Guild Wars 2

I’ve enjoyed the new matchup system because it has given my server new enemies to fight. Sometimes we’re terrifically outnumbered, sometimes the opposite, but I can find fun in either situation. It’s better than always facing the same servers.

However, I’d really appreciate someone explaining to me how it’s meant to fix the system, because I see some wild numbers depending on who we fight. When we fight a tougher opponent our score goes way up because we manage to keep fighting anyway. Then we’re placed with weaker servers and our score tumbles again because we can’t destroy them as much as we apparently should.

So I don’t see much change overall, and I fear when the ranks “even out” we’ll be in the same situation we were before, always fighting the same two servers and that match up is not good week after week. But I might be missing something about the new system that will make it clear. Am I missing something?

if you have been enjoying the new system, you can expect to continue enjoying it. the new system is not temporary. it’s not something meant to be in place for a little while, then removed afterwards.

it’s intended to be in place permanently, although as the ratings become more accurate ArenaNet has said they will tune some of the randomness down a bit (I think this hasn’t happened yet because EU servers are still seeing a lot of movement every week).

there are 3 things the new system is intended to “fix”:

1. the old system led to opponent stagnation — servers whose ratings didn’t change much could find themselves playing the same servers over and over. randomization fixes this problem outright.

2. the old system led to map stagnation — it could make map completion difficult for some players. a good example was EB vs BP vs AR in the old system — those 3 servers had played each other for something like 11 weeks before the new system went in, and during that time EB was always green. this made map completion difficult for BP and AR players who needed map objectives in the green borderland map or in the green corner of eternal battlegrounds. another good example is FoW and Vabbi in EU — they never got an opportunity to get the green objectives under the old system. a side-effect of the randomization system is that color changes are now possible. in fact, in the very first week of the new system, EB ended up playing BP and AR again (getting the same matchup was a disappointment to everyone involved). but, due to the randomization, at least the map colors were different that week.

3. the old system made the rating gap between tiers inaccurate. under the old system, rating points were carried up the ladder any time a server moved up across a tier boundary, but rating points never came back down again (even if a server came down, another server who had more rating points always moved up to replace them). this tendency for rating points to never flow down over a tier boundary led to 2 problems: (1) the ratings for the higher servers were likely to be inflated, and (2) as the size of the rating point pool got smaller in the lower tiers, it became harder and harder to move up because servers needed a bigger piece of a smaller pie. so consistent wins were no longer good enough to move up, you needed consistent blowouts. and eventually consistent blowouts would have been insufficient — to get rating points you would eventually need to consistently crush your opponents, spawn camp them, and make then not want to play. the new system allows points to flow back down, by allowing high rated servers to play low rated servers (and lose rating points if they don’t win by the expected amount). this is the “temporary” problem, and as the weeks go by the ratings should correctly reflect average predicted score outcomes. when the inaccurate rating problem has been fixed, I think we can expect ArenaNet to reduce the amount of randomization so that we still get variation in matchups, but without quite as wide a range of likely opponents.

the rating accuracy problem is a key issue here — in order to get “good” matchups the new system needs accurate ratings, because matchups are made based on how close the ratings are — servers with similar ratings can expect to play each other often, and servers with ratings far apart will play each other rarely (if ever).

under the old system, “how close” the ratings were was largely irrelevant, because matches were assigned purely on the basis of ranking. if you were the #4 server by a big margin or a small margin made no difference at all — you were either #4 or you weren’t, and if you were #4 that meant you would always play #5 and #6.

-ken

Thanks ken, that makes a lot of sense. I’d heard the randomness was going to be narrowed which made me think we were going to get stuck into the same system again somehow, but if we’re still meant to fight different servers I’m glad that will continue.

I’ll continue to be frustrated by how our rating goes up and down and leaves us in the same place, but as long as the randomness continues at least it won’t trap us anywhere.

Sitkaz.5463:

I’ve enjoyed the new matchup system because it has given my server new enemies to fight. Sometimes we’re terrifically outnumbered, sometimes the opposite, but I can find fun in either situation. It’s better than always facing the same servers.

However, I’d really appreciate someone explaining to me how it’s meant to fix the system, because I see some wild numbers depending on who we fight. When we fight a tougher opponent our score goes way up because we manage to keep fighting anyway. Then we’re placed with weaker servers and our score tumbles again because we can’t destroy them as much as we apparently should.

So I don’t see much change overall, and I fear when the ranks “even out” we’ll be in the same situation we were before, always fighting the same two servers and that match up is not good week after week. But I might be missing something about the new system that will make it clear. Am I missing something?

The part you missed is all the posts explaining why ANet’s ruse wasn’t really going to fix anything at all. How anyone expected randomizing failure was going to make it any less of a failure is beyond me.