I don’t get why you say things like this anet…
Currently some worlds, such as Crystal Desert, are stuck in what’s come to be referred to as “Glicko hell”. This is where a few worlds’ Glicko ratings drift off from the rest of the group, leaving a wide enough gap that those drifting away won’t be matched up against the rest.
To help prevent stale matchups, we’ll soon be able to artificially adjust Glicko ratings solely for the purpose of matchmaking.
For example, if we use this for NA’s lowest tier of worlds (T4), bumping them up to be just below T3’s ratings, they’ll have a decent chance of being shuffled into T3 during matchmaking. Given the ratings today, an adjustment we might make for July 29th could look something like:
T3:
- Maguuma: 1,838
- Sea of Sorrows: 1,777
- Stormbluff Isle: 1,767
T4:
- Crystal Desert: 1,512 (plus-185) = 1,697
- Sorrow’s Furnace: 1,398 (plus-130) = 1,528
- Darkhaven: 1,366 (plus-100) = 1,466However, note that this artificial adjustment is invisible to anything but matchmaking. So if a T3 world doesn’t beat a T4 world significantly, the T4 world’s actual Glicko rating will increase more than usual, while the T3 world’s actual rating will decrease more than usual. Both worlds will also have increased deviation and volatility after. This is all because the T4 world’s lower rating means to Glicko “this world is expected to lose significantly to that T3 world”. So in an upset where this doesn’t happen, Glicko both makes larger adjustments to rating, since it was shown to need change, and increases deviation/volatility, to tell itself on the next round that these worlds haven’t settled into their most accurate rating yet.
EU’s lowest-rated world is still within its deviation of the lowest-rated world in the next tier up, making EU much less likely to receive artificial matchmaking adjustment on July 29th.
The second set of pairings was ur idea of ‘more balanced’ and we all know how that ended up for most of the tiers. You need fresh thinking here. Fresh simplistic thinking. Maybe then you will be able to see that 4 servers vs 2 small servers is not a good idea.
————————————————————————————-
IDEA
Why can’t you just reset everything instead (except the pairings)? Then every current pairing can have diverse matchups each week. And EVERY pairing can quickly move to where it should be. Every server can GAIN something from winning.
Week 1:
-All matchups COMPLETELY random. At the end of the week: 1st place servers get 2 points, second place servers 1 point, third place 0
Week 2:
-Everyone is matched most points to least points. That way first placed server are matched up against other first place and some get lucky and are against second place. No first place server is matched with a third place server from the previous week.
Week 3:
-Continue this process from week 2.
Week 4:
-Continue
Week 5:
-TADA! WE will have more balanced/neck-neck matchups ‘overall’ than we have had in any of this ‘beta’ period.
A discrepancy will exist obviously. The top most server may be insanely powerful compared to the others. And the bottom most server may be extremely weak compared to the others. This is the gamemode’s fault though, but it can be remedied for most servers. For the Top most tier, changing little things would work (map que’s for one, since it doesn’t matter how powerful they are if all 3 teams have the same number of players in ebg). More methods to deal with zergs/outnumbered forced would also be a good idea…but honestly since thats never gonna happen u can get away with ‘population spreading’ and putting floodgates on the top most server’s firepower.
The real problem lies with the bottom most server, which might become a dumping ground for endless losses. You will have to get a little creative with how to deal with that.
My ideas on how to deal with a dead-last server: Threaten to kill it entirely if it continues to lose and become too much of an outlier in the point spread. And/OR Merge it with the 2nd lowest server or some other server (if its small enough put it with the lesser server in the mid-upper levels).
Also with the first week you ‘could’ match the servers based on the current data you have obtained…or even just leave it in the current matchups and then just allow people to migrate out of those matchups each week in the process described. Personally I would from best to worst: 1. make it completely random (best idea); 2. match based on the data you have; 3. start matchups same as present (worst idea)
If by some crazy chance u get too many new tiers that don’t work well, then you just change the pairings and reset the ratings! With some ‘educated’ guesses when creating new pairings, based on the ‘data’ of the reset, you can/will come out with better tiers once the 2nd set reaches the same number of weeks as the first.
The Tiny Yuno Sniper of Ebay [EBAY]
(edited by Cerby.1069)