Not knowing the details of the present ranking system (Elo?) I am going to assume that it has the general property of taking into account prior scores to predict an expected server power (the score in the server rankings list). I haven’t been watching for long so please correct me if you disagree with that assumption.
To provide the best match-ups worlds are paired with their nearest expected powers. This has two potential benefits. First, servers are always provided with the closest thing to an even match as is available. And second, close match-ups are great if you want to decrease your uncertainty regarding your server rankings.
But – does that mean the match-ups will be fun to play? To answer that we need to define the personal utility functions of individual players. I can identify two sides to that utility function: “How am I doing in this week’s match-up?” and “How novel am I finding this experience?”.
“How am I doing in this week’s match-up?” – We have a very good measure of this one. The closer the score at the end of the week, the more net fun (net fun = total fun – total misery) being had. I believe they’re tuning the system to satisfy this factor, but in a flawed manner. The flaw stems from the fact that they’re not using their current expected power ratings to create a scenario where the expected game outcome is an even score. This means that even if they know with a 95% probability that HoD will win by 100K points, they will still create the pairing. Before we address how to fix that, let’s talk about factor two in player utility functions “How novel am I finding this experience?”
As far as novelty goes, I believe the current system is very lacking. Worlds are dividing into tiers, the ranking system certainty regarding the tiers is increasing, and within the tolerance of server flux and the ranking system’s memory we’re going to see a high correlation in the worlds we fight. (Granted these are moving targets, more on that at the end)
So how could we fix this? I’ve seen a number of good suggestions on the forum – buffing the losing team would help, but the timing and magnitude of those buffs would be a confounding factor for the ranking system and it would have little effect on novelty.
My suggestion has two parts – the first one is the hard one for ANet. You allow the system to generate built-in alliances, two worlds will split a score and share keeps/towers/supply camps – all of the typical restrictions apply but they share a side. This could take the form of WWvWW, or WvWW which would work with the current map. The second change is to create match-ups where the expected outcome of the game is as close to possible as even. That could be WvWvW (best for ranking), or one of these novel match-ups.
There are myriad benefits to this, but the two most salient are that it would seek out match-ups that were inherently the most fun and that it would allow for a huge increase the week-to-week novelty without compromising that balance.
I want to be clear that I’m not suggesting that the current system won’t work as a long term solution – I believe that there is a feedback loop where people/guilds in repetitive or bad situations will leave a server until that server’s status in the rankings changes. But, that requires people to become either bored or fed-up with their current situation. That’s not a good thing.