Guild Wars 2

A discussion around the application of the Glicko system to WvW started to bud in the TC/CD/FA board, and in order to not derail it, I’m creating this thread to continue said discussion.

A few folks have expressed their distaste for the system because of how it has the potential to create discrete groups that do not move, stagnating any potential rise in rankings. In the extreme case, you could swap the top 3 worlds with the bottom 3, and if they maintained fairly even scores in order to not disrupt their Glicko rating enough to bump any of them out of their bracket, the two groups would persist in the ranking that is opposite their “true” rank.

Personally, I don’t think WvW rankings mean anything whatesoever. They could name the brackets Frank, Ted, Wilson, Mary, Julie, etc. and it would make no difference to me (though I’m sure people would still put weight on the random name that corresponds to the group with the highest Glicko). All I think the system does is attempt to match worlds up that are as closely aligned as possible with regard to their WvWvW variables (total population, coverage, coordination, etc.). When that is achieved, the system succeeds, because matches are interesting when they aren’t total blow-outs. More people play across all servers that are even, and I’d wager that when matches are blow-outs, less people on all servers participates (who likes to defend against no one, and who likes to siege with futility?)

At any rate, feel free to post ideas on a potentially better system, as it was mentioned that a more simple system could potentially be easily conceived in minimal time. My personality loves to pick algorithms apart and think of boundary cases/ways that it might break, so I’ll probably be posing questions that attack each system.

Please note that I have no general love for the Glicko system, I just think it’s working as intended . I do, however, think any alternate algorithm should be stable enough to elegantly handle any situation properly, and I think “fun” (as measured by world presence in WvW or score disparities) should also be a metric of success for the system. Theoretically you could round-robin and follow up with a tournament to determine the “best” server, but the majority of servers would probably come to hate WvW in the process.

So let’s hear it.

In the current state, the top and bottom are polarized. If you think of the rating distributions as those of a normal curve, the highest and lowest ratings are relative outliers and are therefore harder to reach. The ones that fall under the center of the curve (Tier 3-6) are within about 1 Standard Deviation from the center of the curve, approximately 1500 rating. Unless there is absolute parity, which is next to impossible because of servers being “official” this and “official” that compounded with free transfers, there is no upward or downward movement with any degree of swiftness other than those servers in the middle or those who receive a large amount of WvW centric transfers.

The Glicko 2 system is creating the best matches possible for WvW at the moment. If Tarnished Coast went up to Tier 2, they would get blown out. If Blackgate, my server, went up to Tier 1, the same would likely occur. This is evidence at the top end of the spectrum. The same would likely hold true in Tier 7-8. If you are in the middle tiers, getting to the next tier isn’t nearly as impossible, but the higher or lower you go in the tiers, it gets progressively harder to rise or drop depending on which side of the curve you are on from the center. Ultimately, you hit a wall where your server fluctuates within its own standard curve and has its ultimate high point and ultimate low point restricting it to its current seeding +/- 1. Potentially the biggest flaw is that you can lose rating by winning a match, this is likely the largest contributing factor for stagnant matches, but it also saves a server where there is a large rating disparity from getting bumped to a tier where they will get killed or bumped down to a tier that they would dominate.

Currently, the same people are playing week after week in the same matchups for the most part. It is difficult to judge a system where it is working as intended, but the game type lacks retention values that is necessary for an effective endgame or the dynamic qualities that will not only keep players, but bring more players in. The upcoming WvW patch should give a much better indication of the health of this aspect of the game, but I can assure you server stagnation is not a negative thing, rather, just the current state of WvW.

I’m no statistician, but I feel the way they use the modified Glikko rating system could benefit from a modification.

The attached picture highlights the issue …

This week in tiers 1-3, all green servers are in the lead and will likely win the match, yet all three are losing rating on the week. Why? Because the the score differential is too close, and Anet’s implementation of the Glikko rating system uses scoring rather than win/lose/draw.

While this is great in practice, as it means the system has worked and grouped the servers according to their scoring ability, it can also lead to the content getting stale due to facing the same opponents week after week. Unless there is a shock to the makeup of the WvW population on a server (such as a mass exodus ala HoD and ET), the more time that goes on the less likely we will see movement up and down tiers, especially in the top 3 and bottom 3 tiers.

In a nutshell, it seems to me that it is too difficult for green servers to move up a tier, and red servers to drop down a tier. To remedy this and potentially add a bit more spice to the WvW weekly matchups without creating total blowouts, would be to augment the rating system to include bonus rating for winning a match, no bonus for 2nd place, and negative rating for 3rd place. This would have to be a ratio rather than a fixed number so the amount of increase or decrease is relative to the current rating of a server.

Something along these lines anyways. I far prefer seeing the same servers every week if it means I’m more likely to experience a close competitive match, but I also think a little more server movement up and down tiers at this point could help keep things from getting boring.

