Diamond is a bit undersized, Gold is a bit small, Bronze a bit large. http://www.teamliquid.net/forum/viewpost.php?post_id=22415099 June 2014, ~5 months after the patch: Bronze and Silver leagues are both undersized, while Gold and above are all oversized. The effect of the promotions is clear. August to November 2014: pretty much the same thing. However, Master league starts to "shrink" while Plat/Diamond begin to stabilize. December 2014: Master league (arguably) drops below its target. First time it's undersized (barely). Meanwhile, Bronze/Silver leagues remain undersized. ******* | ||
|
HewTheTitan
Canada331 Posts
The leagues in sc2 are far too small... it would be like having the Tetris levels be 3x the size, which rewards improvement a lot less frequently... | ||
| ||
