|
United States12240 Posts
On November 20 2010 08:57 Lysenko wrote:Show nested quote +On November 19 2010 23:26 KaienFEMC wrote:
Do you think there is another factor involved in the generation of top 200 other than the division modifier? Do you think MMR would possibly play a role in here? Though I think you stated that Blizzard had announced that MMR has nothing to do with top 200. This is making things really interesting. Please correct me if I had misinterpreted your analysis.
Again, thanks for the awesome work! Here's a question: Are we certain that there really an "MMR" other than points minus division modifier? Is it possible that Blizzard subtracts the division modifier from displayed point value when deciding against whom to match a player, rather than maintaining a totally separate MMR number? Perhaps there's an error estimate ("sigma" in Trueskill-speak) that's separate, but which estimates error in the displayed-point-minus-division-modifier number.
There's definitely a hidden MMR, and this is proven by matchmaking alone (Bronze being matched against Diamond, for example). If this were just a result of a large sigma then you would be just as likely to be matched against very low level players, which we know is not the case.
|
United States12240 Posts
On November 20 2010 09:07 Kiarip wrote:
I have seen situations where people that have less points are ranked higher than people that have more points in their same division. Can you explain that?
Unless they're a significant outlier like that #200 guy in the TW Top 200 or InSTinK from Aug 10's NA Top 200 (which will be very very rare), then this can be explained by points changing since the time of the Top 200 snapshot. Only points at the time of the snapshot matter, and in this case that was 11/15 in the late morning.
Find out for yourself by tracking the match histories of each player and adjusting their points back to that time period.
|
I was under the impression everything the OP said was actually not true. If it is, than damn Blizzard's ladder is even more f'ed up then most people realized.
btw even if all divisions are not equal, theres still no reason for a division modifier since your division doesn't affect your matchups or points... right?
but I could have sworn that blues had posted in their forum back in beta that divisions actually were random and were not based on player skill whatsoever... I don't think people can say anything is confirmed since Blizzard always confirms different things, if everything, even conflicting ideas, is confirmed its the same as if nothing is confirmed.
|
Catyoul
France2377 Posts
On November 19 2010 16:18 Largeman wrote:Occam's razor suggest that this is false. More than likely the reason there are multiple division with less than 100 people is because blizzard doesn't want divisons to be rediculously small. They likely have a minimum number of players per division rule, example: "All divisions must have a minimum number of 70. If there are not enough players to reach this minimum then players will be drawn from other divisions to meet the requirement. If there were 120 players needed to be placed, then it makes sense to have two divisions of 60 than 1 divison of 100 and one of 20. If there are 204 players that need to be placed, it makes more sense to have three divisions of 68 than 100, 100, 4, or 100, 52, 52. In the case of the above example, where 204 players need to be placed we are 6 members short, the system would draw from a full division of 100 players so the #s would likely be something 75 players, split evenly among 4 divisions.
To add to the counter arguments of Wargizmo and Bosu, I have been promoted into almost empty divisions (maybe 4 people or so) with some of my 2v2 partners.
Since the question of how Blizzard decides to promote which people into which division rank, I offer a conjecture that only works up to 6 ranks. It might be based on the number of games won/lost during placement matches : 5-0 = S rank, 4-1 = A rank, etc. Why not I guess. It is safe to assume people who are currently in the top 200 mostly went 5-0. Some might have voluntarily gone 0-5 for fun (?), some might have lost one game to some cheese and gone 4-1. The 3-2 and 2-3 results would seem more like players who didn't know the game and really improved up to the top 200 (less likely ?).
It offers the benefit of being somewhat disprovable, if we trust the people to remember correctly the result of their placement matches.
|
On November 20 2010 09:27 Excalibur_Z wrote:
There's definitely a hidden MMR, and this is proven by matchmaking alone (Bronze being matched against Diamond, for example). If this were just a result of a large sigma then you would be just as likely to be matched against very low level players, which we know is not the case.
Ah, gotcha, so another case that establishes the existence of a hidden MMR is the player who gets promoted to Platinum after 30 games in Bronze, for example, since 30 games would not be nearly enough to get their points high enough for such a promotion if the displayed rating minus modifier were all that were being used.
|
On November 20 2010 09:33 Excalibur_Z wrote:Show nested quote +On November 20 2010 09:07 Kiarip wrote:
I have seen situations where people that have less points are ranked higher than people that have more points in their same division. Can you explain that? Unless they're a significant outlier like that #200 guy in the TW Top 200 or InSTinK from Aug 10's NA Top 200 (which will be very very rare), then this can be explained by points changing since the time of the Top 200 snapshot. Only points at the time of the snapshot matter, and in this case that was 11/15 in the late morning. Find out for yourself by tracking the match histories of each player and adjusting their points back to that time period.
