EPTLadder Deflation and MMR Decay - Page 4
| Forum Index > SC2 General |
|
guN-viCe
United States687 Posts
| ||
|
ZenithM
France15952 Posts
| ||
|
Plansix
United States60190 Posts
On September 24 2013 00:31 guN-viCe wrote: I've been master since WOL beta, up until about 2 seasons ago when I stopped playing. Now I'm in platinum, stomping noobs. It's strange.. Did you have to play new placement matches? I know that the ladder removes you if you take more than one season off, so it might have just put you as high as it can and you will have to slog your way back up. | ||
|
vaderseven
United States2556 Posts
I think the amount and speed of decay is wrong and needs reduced. Sometimes I take a week break and the first like 2-5 games will be a lose no matter the enemy. I would rather have no decay at that point so that after my hands regain their memory I am facing the right MMR. When it deflates to quickly then I lose a few as I get my chops back it just means I have to ladder like 20+ games to get back to a 50% win rate. I hate feeling like I have to fight my MMR back into place every couple weeks. | ||
|
thurst0n
United States611 Posts
I don't understand why you came to your conclusion in the "Does MMR Decay Result in Ladder Deflation?" Section. Just because half of platinum is potentially inactive by some chosen metric that doesn't deflate the ladder from my ignorant perspective. To me you take the state of the ladder and it is what it is. So if half of a league is inactive that's just how it is. I don't see how gold is "pushed down" because in theory half of them are also inactive anyway. And the ones that aren't are more consistent (in theory) when matched against an inactive platinum the match will be even since there definitely is an true skill decay after not playing, whether that be consistency or not knowing meta, forgetting some basic tech switch/scout whatever, or of course the infamous misclick. My guess is that blizz is working to find the correct fit for this decay so that matches remain as even as possible. If this hasn't been here all along then I suppose I follow your deflation idea (though couldn't you argue for inflation the other way since you're essentially manufacturing MMR #'s based on something that's not win/loss anymore. I guess I'm just thinking through it but I guess I don't agree that an inactive plat will beat an active gold more than 50% of the time. I think silver to plat is closer than people think. Diamond to Masters is a very large gap of course. Bronze to Gold also big gap. | ||
|
Mali__Slon
Senegal163 Posts
On September 24 2013 00:31 guN-viCe wrote: I've been master since WOL beta, up until about 2 seasons ago when I stopped playing. Now I'm in platinum, stomping noobs. It's strange.. Strange indeed. There was no master league in beta, not even diamond... diamond started when game launched, and master in 2011. | ||
|
tenklavir
Slovakia116 Posts
On September 24 2013 00:31 guN-viCe wrote: I've been master since WOL beta, up until about 2 seasons ago when I stopped playing. Now I'm in platinum, stomping noobs. It's strange.. I don't think that existed until the second major patch following release... | ||
|
InfCereal
Canada1759 Posts
On September 24 2013 00:42 tenklavir wrote: I don't think that existed until the second major patch following release... There wasn't even a diamond league in beta, if I recall correctly. | ||
|
MstrJinbo
United States1251 Posts
On September 24 2013 00:45 InfCereal wrote: There wasn't even a diamond league in beta, if I recall correctly. I seem to remember it came in a patch towards the end of beta. | ||
|
tenklavir
Slovakia116 Posts
On September 24 2013 00:46 MstrJinbo wrote: I seem to remember it came in a patch towards the end of beta. Correct, beta patch 13. Removed Copper, added Diamond. Still doesn't make a lot of sense... | ||
Excalibur_Z
United States12252 Posts
On September 24 2013 00:41 thurst0n wrote: I love reading anything you write Professor Z. I have one question I'm hoping you or someone else could expand on. I don't understand why you came to your conclusion in the "Does MMR Decay Result in Ladder Deflation?" Section. Just because half of platinum is potentially inactive by some chosen metric that doesn't deflate the ladder from my ignorant perspective. To me you take the state of the ladder and it is what it is. So if half of a league is inactive that's just how it is. I don't see how gold is "pushed down" because in theory half of them are also inactive anyway. And the ones that aren't are more consistent (in theory) when matched against an inactive platinum the match will be even since there definitely is an true skill decay after not playing, whether that be consistency or not knowing meta, forgetting some basic tech switch/scout whatever, or of course the infamous misclick. My guess is that blizz is working to find the correct fit for this decay so that matches remain as even as possible. If this hasn't been here all along then I suppose I follow your deflation idea (though couldn't you argue for inflation the other way since you're essentially manufacturing MMR #'s based on something that's not win/loss anymore. I guess I'm just thinking through it but I guess I don't agree that an inactive plat will beat an active gold more than 50% of the time. I think silver to plat is closer than people think. Diamond to Masters is a very large gap of course. Bronze to Gold also big gap. You may be right, and that section could be considered sensationalist I suppose. The decay system has been in since around the HotS launch as best we can determine. The percentage of activity in terms of bonus pool per league also doesn't necessarily prove deflation because, after all, we're fundamentally talking about inactive players here meaning they don't play often. The active players are the ones whose placement is actually relevant. If you play against an inactive guy who was previously a league or two higher, and he beats you, then as an isolated incident that's a problem. However, what if that guy only goes on to play three more games over the entire season? I think there are enough of these incidents that it creates a mental problem for a lot of players, because they can't reliably determine how good their current or future opponents are going to be. As for the MMR ranges per league, I believe the MMR-Stats guys are much less confident in the Bronze through Gold leagues because there simply aren't enough data points from users who play in those leagues. According to the tool, though, each league is estimated to cover almost the same numerical range. There's a bit of guessing involved because that's how the tool works (it takes guesses based on each game you play) so it may not be perfect, but it's probably okay for purposes like these. | ||
|
Ben...
