| Forum Index > SC2 General |
|
Lysenko
Iceland2128 Posts
On January 24 2012 13:27 perser84 wrote: can somebody tell me how much i must win or how far/close i am from master leauge ? Based on your score and where you fall in the population, I'd guess you could get promoted at any time. But, it's not possible to be absolutely certain. | ||
|
jjhchsc2
Korea (South)2393 Posts
I think i just had one but i don't understand the logic behind it. I was ranked 3rd in my gold division with around 30 BP and about 80 points behind number 1 yesterday. Now today my rank went down to 6th with around 90 points behind number 1. But with 230 BP. BOOM PLUS 200 BP Any reasoning behind this? o.o | ||
Excalibur_Z
United States12235 Posts
On January 25 2012 14:31 jjhchsc2 wrote: have any of yous had massive bonus pool fluctuations? I think i just had one but i don't understand the logic behind it. I was ranked 3rd in my gold division with around 30 BP and about 80 points behind number 1 yesterday. Now today my rank went down to 6th with around 90 points behind number 1. But with 230 BP. BOOM PLUS 200 BP Any reasoning behind this? o.o Can you actually spend that extra bonus pool, or has that display bug reappeared? Does your bonus pool total return to normal after logging out? And, are you looking at the website or in-game? | ||
|
jjhchsc2
Korea (South)2393 Posts
On January 25 2012 15:53 Excalibur_Z wrote: Can you actually spend that extra bonus pool, or has that display bug reappeared? Does your bonus pool total return to normal after logging out? And, are you looking at the website or in-game? lol what the hell it returned to normal after logging out. damn got excited and happy for nothing. Thanks for the reply ^.^ | ||
|
tpmraven
United States833 Posts
On January 21 2012 06:01 Mendelfist wrote: I've been thinking about this. I haven't seen any obvious league threshold adjustments. I may have missed some, but I don't think that's what's going on. A matchmaking system needs some sort of "anchor". Otherwise the MMRs of the entire population could drift up or down without control. Blizzard decided to implement this anchor as an MMR cap. In other words they use the best players on the ladder as anchors. They play a lot and their skill probably doesn't change quickly, so they are good for this. But as you remember from our previous discussion, the MMR difference between two players is directly related to their skill difference. New players are still newbies, but the best players get better and better. The skill range gets stretched out over time, and so must the MMR range. If the best players MMRs are fixed by the MMR cap, it will push all the other players downwards. That's what's happening. I think this implementation is a terribly bad idea. It is very discouraging. The MMR range should be anchored at the bottom, not at the top. Blizzard could then add new leagues at the top as masters fill up more and more, and that would much more accurately reflect the actual skill change of the population. An effect of the current system is that the downwards push of the population counters the fact that many bronze players stop playing. Blizzard probably thinks this is convenient. I think it's just bad. Being demoted despite practicing and getting better is not an incentive to keep playing. But all this is just my theory of how it works, of course. I dont think this is how it works. I did not play more than a few games for about 4 months, because i played a few games each season my mmr stayed the same and i stayed in the same league. i just came back at the end of last season, and started playing like crazy again, going from high gold to low diamond, in one season. Before i stopped playing i struggled to be top 8 in gold, hardly able to keep a 50% win ratio, i had been in gold since season 2. when i came back i was STOMPING high golds, and plats left and right. It seemed to me like the competition in the leagues had been watered down. | ||
|
Mendelfist
Sweden356 Posts
On January 25 2012 16:46 tpmraven wrote: Before i stopped playing i struggled to be top 8 in gold, hardly able to keep a 50% win ratio, i had been in gold since season 2. when i came back i was STOMPING high golds, and plats left and right. It seemed to me like the competition in the leagues had been watered down. I think most people would disagree with you. The general consensus seems to be that all leagues have become a lot harder since the beginning. I just watched a few games that I played a year ago, and they were quite ridiculous. No expansions, no wall-ins. Make a few units when you feel like it and attack. That was enough for gold back then. It isn't now. The difference is dramatic. | ||
|
Let it Raine
Canada1245 Posts
| ||
|
Cornstarched
Canada74 Posts
| ||
|
Lysenko
Iceland2128 Posts
