|
United States12224 Posts
Levin is in an S-Rank Diamond division on LA with 2192.
aStrO is below him in Master with 2027 at the time of the snapshot. BiOZeRg is above Levin in Master with 2043.
That means that 2192 S-Rank Diamond is between 2027 and 2043. A possible difference of 149-165 points (probably 150 again).
Madness is in an S-Rank Diamond division on LA with 2123.
cREATiVe is in Master with 1971.
OXOmgmt is in a B-Rank Diamond division with 2246 (unmodified: 2120).
That puts 1971 points in Master between 2123 and 2120 points in Diamond, for a possible range of 149-152 points (again probably 150).
Khas is in Master with 1960.
Oxente is above him in an E-Rank Diamond division with 2427 (unmodified: 2112)
pOwGuasy is below Khas in a C-Rank Diamond division with 2298 (unmodified: 2109)
That means 1960 Master is between 2112 and 2109 in Diamond, possible range 149-152 again.
|
I'll try to make at least a 1/4 parse again.
The 168 guy (Boxer) actually won 18 points in his last game, probably the promotion game, that remembers me of the LEFT game bug, do you remember that one Excalibur?
So there is hope there is indeed a offset of 150...
|
United States12224 Posts
That's possible. It sure sounds like an anomaly given these other data points.
|
Also, before I forget it:
Bonus pool right now:
2422
40% means = 969
Anyone with 969 bonus points to get or more shouldn't be there if my theory is right, let's hunt these players XD
Edit:
I am confirming over and over the 150 "offset", the 168 guy is most likely a bug/anomaly, our dream lives XD
Edit2: Ok, master -> S diamond difference confirmed: 150 points.
Top 200 points requirement this week: 2050 if S diamond, so 1900 if Masters.
I'll look for someone with more than 1900 that didn't make it.
|
United States12224 Posts
That's proof enough for me. I added the info to the OP as a new tier of -150. I wonder if each other league offset is +150 also?
|
On January 25 2011 08:17 Excalibur_Z wrote:
That's proof enough for me. I added the info to the OP as a new tier of -150. I wonder if each other league offset is +150 also?
That one is impossible to know, and the plat <-> diamond difference is even wierd, as we do have F-Rank divisions confirmed on LA, SEA, but not US. So it would be 150 points from F? E?
Anyway, Master could have a special number, if we ever want to make a consolidated league we'll need to ask Blizzard to spit the numbers! ^^
|
Looking at non-filled division on 2x2, 3x3, 4x4 on all servers (NA, LA, SEA, Ru, Eu etc) we can verify that there are at least:
1 Master division (confirmed)
7 diamond divisions (confirmed)
2+ plat divisions
2+ gold divisions
2+ silver divisions
5+ bronze divisions
Unfortunately we don't know the offset from none of them, but it's interesting to know that they do exist.
I also think it is interesting to note that gold has the least amount of tiers while still having as many (active) players than diamond(+master) and bronze. That probably means that a Graph for Number of Players and MMR (for active players) will be something like this:
![[image loading]](http://imgur.com/97i1r.jpg)
(Ok, my Excel + Paint skills sux)
What I am trying to say is that the worst gold and the best gold player have a difference of "skill" very minimal, but the difference of skill between the best diamond and worst diamond is huge. That explain the need for 7 diamond tiers, but only 2 gold tiers. (and that also means we need a lot of bronze tiers, maybe more than 5)
Edit: Note: The negative MMR bug would only happen then for inactive players that have a "skill" lower than of the last (and worst) active player.
|
|
|
My mind is blown that you would even post here claiming that MMR is merely an offset away from displayed ladder points. MMR is SO much more complex than what you're suggesting it is. If it were directly offset from ladder points then we'd see that all diamond players of one division class would jump to Masters at hitting a certain number of ladder points--and this isn't the case.