Cogbyrn.7283:

Personally, I don’t think WvW rankings mean anything whatesoever. They could name the brackets Frank, Ted, Wilson, Mary, Julie, etc. and it would make no difference to me (though I’m sure people would still put weight on the random name that corresponds to the group with the highest Glicko). All I think the system does is attempt to match worlds up that are as closely aligned as possible with regard to their WvWvW variables (total population, coverage, coordination, etc.). When that is achieved, the system succeeds, because matches are interesting when they aren’t total blow-outs. More people play across all servers that are even,

Maybe but stagnation itself leads to people stopping participating as happened with YB by the 3rd week of the same matchup in Tier 3. Its an issue with this system.

With regard to stagnation/feeling stale, two things:

1. If the 1/3 slots start rhythmically bouncing between tiers, would that get boring as well? I mean sure you’re facing “new” opponents occasionally, but in one bracket you face opponents you’ve already faced and win without much issue (typically), and in the other you’ve faced opponents that typically beat you without much issue. Would that change just delay another claim of gameplay feeling stale?

2. A stale match-up is only as stale as the people participating. If you do the same strategies each day you log in, it’ll probably start to feel stale. However, nothing’s stopping you from spicing it up. Dive deep into enemy territory an harass Doylaks. Cap deep supply camps to divert enemy forces. Do something you normally wouldn’t do and see if it works better.

Naturally, there’s a concept of “playing to win”, and so if you are already winning, you can just keep doing what you’re doing. However, I feel like that mentality falls a little flat when you’re dealing with so many variables outside of an individual’s control.

As an aside, I have a theory that “Gamer A.D.D.” is a serious concern with regard to long-term retention for larger-scale games. It seems like people need something “new” to maintain interest in end-game, whether it’s new bosses to kill, or new maps to fight on. However, that can be saved for a separate discussion.

I still personally think the system is working as intended, and as Esoteric mentioned, the complaints about the system or stagnation are symptoms of a deeper problem people have with WvW.

Maybe the solution to this is really quite simple: Win your tier and you move up, lose your tier and you move down. (With the obvious exceptions of 1st and 24th ranked servers.)

This would keep the match ups constantly changing, but each server would stay in about the same place.

As far as which server is red/blue/green, the green server can be the higher ranked team that moved down, blue is the team that stayed in the tier and red is the lower ranked team that moved up.

Yeah, this may create some blowout matches, but I’d gladly take that once in a while over fighting the same servers all the time. Besides, a blowout win and you’re up into the next tier, a blowout loss and you’re down into the next lower tier.

DemonCow.5328:

Yeah, this may create some blowout matches, but I’d gladly take that once in a while over fighting the same servers all the time. Besides, a blowout win and you’re up into the next tier, a blowout loss and you’re down into the next lower tier.

The sheer number of blow-out matches every week with a system like this would probably be discouraging for many. You say you’d gladly take a blow-out match “once in a while”, but if you’re in T4 and move to T3, and a T2 world simultaneously moves down to T3, odds are you’ll be getting blown out every other week.

As a Maguuman, I’m pretty comfortable with how it’s done now. We pretty much chill out in tier-Maguuma week after week and every other match-up we get a new opponent or two. We essentially get the same end score every single week, regardless of who we fight, so where we are seems to make sense.

- tiers/scoring is primarily driven by population dynamics
- actual tiers/scoring are pretty meaningless (and should be)
- tiers/groupings are such just to ensures that you are not facing an opponent that out numbers you to a degree that you are spawn camped 24/7
-**** There are a limited number of servers and very few “like” servers from a pop perspective. We’re all going to fall into a small grouping and be pretty stagnant +/-.
- Xfers are accounted for poorly in the current system, and really are the only way for most servers to dramatically impact their performance, with lower pop servers/tiers impacted the most (re: pop>all)
-matches are as fun as you make em.

When I look at the score, I think the following is happening (I’m TC and using it as an example) :

The algorithm says “TC is expected to comfortably defeat FA”.

The result comes in, TC wins with a comfortable margin.

Algorithm concludes that TC & FA have correct ratings.

And this would be correct, but it doesn’t make for servers shifting tiers until people shift servers.

If you want this to change, it seems obvious to me that the thing to do is to change the coefficient. If TC has won tier 3 for a full month running but still does not overtake the 6th seeded, it is because the coefficient (the number that you multiply into the difference between the achieved and the expected score) is too small. The amount of points that TC gets for winning is too small compared to the distance it needs to cover to get to tier 2. But it’s all a question of balance, obviously. If you make it larger, people will start running up and down the tiers, but personally I’d rather be beaten by someone far better than myself once in a while than never playing them.