On January 21 2012 06:01 Mendelfist wrote: A matchmaking system needs some sort of "anchor". Otherwise the MMRs of the entire population could drift up or down without control. Blizzard decided to implement this anchor as an MMR cap. In other words they use the best players on the ladder as anchors. They play a lot and their skill probably doesn't change quickly, so they are good for this. But as you remember from our previous discussion, the MMR difference between two players is directly related to their skill difference. New players are still newbies, but the best players get better and better. The skill range gets stretched out over time, and so must the MMR range. If the best players MMRs are fixed by the MMR cap, it will push all the other players downwards. That's what's happening. A couple thoughts: First, MMR won't have population-wide inflation or deflation issues if the MMR adjustment after each game is zero-sum. So, if a better player plays a lower-MMR player, and the lower-MMR player wins, the better player's MMR goes down by 20 and the lower player's MMR goes up by 20. The average MMR doesn't move if one player's increases by X and another's decreases by X, at the same time, every time. Second: an MMR cap serves a different purpose, in that the very best players may be statistical outliers so far beyond everyone else that they have difficulty finding a match within whatever the usual threshold is. However, a player's MMR might be capped for purposes of the match-finding process without actually placing a hard limit on the player's MMR number such that they'd receive 0 points for winning any game regardless of the quality of their opponent. A complete, hard cap like that would probably create issues, but I don't know that we can tell whether that's exactly what Blizzard is doing, and if they are, they may well have a way to compensate for it. | ||
|
Lysenko
Iceland2128 Posts
On January 25 2012 16:46 tpmraven wrote: Before i stopped playing i struggled to be top 8 in gold, hardly able to keep a 50% win ratio, i had been in gold since season 2. when i came back i was STOMPING high golds, and plats left and right. It seemed to me like the competition in the leagues had been watered down. Aside from adjustments to MMR thresholds for leagues, which could account for this, I have the feeling that play styles in the middle leagues have shifted toward less aggressive, longer-term games over the last few seasons, and someone who bucks that trend might find their opponents often unprepared because they're seeing strongly aggressive play less on the ladder. The players might not be worse than they were before, but they may be playing differently in a way that's more vulnerable to what you're used to doing. | ||
|
Mendelfist
Sweden356 Posts
On January 27 2012 13:52 Lysenko wrote: First, MMR won't have population-wide inflation or deflation issues if the MMR adjustment after each game is zero-sum. A zero sum system sounds unlikely. How would it handle a game where one player is playing placement matches against an old player with a very low sigma? It does not feel reasonable to make large changes to the stable players MMR when he is playing against someone with an unknown skill. | ||
|
Lysenko
Iceland2128 Posts
(I say that based on league placements of my opponents after winning and losing streaks. Incidentally, it feels like they've chosen to let MMR fluctuate more and just make promotion/demotion thresholds harder to achieve rather than design a system where MMR is more stable and the promotion/demotion thresholds are closer together for each league boundary.) Also, it may be possible to avoid inflation or deflation with a system that's designed to be zero sum in the aggregate without necessarily being so in an individual game. I'm not entirely certain how one might put that together though. | ||
|
Excalibur_Z
United States12235 Posts
On January 27 2012 15:37 Mendelfist wrote: A zero sum system sounds unlikely. How would it handle a game where one player is playing placement matches against an old player with a very low sigma? It does not feel reasonable to make large changes to the stable players MMR when he is playing against someone with an unknown skill. I think Placement matches just have to be treated somewhat differently because those are the only matches where it's possible to be placed anywhere from Diamond to Bronze in only 5 games, and in a good portion of those matches you're most likely playing against other Placement players as well. It's very difficult to gauge Placement match data because those are going to constitute a minority of most players' games, but from what I've seen it's uncommon that Placement players are matched against very stable players (that's not to say it's impossible though). I'm reasonably confident that MMR is zero sum for regular games though. | ||
|
Mendelfist
Sweden356 Posts
1: There is an upper MMR cap. 2: The MMR difference between two players is directly related to their skill difference. 3: The increasing skill difference between beginners and the best players causes no downwards MMR push of the lower leagues. In one or more of those three you would have to make exceptions. Or there are some tricks that I can't think of. I also feel like there would be some problems with an MMR cap and a zero sum system, but I haven't thought it through. | ||