It's obvious that MMR is so much deeper than ladder points. How do you explain the fact that when Master League was first implemented, top players with 0 bonus pool were both FAVORED against each other? According to you their MMR would be the same as a low Master player who also just got promoted.
|
1) The 50% wins / losses aren't enforced, just a consequence for most players. The system doesn't "want" you to be 50%. Most of the times it just means how many bronzes-platinum you "smurfed" before facing real players. Or should I say, how long the system took to put you against you adequate skill range. That said, I doubt you should put it in the math.
2) You don't know what MMR looks like, so stop calling it 'MMR'. 63 "points" could mean a difference of 20 MMR, we just don't know. Call it something else like PMMR...
3) It takes 112 minutes for everyone to gain 1 bonus pool (and not 2 hours).
The rest is fine, you can refine what you have and it will be pretty much what we already have... hm...
I'd like to add that when ladder resets everyone will still have a MMR, and everyone will start with 0 bonus pool and it'll take a lot more than 5 games to be able to calculate this faked / improvised MMr to be the least accurate!
Sincerally, though, I think you are focusing in the wrong "problem". None needs to know how much MMR they got. What we want to discover is:
1) Why am I not in master/diamond/whatever yet?
2) Why was I demoteed if I was top 10 in my diamond/platinum/whatever league?
And this questions can't be answered easilly specially cause it's seemingly impossible to figure out the division offsets of every division in all these servers. If we knew that already, we could answer so many more things.
So instead of trying to figure out "MMR", I want you to think about how will we know what tier the division aba kadraba on silver is.
|
|
|
|
|
|
|
Please show how you established that you must be promoted after 475 adjusted points. Second, please demonstrate why sigma still matters when we know that the moving average is what needs to stabilize within a league for a promotion.
|
Holy eff I think I remember your thread back then but it has disappeared, I read this from you late at night and you have shown even better evidence (proof?) But holy batman as you say haha, I don't get it! May be I'll put more time into it but it seems you have put a lot of time so for the time being I'll believe you. Especially if what you said was true that it is very similar to the old WoW arena system.
Thanks for doing all this work and I hope you are right or very close and that your work isn't wasted 
|
|
|
My first theory when I saw that bronze guy was that his moving average just wasn't stable enough, but he did lost 11 points on recent 1x1 games, that probably means he is actually stable enough.
So I think your idea is quite close to some truth.
There is no need to divisions tiers offset on bronze to be 63, it could be higher, so there would be no need for 11 tiers on Bronze. If tiers on bronze are 120-180, then we could still have the 5-7 tiers, though none of this is easy to confirm, it's good to expand the theories, thanks for sharing yours.
I'd like to add that the system is conservative. He won't promote you as soon as you reach the "barrier", he'll need you to go further AND to stabilize there.
I think that the +73 points that you get when you're promoted/demoted could be a reflex of that extra "confidence" the system wants on you before you get promoted.
Se if you are a Diamond S and wants to be promoted to master, you not only need 150 points, but this extra 73 points. Obviously MMR is still hidden and there could be even more variables into the math, but I think that for most people it will be accurate for most of the time, for most of them.
I think that the next step is stalking these players that we can grasp they'll get promoted "soon" and check if they indeed get promoted, when exactly and things like these. I'll need time and lots of stalking 
|
|
|
I said it on the other thread, but I'll repeat it here.
I think the problem is that this, or whatever similar formula, work for SOME players but not ALL.
If you generally win the games you should win, and lose the games you should lose, then you fit into the case where it does work.
If you are erratic and have a lot of wins of less than 9 points, and/or losses of greater than 14 points, then you are going to deviate from this formula because the matchmaking system does things to try and reduce your sigma and now you have a non-linear correlation.
If you really want to show us your formula works, you should not try to find cases where people fit your formula. Instead, you should try and take a broad spectrum of people (ie a representative sample of the population) and show that that entire spectrum fits your formula.
In regards to finding out about promotions, I'll take a couple samples now and post them in about 8 hours. In addition to the people who do get promoted, I'll post a list of people who still played and didn't get promoted, and we'll see if the formula can explain both cases.
|
|
|
|
|
|
|
|
EPT