Canada3485 Posts
| ||
|
hzflank
United Kingdom2991 Posts
On September 24 2013 00:36 Plansix wrote: Did you have to play new placement matches? I know that the ladder removes you if you take more than one season off, so it might have just put you as high as it can and you will have to slog your way back up. I have heard that many times, but is it actually true? I have had many long periods of only playing occasional team games with friends, but never had to play more than one 1v1 placement match. My MMR massively dropped, though. | ||
|
vesicular
United States1310 Posts
On September 24 2013 00:56 Excalibur_Z wrote: If you play against an inactive guy who was previously a league or two higher, and he beats you, then as an isolated incident that's a problem. However, what if that guy only goes on to play three more games over the entire season? I think there are enough of these incidents that it creates a mental problem for a lot of players, because they can't reliably determine how good their current or future opponents are going to be. I think this is the major issue, if any. The whole demotion of league is less of an issue since leagues are just a facade anyway. What matters is the matchmaking and matching of players of like skill (regardless of league). While decay of MMR during idle periods makes sense I think it needs to scale and I'm wondering if it does. For example, the higher your MMR the more the metagame has influence. Taking time off means missing out on game shifts and thus the decay makes more sense when you're up at GM or Master level. At a lower level, metagame shifts do not matter and it is simply mechanics that influence result, thus decay here makes less sense since a Gold player will almost always have better mechanics than a Bronze player even after a long period of inactivity for the Gold player. You don't tend to forget how to ride a bike. | ||
|
anessie
180 Posts
Ended previous season with about 1800 unused bonus pool. | ||
|
isospeedrix
United States215 Posts
| ||
|
Avos5
27 Posts
nice writeup | ||
|
Excalibur_Z
United States12252 Posts
On September 24 2013 02:20 vesicular wrote: I think this is the major issue, if any. The whole demotion of league is less of an issue since leagues are just a facade anyway. What matters is the matchmaking and matching of players of like skill (regardless of league). While decay of MMR during idle periods makes sense I think it needs to scale and I'm wondering if it does. For example, the higher your MMR the more the metagame has influence. Taking time off means missing out on game shifts and thus the decay makes more sense when you're up at GM or Master level. At a lower level, metagame shifts do not matter and it is simply mechanics that influence result, thus decay here makes less sense since a Gold player will almost always have better mechanics than a Bronze player even after a long period of inactivity for the Gold player. You don't tend to forget how to ride a bike. It "scales" in the sense that you lose more rating over time up to the max, but it's always the same amount whether you start at 2000 or 1000 (I believe, anyway). The idea behind it is that you come back rusty and the first person you face is someone you would have been favored to win before you went idle. Because in an Elo system the difference in rating between two players is what matters, then it makes sense that the decay rate is uniform for everyone. Let's say that a 100 rating gap represents a 10% win rate probability change, so 2000 vs 1900 is a 60%/40% pairing, but 1000 vs 900 is also a 60%/40% pairing. It wouldn't be fair for MMR to decay faster for higher-rated players so that they have a 70/30 pairing after a month off whereas a lower-ranked player would have a 60/40 pairing after the same time. | ||
|
CamoPillbox
Czech Republic229 Posts
| ||
|
Minkus
United States29 Posts
| ||
| ||
