|
as stated previously, it is only currently possible to determine your division's modifier if someone from your division makes the top 200. otherwise, you cant =\
|
as stated previously, it is only currently possible to determine your division's modifier if someone from your division makes the top 200. otherwise, you cant =\
Thanks for that quick answer!
Too bad that guy being 1st in my division isn't in the top 200..
|
If matchmaking is done by comparing hidden MMR ratings and point values are assigned after division modifiers are applied, can people estimate what their division tier is if they are matched up against someone in a known division tier?
For example:
Player A is a displayed 3000 point diamond player in a known S-class division.
Player B is also a 3000 point diamond player in an unknown class division.
They are matched against each other and with a displayed "Even Match."
Assuming both players have played a large number of games (500+) to where the uncertainty factor is quite small, can we infer Player B is also in an S-class division?
|
United States12240 Posts
On December 26 2010 06:58 DavasiaN wrote:
If matchmaking is done by comparing hidden MMR ratings and point values are assigned after division modifiers are applied, can people estimate what their division tier is if they are matched up against someone in a known division tier?
For example:
Player A is a displayed 3000 point diamond player in a known S-class division.
Player B is also a 3000 point diamond player in an unknown class division.
They are matched against each other and with a displayed "Even Match."
Assuming both players have played a large number of games (500+) to where the uncertainty factor is quite small, can we infer Player B is also in an S-class division?
It's quite risky to do this and that's why I haven't tried. MMR can theoretically swing 63 or more per game, and if your MMR is at least a tier ahead of or behind your points, you've already invalidated the test. According to Blizzard, sigma never actually gets small enough -- even after thousands of games -- for you to gauge your MMR, by design. If it did, it would suffer from the same problem that TrueSkill has in charting sharp rises or drops in skill, so they keep sigma large enough to remain flexible.
|
On December 26 2010 06:58 DavasiaN wrote:
If matchmaking is done by comparing hidden MMR ratings and point values are assigned after division modifiers are applied, can people estimate what their division tier is if they are matched up against someone in a known division tier?
For example:
Player A is a displayed 3000 point diamond player in a known S-class division.
Player B is also a 3000 point diamond player in an unknown class division.
They are matched against each other and with a displayed "Even Match."
Assuming both players have played a large number of games (500+) to where the uncertainty factor is quite small, can we infer Player B is also in an S-class division?
That makes sense, add to their points their rested points and it will be even better. But as the difference of one rank and another is of only 63 points, it might mislead you one or two ranks, also it's better confirming it through a lot of games and other players in that same division.
That could work, but again, we won't be able to verify it unless someone at that division appears on top 200. But for personal use, that might help.
It's not worth the effort in my opinion, as the reset is probably so close and we'll have the master leagues anyway.
|
On December 23 2010 01:09 Excalibur_Z wrote:Show nested quote +On December 22 2010 21:13 SDream wrote:On December 22 2010 20:51 Buruguduy wrote:On December 22 2010 20:46 Comma20 wrote:On December 22 2010 20:28 SDream wrote:
I got it from the beggining, my friend.
There are also other prerequisites that could be hindering him from getting on top 200. Therefore this division offset could still be F, E etc. That alone is not proof that it's a rank G.
Got it? Sorry, I'm being a little bullheaded. I think the word must is probably a bad choice. Doing some backwards reading, using SC2Ranks Graphs and the fact that my division -did- have 1 player (KingSphere) in the 7th of December Top200. His points value was that of an E Division compared to those above and below him. So I'm more curious about your thoughts. -IF- My division is an E ranked division what is stopping TheMuscle from being Top200 and what more importantly is stopping Me from being Top200. Thoughts? Unused bonus pool points from the lowest top 200 players? Definitely not related with unspent bonus points. It's probably related with their MMR moving average not being where Blizzard requires it to be in order to make them elegible for top 200. But we can't confirm it, it could be other unknown problem. Yeah I still think this is a possibility. If TheMuscle has an MMR moving average below the rest of the top 200 then maybe his points don't qualify him. I don't want to say definitively that that is the case, but we have seen prior anomalies that appear to point toward that conclusion.
Well, I play about 5 games on my SEA account, expend my bonus pool and then win them all usually, since my skill has improved from laddering on NA. I think there may be 'insufficient' data to either a) Be Active or b) Compare my moving average...