I remember this being the case some time ago back when Fenix was #1, because I also hypothesized that different players of different divisions all just got some kind of bonus/multiplier based on the difficulty of their division, but I remember finding an extreme example where this didn't hold true, and sc2ranks match history also confirmed that.
maybe it could have had something to do with how many points they have in the bonus pool? I don't know. I'll try to find an example of this in the current top 200
|
So i just got promoted and I lost exactly 126 points when getting promoted into the new league. The new league i am in is Boros Gravity
Wow, maybe it's not a coincidence after all.
|
On November 20 2010 09:34 Disastorm wrote:
I was under the impression everything the OP said was actually not true. If it is, than damn Blizzard's ladder is even more f'ed up then most people realized.
btw even if all divisions are not equal, theres still no reason for a division modifier since your division doesn't affect your matchups or points... right?
but I could have sworn that blues had posted in their forum back in beta that divisions actually were random and were not based on player skill whatsoever... I don't think people can say anything is confirmed since Blizzard always confirms different things, if everything, even conflicting ideas, is confirmed its the same as if nothing is confirmed.
I believe my division (Scourge Rho) is overwhelming evidence of the skill level per division. As I've said before, it is only low amount of games played with high win % (meaning very high MMR at the time of promotion). There is one outlier who went on a massive losing streak after being promoted (as far as I can tell through his match history).
I've been watching my division, people are being added very slowly, and the trend of low games played/high win% is continuing. The division is clearly waiting for certain players to meet the requirement, since it's been 4-5+ days since my division has been created and it only has ~20 players.
|
This all makes a lot of sense, and explains a lot of the weird things about the ladder.
Great work guys!
|
Page up: http://sc2ranks.com/masters
It's not linked anywhere else right now since this is just testing. This only shows US, and only divisions with modifiers set in 1v1 diamond. Gold number is modified points, white is original, everything is sorted by modified.
|
I started playing on LA using the free weekend and i can confirm 2 things.
1.Division are made solely regarding MMR, i assume blizzard estimates a range and put in each division certain people that fit there. It has absolutely nothing to do to creating divisions once enough people are ready to be promoted, i got promoted to diamond in something like 23 games and the Rank 1 of my division had more than 100 games[2k+points] and he had been playing for at least 2 weeks and after i joined more people joined who had kind of the same win rate. Also im pretty sure most of them were already diamond in other servers.
2-Bonus points are not considered in the TOP 200. Or they consider them anyway regardless of you using them or not. This explain why Idra for example kept his rank despite not playing.
3.MMR is an extremely efficient system, i started playing with people around my level in around 10 games. (first 5 practice games considered) This is probably because in the first game they put you against a platinum player now. (I believe they implemented this after the 2v2 abuse).
|
Akilae Omicron and Medivac Alamo are currently missing from the masters page. The kinks are obviously still being worked out 
|
Oops, it was setting the division to whatever division it could find not 1v1 diamond. It's fixed, look again.
|
On November 21 2010 03:40 Shadowed wrote:Page up: http://sc2ranks.com/mastersIt's not linked anywhere else right now since this is just testing. This only shows US, and only divisions with modifiers set in 1v1 diamond. Gold number is modified points, white is original, everything is sorted by modified.
Awesome work Shadowed! Hopefully with a few more top 200 lists we'll be able to add more divisions to the list
|
|
|
198 AlwaysYou 227-175 2250 Shuttle Nu
199 VTGoddeR 569-495 2553 Void Ray Tau
200 SeaofStorms 172-120 2245 Talematros Eta
I was curious about my division and happen to be in Void Ray Tau, and checked VTGoddeR's match history, and subtracted the amount of 1v1's to equal his record on the top 200 list. He had 2499 points as of the date Blizzard ran the top 200. Subtracting 252 (D Rank) puts him right between #198 and #200 with 2247 points. Would this be accurate enough to determine that Void Ray Tau is in fact D rank?
|
Im in NA but my division isn't listed? "Tarsonis Alpha" is the name and I guess it must be so low it's off the charts. meh.
|
On November 21 2010 08:33 Fusionbomb wrote:Show nested quote +198 AlwaysYou 227-175 2250 Shuttle Nu
199 VTGoddeR 569-495 2553 Void Ray Tau
200 SeaofStorms 172-120 2245 Talematros Eta I was curious about my division and happen to be in Void Ray Tau, and checked VTGoddeR's match history, and subtracted the amount of 1v1's to equal his record on the top 200 list. He had 2499 points as of the date Blizzard ran the top 200. Subtracting 252 (D Rank) puts him right between #198 and #200 with 2247 points. Would this be accurate enough to determine that Void Ray Tau is in fact D rank?
Yes VRtau is D rank.
Well done putting together the implementation on sc2ranks, shadowed. we'll confirm excalibur's theory at the next top 200 by comparing it to the sc2 ranks list
|
Damnit. I'm in Scout Kilo.
wtf, lol.
|
On November 21 2010 03:40 Shadowed wrote:Page up: http://sc2ranks.com/mastersIt's not linked anywhere else right now since this is just testing. This only shows US, and only divisions with modifiers set in 1v1 diamond. Gold number is modified points, white is original, everything is sorted by modified.
This is REALLY nice. Good work.
|
|
|
|
|
|
EPT
