StarCraft II Ladder Update -- March 2, 2015 - Page 6

On March 04 2015 00:54 opisska wrote:


hey, that's cool. It actually makes me feel better to see that while we are "just plat" in 2v2 EU (the mode I play the most), we are among top 12.5 percent of teams (which sould have be diamond under the target percntages).

Still doesn't mean anything, but the Blizzard feel-good-system needs to be fixed so that more people feel good

On March 04 2015 00:46 Cyro wrote:



They changed it a while ago, making gold something like 35% target IIRC

however actual distribution is miles away from their targets.

Since people can't be demoted, if getting top 2% MMR meant masters promotion, the result would be an inflation in masters because there were people there that didn't deserve to be there, and new people got promoted in as they hit top 2% MMR. Masters would go to ~3-4% (the extra 1.5x for example from people who should be demoted, but have not been) but instead it's at like ~0.8%. It's way off

On March 04 2015 00:48 JacobShock wrote:
This might be the most vague statement blizzard has ever released. What are you going to do I wonder, you are just pointing out issues.

On March 04 2015 00:02 virpi wrote:
For me as an on-off player who has been playing in master / diamond for most of his sc2 life, the possible changes sound good. I've picked up the game again a few weeks ago and found myself in gold league, which wasn't really where I belonged. Got back to diamond with a 90% win rate, but I ran into lots of players who were complaining about opponents of higher skill level. Once you've reached a certain level, you just don't fall below that anymore. Especially if your mechanics are kind of solid. I really hope they'll get rid of mmr decay and make the whole ladder more transparent.

On March 03 2015 13:57 Excalibur_Z wrote:


In the short term, and on an individual basis, decay is probably a fine thing. After all, you take a break, you come back rusty, you aren't going to be up to the task of playing against people that required your best game a month ago, so decay handles that for you. They ran the statistical models and found that after a month or more of inactivity, the average player's skill level falls to a point where opponents who were 75/25 matchups are now 50/50 matchups, so that's how it was tuned.

The ripple effects are far more damaging. Heart of the Swarm launched two years ago, that's two full years of decay and general rating deflation. Think about it: when you come back to play a game, you are much more likely to face an active player than another inactive one (obviously), and the outcome of that game impacts your opponent's rating positively or negatively. You are, on average, playing against a different tier of opponents and your games will push them up or down. Again, in the short term, that's probably okay because if you play only a few games your impact on the ladder around you is minimal, and if you play a lot of games you'll eventually climb back to where you used to be. But what happens if you decay again a few months later? And again a few months after that? That's exactly what's resulted in the ladder's current state, with mid- and high-level players decaying to the bottom and middle of the ladder after multiple seasons of inactivity.

In the immediate term, their fix is to adjust the league boundaries to match the target population percentiles, but that doesn't resolve the underlying issue. Each time they do this, it causes a league "crunch" where the rating ranges for the lower leagues become smaller and smaller (because if X rating range used to cover 8% of people and now it covers 20%, you need to find the new 8% intercept) while Diamond becomes a wide expanse collecting the chunks removed from the crunch, and that's untenable. Deflation and decay compounded over those two years have caused the league badges to mostly lose meaning, and that's important considering that's a player's only reference point.

On March 04 2015 04:21 KrazyTrumpet wrote:


Why do some people keep saying this? Did you even read the statement? They outlined areas of concerns, and steps they plan on taking immediately and in the future to try to resolve some of these issues. Frankly, this is the most specific statement we've ever heard from Blizzard on SC2 that wasn't a list of patch notes.

On March 04 2015 04:55 NonY wrote:

Has it ever been the case, or has it ever been considered, for inactivity to increase MMR volatility? It seems like the main complaint is players being "stuck" somewhere and having to play 20+ games, winning them all or losing them all, to get to where they belong. A pro player starting on a fresh account can expect at least 30 "free wins" before getting close to his true MMR. Maybe this was sensible when the game first came out, but it seems like for an ongoing system with players coming and going a lot, MMR volatility for "new" players (players returning from inactivity) should be much higher.

And I'm just curious, since I'm pretty certain that MMR did actually reset if the period of inactivity was long enough. Well, was that actually the case or am I misremembering? And with the removal of MMR decay, will that no longer be the case? Or is that form of MMR decay still going to exist?

On March 03 2015 13:57 Excalibur_Z wrote:


In the short term, and on an individual basis, decay is probably a fine thing. After all, you take a break, you come back rusty, you aren't going to be up to the task of playing against people that required your best game a month ago, so decay handles that for you. They ran the statistical models and found that after a month or more of inactivity, the average player's skill level falls to a point where opponents who were 75/25 matchups are now 50/50 matchups, so that's how it was tuned.

The ripple effects are far more damaging. Heart of the Swarm launched two years ago, that's two full years of decay and general rating deflation. Think about it: when you come back to play a game, you are much more likely to face an active player than another inactive one (obviously), and the outcome of that game impacts your opponent's rating positively or negatively. You are, on average, playing against a different tier of opponents and your games will push them up or down. Again, in the short term, that's probably okay because if you play only a few games your impact on the ladder around you is minimal, and if you play a lot of games you'll eventually climb back to where you used to be. But what happens if you decay again a few months later? And again a few months after that? That's exactly what's resulted in the ladder's current state, with mid- and high-level players decaying to the bottom and middle of the ladder after multiple seasons of inactivity.

In the immediate term, their fix is to adjust the league boundaries to match the target population percentiles, but that doesn't resolve the underlying issue. Each time they do this, it causes a league "crunch" where the rating ranges for the lower leagues become smaller and smaller (because if X rating range used to cover 8% of people and now it covers 20%, you need to find the new 8% intercept) while Diamond becomes a wide expanse collecting the chunks removed from the crunch, and that's untenable. Deflation and decay compounded over those two years have caused the league badges to mostly lose meaning, and that's important considering that's a player's only reference point.

While investigating these areas, we also took a careful look at player win-rates. When looking at players who are consistently active on the ladder, win-rates are right around 50%. However, players who are playing only a handful of sessions per season may be seeing less accurate matches, due to both a less measureable skill level,

On March 03 2015 23:21 SorrowShine wrote:
what is league Distribution atm?

Edit:
found it nvm
http://www.sc2ranks.com/stats/league

On March 04 2015 17:27 Rollora wrote:

Interesting that it is considered "Skill level" if you got 50% winrate.

I've seen enough one-tick ponys and cheesers to know that there is no skill involved in a lot of time. If the trick is discovered, there is no "skill" behind that.But I understand that the ladder cannot really be skillbased (you will not include something that reads build orders and their execution). However this 50% thingie leads to the simple fact that the "skill distribution" isn't accurate in the leagues up until diamond, since in lower leagues you just choose some bullshit and win your 50% until you get placed somewhere. Naturally the race with the most bullshit builds gets more high placements (because of these many one-trick-ponys) then the others.

Actually the only thing I wanted to say here: I don't see much correlation between skill and winrates in many if not most cases.

On March 04 2015 17:51 Pokebunny wrote:

This is not accurate, use nios.kr for better numbers.

On March 04 2015 18:45 opisska wrote:


Those numbers lead to very similiar conclusions, even if they differ by a couple of percent here and there.