Guild Wars 2

Evan Lesh.3295:

When comparing two rosters, we first assume they are a perfect match with a score of ‘0’.

This gem is explained everything.
Most of people that read this will think that this is self explanatory. If both team are in mirror setup and plays equal the game will end at “no progress”.
But explain “no progress” you can by only one way – both team plays same, got same gear up and probably not dying at all. Even now Zero is a rare. Even in Bunker meta Zero was achieved not buy game balance but players errors and miss communication.

We got 9 professions and 9 elite specs for them.
The only way to do decent balance is a matchmaking that will.
1. Find equal skill players
2. Clash them
3. Make statistics out of it
If some profession combination have more popularity – need to calculate winrate vs rare appear combination.
If some profession combination got high impact(kills, captures, damage, heal, incoming damage) – need to calculate how systematic it is.
If something happens once – it can be never happened again, if something happened twice – it will happen again.
All statistics of semi-mirror setup is important!

Only statistic zero is something close to balance. That why players play BO3, BO5, BO7 etc…

Your quote is Evan talking about the matchmaking algorithm score (who is picked for a match), not the in-game score or anything related to progression/rewards.

out of context much?

With the context unknown, that statement means absolutely nothing. When I see it, I assume he’s talking about the match prediction code starting both teams at at equal value before proceeding to consider different aspects which would influence either team’s chance of success. This means both teams are considered equally. They start at the same value, and go through the same test, through which each condition has an effect on the initial value. To have them start at differing values before any tests indicate what they should be would be tantamount to using RNG for match result prediction...

Now if I’m right about what I assume he’s saying, I’d expect nothing less. If I’m wrong, it’s to be expected with a sentence taken out of context like that.

That quote came from the new matchmaking rules where someone said they needed to to remove the algorithm from the matchmaking to lower the chances of rosters having stacked professions. Evan explained, per request, why a change on the algorithm wouldn’t have such influence in the rosters and what changed comparing class stacking before and after the update to the rules (as to why they weren’t so abundant before as asked):

Evan Lesh.3295:

Sure! To get in the right mindset, it helps to understand the goal of the matchmaker a bit. When comparing two rosters, we first assume they are a perfect match with a score of ‘0’. We then reduce this score for things like how far apart the ratings are, difference in party sizes, etc. This way we can decide if one roster is a better choice than another when filling up teams because one roster will have a higher score.

Before and after the change, we reduce the score of a roster if it contains profession duplicates over the desired(configurable) maximum for a team. However before the change, we were also increasing the score of a roster if it had professions that were not yet on the team. They were both trying to accomplishing the same goal.

The are three problems with the second part; the piece we removed. First, increasing the score of a roster effectively cancels out some of the negative scoring we added for other parameters. For example, a roster could have a poor rating match, but ‘oh well, they have unique professions!’ Second, the score was erroneously given to the first roster added to a team, so each profession was guaranteed to be considered unique and result in a score increase. Thirdly, the matchmaker has no choice in picking unique professions when dealing with large premades, but their rosters were still getting huge score increases because most of the time it had 5 unique professions. All combined, this meant premade rosters were getting a score increase that cancelled out most if not all negative aspects of the match up.

After this change we’ve seen the win rates of larger premades drop a noticeable amount. Now that rating differences aren’t being lost in scoring due to profession bonuses, we can see that the party rating boost was too high (which we’ve now lowered).

I’d like to say that this change was a direct result of a screenshot posted to the forums. While investigating the details of the match, I saw that a premade had a wildly abnormal score in relation to the other rosters. So thank you for the feedback!

https://forum-en.gw2archive.eu/forum/game/pvp/New-Matchmaking-rules/first#post5952756