|
Lysenko
Iceland2128 Posts
So, player A has MMR of 9001, but the soft cap is 3000, so when player A hits the find match button they're matched with another player as though their MMR were 3000, but point awards affect the uncapped MMR of 9001. The issues with grandmaster league might have been an implementation mistake with a soft cap like that rather than a hard cap that affects the actual stored MMR number. | ||
|
Mendelfist
Sweden356 Posts
| ||
|
Lysenko
Iceland2128 Posts
On January 28 2012 04:50 Mendelfist wrote: Isn't it common at the very top that all players see themselves as favored regardless of their opponents, and isn't this an indication of a hard cap? Not necessarily. If a capped MMR were used for matchmaking only, that display might well use that number, while the actual MMR adjustment taking place behind the scenes could be correct. Also, such a soft cap would greatly increase the likelihood that top players would be matched against other players vs. whom they'd legitimately have a "favored" display, because the purpose of such a cap is explicitly to match them against much weaker players so that they can find games when they hit the button without an undue wait. | ||
|
Excalibur_Z
United States12235 Posts
On January 28 2012 04:50 Mendelfist wrote: Isn't it common at the very top that all players see themselves as favored regardless of their opponents, and isn't this an indication of a hard cap? Yes I don't believe there is a matchmaking cap in place because David Kim has indicated to top players that there is a cap, and the behavior at the top is just as you say which is expected from a hard cap. | ||
|
Excalibur_Z
United States12235 Posts
On January 28 2012 05:21 Lysenko wrote: Not necessarily. If a capped MMR were used for matchmaking only, that display might well use that number, while the actual MMR adjustment taking place behind the scenes could be correct. Also, such a soft cap would greatly increase the likelihood that top players would be matched against other players vs. whom they'd legitimately have a "favored" display, because the purpose of such a cap is explicitly to match them against much weaker players so that they can find games when they hit the button without an undue wait. So for the purposes of matchmaking and point determination they have a capped MMR, but what other reason would there be for MMR to actually go beyond the cap? I suppose maybe acting as a buffer so that if you slip a little you still stay above the cap, but that sounds kind of weird to me. | ||
|
Lysenko
Iceland2128 Posts
On January 28 2012 05:25 Excalibur_Z wrote: So for the purposes of matchmaking and point determination they have a capped MMR, but what other reason would there be for MMR to actually go beyond the cap? I suppose maybe acting as a buffer so that if you slip a little you still stay above the cap, but that sounds kind of weird to me. I'm sorry -- I was describing a case in which matchmaking only would be affected by a capped MMR, points and MMR adjustment would not be, and the "favored" display were a mistake caused by using the capped rather than uncapped MMR only when displaying the loading screen. (Note that the bug people have intermittently mentioned where the loading screen doesn't agree with the score screen on who's favored suggests there is a different code path for that loading screen's display.) Even if post-game points do use the capped MMR (which should be easy to tell, and it sounds like you're sure that's the case) maintaining a behind-the-scenes uncapped MMR and adjusting the hidden MMR based on that number would prevent the problems that Mendelfist is describing with the entire population shifting around as a result of the cap. It also does occur to me, though, that if there were a hard cap at the top and a hard floor at the bottom (to fix the problem where people at the bottom can't earn any points) those might balance each other off to fix any tendency for the population to drift. It just seems to me that using a cap only for matchmaking solves the game-finding problem for top players without introducing all these potential issues a hard cap might. If they went with a hard cap, I'm not sure why, but then there might be implications one might not be able to see without looking at the implementation or the actual math behind the system. P.S. To hell with asking David Kim about carriers. I want to get the guy who designed this system in a room with a whiteboard and the freedom to talk freely. | ||
| ||
|
|
Wardi Open
Replay Cast
Replay Cast
RSL Revival
Cure vs Percival
ByuN vs Spirit
PiGosaur Monday
RSL Revival
herO vs sOs
Zoun vs Clem
Replay Cast
The PondCast
RSL Revival
Serral vs SHIN
Solar vs Cham
Replay Cast
[ Show More ] |
|
|
EPT