Thoughts?
|
Would I be asking too much if there could be an overly simple version of the posts you've made Excalibur? I love the posts but I feel I actually have to think a little when I skim over them  Great work btw.
|
awww my division aint even on there
Division Forge Sigma D:
|
I'm at 2456 at the moment. TheMuscle has been active and is at 2624. Perhaps we can see whether or not either of us make the Top200 this week, infer my division rating and then consider what would have stopped either of us from making the list.
|
On December 27 2010 23:47 Comma20 wrote:
I'm at 2456 at the moment. TheMuscle has been active and is at 2624. Perhaps we can see whether or not either of us make the Top200 this week, infer my division rating and then consider what would have stopped either of us from making the list.
u need over 2800 points to make top 200 this week
|
SeleCT 557-215 2811 Medic Mu O.O wow select!
|
|
|
Anybody can play top 200 players, doesn't mean a division is class S. Now I agree, looking at the players, this is probably an S-class division, but he's not going to add it until a player shows up in the top 200 so he can know for sure.
|
On December 28 2010 02:53 Pookie Monster wrote:Show nested quote +On December 27 2010 23:47 Comma20 wrote:
I'm at 2456 at the moment. TheMuscle has been active and is at 2624. Perhaps we can see whether or not either of us make the Top200 this week, infer my division rating and then consider what would have stopped either of us from making the list. u need over 2800 points to make top 200 this week
SEA Top200 is lower threshold due to the fact there are less player to feed points upward and less players over all.
|
Top 200 is up, get into work guys!!
There is more than 100 players with more than enough points to take the last place on top 200 LA, but they didn't make it, enough proof to me that MMR plays a role.
|
new top 200, I was 65th but I was around 50th on masters.... unless snapshot was earlier than I expected.
|
On December 29 2010 06:57 Pokebunny wrote:
new top 200, I was 65th but I was around 50th on masters.... unless snapshot was earlier than I expected.
Masters is more up to date, if you look at the top 200 your record was 281 - 199 when the snapshot was taken, while on sc2ranks you're 290 - 203. The difference between 50th and 65th in masters is ~60 points, so being higher on Masters is to be expected since you played 13 games after the snapshot was taken.
|
Wow, no zergs in the top 10 and only 4 in the top 30.
|
Really nice job for figuring out the 63 point modifier, goes to show how difficult it is to hide rating systems from a subgroup with disproportionately higher IQ when compared to that of the general populace.
Here are some questions though, is the top 200 purely rating based? According to your data in the vast majority of circumstances the main determining factor of division rank number of games was the most significant.
Some new information that I seem to have noticed that i'm unsure if your aware of is that if one were to check their opponents match history of the point exchange, independent of bonus pool if you won and gained 8 points, you should be able to check your opponents point change and instead of losing 8 points they lost 6 points.
If this is universal across the board and not a calculated occurrence based on the current MMR calculation this could shed some light on why more games = more rating. As even if your percentage was a flat 50%, you would still gain an average of 2 points than lost, including bonus pool inflation.
Can anyone confirm/deny this observation?
|
On December 29 2010 09:17 NearPerfection wrote:
Really nice job for figuring out the 63 point modifier, goes to show how difficult it is to hide rating systems from a subgroup with disproportionately higher IQ when compared to that of the general populace.
Here are some questions though, is the top 200 purely rating based? According to your data in the vast majority of circumstances the main determining factor of division rank number of games was the most significant.
This is in the OP, and on sc2ranks, it seems to be 100% points - division modifier with 100% prediction rate for known division modifiers (many division modifiers are unknown).
Some new information that I seem to have noticed that i'm unsure if your aware of is that if one were to check their opponents match history of the point exchange, independent of bonus pool if you won and gained 8 points, you should be able to check your opponents point change and instead of losing 8 points they lost 6 points.
If this is universal across the board and not a calculated occurrence based on the current MMR calculation this could shed some light on why more games = more rating. As even if your percentage was a flat 50%, you would still gain an average of 2 points than lost, including bonus pool inflation.
Can anyone confirm/deny this observation?
You do not net gain points from playing games, look at how many points your opponent lost when you win, it will always exactly equal the pre-bonus points you gained.
|
|
|
|
|
|
EPT
