Guild Wars 2

The algorithm while sound in theory is horrible in practice. Consider the following scenario:

3 sides, mutually matched well together and they lose by a very small margin. The algorithm works.

3 sides, mutually matched, but 1 side takes the lead and just absolutely dominates the entire match. Now the other side or sides generally quit or are so oppressed that they cannot get a foothold again. This is where it fails.

The reason? The sides are evenly matched but any number of situations or circumstances could cause the aforementioned problem. But what happens is that the losing server now drops… 4 rank places. This may put them into a division 2 divisions below them in which they are absolutely overdominant in.

Let’s continue with the scenario, they now dominate and move up the point system until something unexpected happens again. Now, that server is stuck in a feast or famine situation.

So ya, discuss… tell me I’m wrong… tell me how to fix… but generally, I’m pretty sure I’m right.

Do the algorithms take into consideration the server scores from the past weeks? If it does then server scores will even out.

they do, what happens is that the scores are put in an equation to calculate the ranking score then is added to the server’s ranking score. there was a post about it and they also gave the formula, however it was hard to follow at best, would need somone to work out a spreadsheet to make it more simplified

I don’t think it takes 2 weeks, it only takes the past week.

This is why when SF got first place in week match, it still fell 5 points (because it was so close of a matchup and we were the point leader going into the matchup).

this is what a posted in the last scoring thread.

to me, its the only fair and viable option to help even out the scores. it wont give anyone an unfair advantage, just makes it less of a faceroll during the times when a server doesn’t have as many players as the other

if you have 100% of your people and the other two teams only have 20% of their people, you should not be getting 100% of your score… basically winning because of an unpopulated server, it should be giving you less points per tick because you can simply hold ground 100 times easier due to numbers.

if its a full queued up server vs 2 unqueued and undermanned server, they should realistically get 20% of their total points if the other two teams average % of people to full is 20%.

so say team 1 is at 100% capacity,

team 2 is at 75% capacity, and team 3 is at 25% capacity.

that means that team 1 will only get 50% of their total ppt each time.

team 2 will get 62.5%

and team 3 will get 87.5% of their ppt

so if team 1 held 325, then they would get 162.5

team 2 held 250, then they would get 156.25

and team 3 had 120 then they would get 105

as you can see. it doesn’t give people a huge advantage for purely outnumbering anyone, and we all know that even though they won’t get tons of points per tick, they will still be gaining score, but it wont be a blowout overnight.

this would, IMHO, steer the points system in the correct direction to allows more fights between servers with obvious population gaps at times to be a more fair game.

when outnumbered, you won’t totally be obliterated with nothing you can possibly do, even though this is your normal play time… and you wouldn’t have to try and switch servers just to find a good one.

this also means that when the servers are getting full queues vs each other that if one really is 250 points ahead of another, then they will get 100% of their tick score because they are fighting against full forces.

perfect? no, obviously no, nothing is perfect.

but it feels like the best solution to help alleviate some issues with population differences that cant be solved by asking people to politely transfer on the forums…

^ best example right now i can give is, Team legacy (yes the same one that claimed they would be amazing and then folded), has now switched from ET, which was suffering from enough population to stay in tier 1 consistently… to JQ… who is in tier 1 right now too and has a higher pop than ET already… this does not help the game and the way things are going right now. and these kinds of people are the problem that scoring system changes can help fix

Equally though if in a level battleground, you crush your opponents in the first 2 days and their spirit and hope is so low that they put up a feeble effort for the rest of the week…..you shouldnt be punished…..its not your failure that your foes were weak.

In essence with a ranking system, over time the league table will stabilise and like for like servers will be playing each other. Now obviously free transfers and the emergence of well organised guilds as well as the player composition geographically will have an influence, however after a couple of months, things should settle. Until then expect the to get the odd thrashing and give the odd thrashing.

The Tier 2 (4, 5 and 6) battle ongoing now is a good example. One fairly well organised populated server is rogering two reasonably organised, reasonably populated servers, who are within about 5% of each other on the scoreboard…..result….much fun and appropriate disdain for the powerful one/pride in being the powerful one.

At the end of the day, the server with the best organised guilds, evenly distributed timezones will triumph…with for me the only caveat being free transfers….in saying that free transfers are great for everything except WvWvW balance…..in saying that in RL war….defectors/changes of alliance often turn a battle.

^ of course if you crush your opponents you shouldn’t be punished… you are still winning though, and no matter what, if you are holding more territory for the numbers, then you will be getting the appropriate score..

if the opponents give up. you still win, its not like they can just give up and you suddenly go into last place.

in a game like this, it will eventually get to the same point its at right now, where X number of servers completely dominate because of population spread.

and then the problem with people just giving up comes from seeing completely unbeatable odds when it comes to points.

its a much better drive for players to log on and know that they may be outnumbered 5 to 1 vs the top server, but if they can take and hold at least 150 points worth of stuff, then they can stay in the game, and hopefully make it to their next prime time..

this isn’t just for tier 1 matchups.. this is for all the way down to the last matchup where kaineng is always outnumbered, but keeps fighting… if they could possibly put up a fight worthy of getting enough points, then thye would lose by 20k instad of 400k. which would make it not only more of a fight for everyone by the numbers, but would encourage more people to come back to WvW simply because “they have a chance to come back” and aren’t completely dominated by the number board

EnochDagor.6185:

I don’t think it takes 2 weeks, it only takes the past week.

This is why when SF got first place in week match, it still fell 5 points (because it was so close of a matchup and we were the point leader going into the matchup).

It does take past performance cumulative performance into account.

This actually discusses ELO but it’s still on point for Glicko-2

A Comprehensive Guide to Chess Rating

The attenuation factor 8 in formula ( rpost = rpre + 8 ( S – Sexp ) ) can best be interpreted as the amount of weight given to the new tournament performance relative to the pre-tournament rating. The larger the value of 8 , the greater the amount of change allowed in one’s rating. It can be shown mathematically that for a 4-round tournament, setting 8 = 3 2 corresponds approximately to computing a weighted average of a pre-tournament rating and a performance rating with weights equal to 94.7 % and 5.3 % , respectively. 2 ± This implies that each time a new tournament is observed, 94.7 % of our belief is invested in the old rating, but we let 5.3 % of our belief be guided purely by what happens in the tournament. If computing a tournament performance rating, rperf , were a straightforward calculation, then an alternate method for computing a post-tournament rating corresponding to 8 = 3 2 would be

rpost = 0 .947 rpre + 0 .053 rperf .

Analogously, when 8 =2 4, the weights become 96.2% and 3.8 % , respectively, and when 8 = 16 the weights become 97.5 % and 2 .5 % , respectively. These approximations only hold when the discrepancy ( S – Sexp ) is not too large.

In GW2’s case either match up is equivalent to a tournament and If I remember correctly K would be calculated based on several things such as rating and deviation.