Matchmaking algorithm suggestions

[Deadmoose.6594]

What they should do is start out by looking at the community rankings of class/specs. S-tier, A-tier, etc. When it comes to matchmaking, there should be an algorithm that factors in the spec ranks. This clearly isn't being done right now. The game should then start recording overall win/loss ratios of all of the specs to verify the community rankings. The game should look at individual win/loss ratings. Lastly, the rock-paper-scissor affect of classes should be factored in if possible. I'm not smart in this area but someone posted about it recently, all I know is thiefs take out mesmers (because I player mes). Memsers take out warriors, etc. etc. So in summary you would get:

 

-Community-influenced spec tiers being factored into matchmaking, because clearly Anet needs help in this area

-Matching an S-tier spec on both teams, circumventing the constant need for nerfing OP specs. Additionally, you could have things like two A-tier classes matched against an S-tier if two S-tier's aren't available. If only one S-tier is queueing and no A-tiers, pair S-tier with the lower ranking specs and have medium specs on the other team. There's a lot of balancing that could be done here

-Machine learning win/loss ratios of specs after each balance update, with past balance data being deleted to allow the process to start over again

-Factoring in individual win/loss ratios which can be used to validate player skill. Example, a B-tier player that tears it up in spvp can be factored higher than an A-tier that underperforms

-Factor in rock-paper-scissors affect of class specs.

 

I was thinking about this when looking at my game history (I only play Conquest), ideally I would think you want to have a match be at least 500 vs 300+ when all is said and done. Ending a match with at least 300 points means everyone tried, but they were bested for whatever reason. Tbh, you really want to aim for 500 vs. 400+, those are the good matches that could've gone either way. 300+ just means one team was clearly better, there was an MVP, or the other team just didn't play together well or the matchup wasn't great. If the matchmaking were better, we could avoid stomps where the outcome is 500 vs <100. Overall I would say it succeeds in this area, but there are some games where you're like, this is complete bs and it leaves you salty about everything: Anet's matchmaking, OP specs, your incompetent teammates, and your underperforming class for whatever reason.

[Deadmoose.6594]

One thing I wanted to add, if more time and effort were placed on spvp, more people would play it, which would mean the algorithm would become smarter and the matches would be overall more fun I think. Anet would want to sort of communicate that they were revamping their matchmaking algorithm to entice players that it can be a fun and enjoyable experience (I have fun but I like competitive play, being salty from time to time is just part of the experience imo).

 

The algorithm could also learn from negative-experience matches, and really it should be learning from those matches. If there was a stomp match, it should get flagged and it should review more in-depth data to try to ensure that the next match involving the players in question is more balanced.

[Deadmoose.6594]

Just fought against two guardians, should literally never happen.

[Disney.7826]

I duoq as double firebrand and get to play vs core eles, fbs… I reroll.

 

UR THEORY IS BS, DID U RLY NEED TO POST THAT

Edited January 22, 2023 by Filip.7463

[Deadmoose.6594]

Additionally, the matchmaking system should show a grade of how well it matched up players. If there's not enough players to queue, it'll do the best it can and it should show a grade to reflect that it won't be a good matchup. If the underdogs win they should get an increased reward.

[Disney.7826]

1 hour ago, Deadmoose.6594 said:

Additionally, the matchmaking system should show a grade of how well it matched up players. If there's not enough players to queue, it'll do the best it can and it should show a grade to reflect that it won't be a good matchup. If the underdogs win they should get an increased reward.

It tries to match same classes, NOT SPECS. U know its bad matchup if the q is 4+min. People with lower rating than opponents average will get an increased reward (MORE RATING). Why else do you think sometimes its +10, sometimes +15? Because top stats damage?

 

There is no enough players in Q at any time of day for more than past 1 year. What you get now is the best it can be.

Matchmaking tends to make same TEAM RATING AVERAGE on both sides, not per player.

 

For example; average of one team is 1050mmr, that means the other one can be made out of 5x 1050mmr or 1x 1250mmr + 4x 1000mmr.

(For example) Base rating gain / lose is 10 points. Players below the average of opponent team will get +11 / -9. Above average will get +9 / -11.

 

Matchmaking tends to put same comp on both sides, but having similar team average has priority. You can not track elite specs, only classes.

Neither the game can recognise if someone is guard sup with avatar amulet or he is bunker guard with avatar amulet.

 

Unproven fact we believe is that matchmaking will after a certain time either allow bigger team average deviations OR spread maximal allowed mmr difference between highest and lowest player in match. 
 

We used to believe that we should req after 4min but i don’t think anyone really cares about it at this state of game.

Edited January 23, 2023 by Filip.7463

[Dawdler.8521]

Why not just have 5 random mirrored class setups with preset builds and skills, no options to change. First come first serve for the pick. Basicly like an Overwatch version of GW2.

Skips all the extra steps with rules and algorithms since this appears to be the endgoal for most suggestions.

Edited January 23, 2023 by Dawdler.8521

[shion.2084]

So my votes are for....

1) if you are going to allow duo q, then both players are assumed to be the higher ranked players rating.  The advantage of duo q, should be offset by considering them higher in ranking.

I'd prefer there was no duo q, but at least rank both players as the highest players rating.

2) I'd allow users one free non-penalised DC per period of time (so they don't come back and screw their team over, plus life happens).

3) A player should be able to set a match making criteria where they are only interested in playing matches with players whom are within two tiers of them (P1 will be matched down to G2 maybe).  Something like that, the exact spread can be tweaked.  They would be understanding that they simply might not get a match if they did this.  

IT is silly that when I get up to spot 74 I'm placed with some silvers I don't recognise against decent Gold2/3's that I recognise.  I want to be able to say G2 or higher players, or don't give me the match.

4) A player should be able to request match maker not put them with duo q's, on pain of simply not getting a match.

5) Getting 4 top scores should result in no loss of rank (and not highest deaths).

6) Rank loss should be graduated by the degree of the loss.  You lose the current rank levels if your game is lost by 200 points, and less if the match is closer (wills top folks from just yelling GG the loudest after the first mid fight is resolved).

7) I should be allowed to switch my build 1 time per game while dead.

8 ) In lower tier rank games, to encourage new players there should be some single hit caps for damage, some code around CC immunity for a short time after being CC'd, A buff to toughness after respawn which decays over time.   Make this friendly for noobs.

[Trevor Boyer.6524]

Honestly it's more about the amount of match manipulation going on, not problems with the algorithm.

Hard to believe, but it's completely true.

Anyone telling you otherwise is helping sweep under the rug.