Is the random roll bugged?

#1 - June 8, 2013, 9:34 a.m.

I’ve been trying to think through how ranks 21 (DR) and 22 (HoD) got matched up with a 14 (BP). If you look at the raw tiers, this is a +/- 3 tier spread. BP is tier 5, DR is tier 7 and HoD is tier 8. The description of the new system states that BP should never be matched up with a tier 8 because that’s three tiers away.

However, if you look at the raw server rankings and break THOSE up by tiers, you’ll see this:

14 Borlis Pass
15 Anvil Rock
16 Darkhaven
—-
17 Isle of Janthir
18 Gate of Madness
19 Sorrow’s Furnace
—-
20 Northern Shiverpeaks
21 Devona’s Rest
22 Henge of Denravi

When you just look at the tier distribution based on static triplets, HoD is top t8 and BP is middle t5, and they’re four tiers apart (making it +/- 3). If you look at it this way, on a strict numerical distribution, HoD and BP are polar opposites in a three tier spread, making it viable for a +/- 2 tier distribution match up like what happened.

There is no mention that I can see in the official blog post about whether or not this spread is based on raw rank or on the existing tiers.

If this is as designed, +/- 2 is too high of a spread in my opinion. I’m sure HoD/DR isn’t the only pairing that is getting stomped so far this week.

DevonCarver

#3 - June 8, 2013, 9:58 a.m.

The concept of tiers doesn’t relate to how the matchups are generated. They are generated based on server ranking. While 1 tier may be next to another, they may have a significant gap in rating or they may have an insignificant gap. That is how the matchups are rolled. If the servers are rated close enough to each other, they can end up matched up against one another.

DevonCarver

#8 - June 8, 2013, 10:18 a.m.

One up and one down solves some problems but creates a bunch more. Including that it invalidates ratings entirely. It no longer takes into account quality of victory but simply uses the fact of victory to move a server out of the matchup they were previously in. The system we are using now creates varied matchups, which will allow us to get more accurate ratings. As those become more accurate we can dial down on the amount of variance we are introducing to have matchups that are varied, but less prone to large rating gaps between the servers. But, again, this is entirely based on server rating, which is the thing we have actually been calculating this entire time. Until that number is a more accurate reflection of server quality, we are going to keep mixing the matchups up.