Cogbyrn.7283:

In the extreme case, you could swap the top 3 worlds with the bottom 3, and if they maintained fairly even scores in order to not disrupt their Glicko rating enough to bump any of them out of their bracket, the two groups would persist in the ranking that is opposite their “true” rank.

I see you remember my example from when we last had this discussion, lol. But I must state again, the point of the example is not to say that the servers in the example are being hurt in any way, but to point out the fact that the rating system in place does not take overall rankings into account. This creates the problem where servers may get stuck with each other when there may actually be better pairings.

Let us assume for the sake of discussion that right now Tarnished Coast has become a stronger server than Isle of Janthir. If this is the case, then we would have two better matches if these two servers swapped places. But currently, IoJ is pulling its weight against BG and SoR while TC is being held back by CD and FA. This week, IoJ is gaining in rating while the opposite is true for TC. Under the current system, they won’t swap. We don’t know who the stronger server is between IoJ and TC in reality. They haven’t faced common opponents for a long time and all servers have evolved since then. Maybe IoJ is far superior to TC, but maybe not. Under the current system, we won’t know until something drastic happens to shake up the rankings. We are potentially getting deprived of better matches because Glicko is creating bubbles that servers (at least in the top four tiers) are having trouble escaping from. Interestingly, the bottom four tiers provide a lot of examples of servers that got stuck in tiers when they deserved to be in other tiers. They were/are getting fixed by HoD/ET shaking things up, but if not for those two, they would be in the wrong matches.

I suggested this before and it seems someone else suggested it again in the TC/CD/FA topic, but it would be better if after TC won a certain amount of times and after IoJ lost a certain amount of times, those two get switched with no changes to rating. The other person who suggested it came up with the number 2 for how many matches it should take. I would say either 2 or 3.

So, what your suggesting Arcadio is something like

if server (wins x 4) they move up regardless of rating?

If that’s the case while for the most part it would be fine, but a point in case would be if 2nd and 3rd are constantly switching from day to day week to week some might argue it’s a little unfair on those 2 servers.

also for the record, I like the old fashioned ways where if you win you go up if you lose you go down, add’s a little varity in there. But yes this could create big blow outs more often, but it also might not. Only one way to find that one out.

If the 2nd and 3rd server keep swapping places under the current system, the slightly modified system would take no action. For example, if SoR and IoJ constantly swapped places, the modified system would not force TC into T2. It would only do that if TC won 1st place X amount of times and IoJ got 3rd X amount of times.

Really, my idea is a far less extreme version of if you get 1st you go up and if you get 3rd you go down. There are very obvious problems with that system even if it does solve some problems. A less extreme version of that system will reduce those problems.

Arcadio.6875:

If the 2nd and 3rd server keep swapping places under the current system, the slightly modified system would take no action. For example, if SoR and IoJ constantly swapped places, the modified system would not force TC into T2. It would only do that if TC won 1st place X amount of times and IoJ got 3rd X amount of times.

Really, my idea is a far less extreme version of if you get 1st you go up and if you get 3rd you go down. There are very obvious problems with that system even if it does solve some problems. A less extreme version of that system will reduce those problems.

I thought about 2 or 3 times as well, but that leaves no room for a situation like the following:
Let’s say TC is comfortably winning Tier 3 every week with CD and FA pretty close to each other, and IoJ and SoR were pretty close in score, while Blackgate is comfortably winning Tier 2, but neither IoJ or Sor (or FA/CD for that matter) are losing the 3 matches in a row required to move down a tier.

Clearly this isn’t the exact case, but it is quite possible.

Another solution may just be to keep a running tally of wins/losses in a tier. Once any one server hits 3 total wins in their tier and, in the next tier up, any one server hits 3 total losses, they switch. Then, the tally for the server that moved down is automatically given 1 win since they came from a higher ranking, and the server that moved up is automatically given 1 loss, since they came from a lower ranking. This would keep them switching less often….. anyways, just another idea.

The bottom line is that the top 4 tier matches are hardly ever changing. There’s no excuse for one server to win (or lose) repeatedly in their tier and not get moved because they didn’t win (or lose) by enough points. It’s very difficult, if not impossible, to actually give the servers a relative rating if they aren’t ever playing each other (especially given that there are still players and guilds moving between the servers).

Maybe we should do a little of what everyone has suggested. Change the coefficient in the Glicko system, like Lord Emamah suggested, while still having a fail-safe of sorts where if one server wins several matches in a row and the bottom server in the tier above them loses several matches in a row, switch them, regardless of rating. Since some servers would occasionally be forced into a tier above/below their rating, this would also prevent the “bubbles” forming in the current ratings. If the tier 5 servers were drastically better than tier 4, then as soon as the top tier 5 server moved into tier 4, their rating would also change drastically until all of the tier 5 servers moved ahead of the tier 4 servers.