9 replies to this topic

[Mawai]

Hi,

In the following thread the developers described the concepts behind matchmaking stage 3:

http://mwomercs.com/...79-matchmaking/

The matchmaker takes the average ELO of the team into account to determine gain or loss of ELO as a result of a match. It uses ELO to estimate the odds of winning or losing.

However, from my experience, the determining factor in winning and losing most matches is not so much the individual skill alone but the communications and coordination of the players acting as a team. It is the GROUP dynamic that is the dominant in determining the effectiveness of the team rather than the individual player skill. i.e. folks dropping in pre-mades with TS are much more effective than the same players dropped separately.

If the matchmaker does NOT take grouping into account in balancing matches then it will fail.

If you have two teams with identical ELO ... one grouped and one not ... the one which is grouped will usually ROFL stomp the other team. The odds of winning are not even in this case as the matchmaking algorithm described in the thread above would imply.

This might be addressed by adding an effectiveness multiplier to the ELO of players who are grouped. This multiplier could be smaller for two man groups and larger for 8 man groups ... but if the effect of groups is ignored in favor of an ELO system that does not take into account whether a player is grouped or not in a specific match ... then it can't work.

Here are some other examples:

1) A player plays a lot of grouped matches and has a very high ELO decides to drop in a PUG. This high ELO value does not reflect the player skill ... it reflects the effectiveness of the team he has been dropping with. Using this high value of ELO will unfairly impact the others on this team.

2) A player who plays mostly PUG may have a lower ELO but when they occasionally join groups this low ELO value will underestimate their effectiveness as part of a group.


The effect of being in a coordinated pre-made group can be more significant than the individual player skill ... thus calculating an individual player skill based on the outcome of a match without taking this into account can not work without significant modification.

Possible suggestions to address this issue:

- introduce two ELO values for each person - one for group play and one for individual or solo play
- introduce an ELO multiplier for grouped people to reflect the increased effectiveness of people in groups in order to generate more balanced matches.

[superteds]

No idea. This bacon sandwich is really good though!

[Roland]

superteds, on 08 January 2013 - 07:36 PM, said:

No idea. This bacon sandwich is really good though!

Bacon sandwiches are OP, and only n00bs eat them.

[TheTrap]

Excellent concerns that should be put in place. Thumbs up.

[Titan Osis]

No, no it isn't because it will keep the bad players teaming up against other bad players until they realize they are bad and its not just the game. Then they will want to do something about it and get better at which point they will start playing with better players and get better.

[Solis Obscuri]

Is the bacon in the sandwich, or is the whole sandwich bacon (e.g. bacon sandwiched between more bacon)?

[BlightFang]

Good points. Although I don't think just having an averaged elo is flawed as its quite better than the current non-elo system just as that. Sure it can be abused in some ways(like having someone take a new account to so you can pugstomp easier) but you don't have need to do any manipulation to "abuse" in the current system.

I believe League of Legends does some of the tricks you mentioned to help balance a bit better, but that can be adjusted with time. They could probably even hire some consultation from riot games to accelerate this process since those guys have gone pretty deep into the subject.

[superteds]

Solis Obscuri, on 08 January 2013 - 07:45 PM, said:

Is the bacon in the sandwich, or is the whole sandwich bacon (e.g. bacon sandwiched between more bacon)?

Awww man that sounds like a dream. This example is (was) bacon (lots of) sandwiched between two (very soft) slices of white bread, and lathered in (brown, red is for heathens) sauce.

I think the only thing that could have further improved on that experience would be some nice cheddar, but it's nearly 4am and that was that little step too far in terms of the effort:food ratio!

[TheMagician]

PUBs are lame, and nobody cares who wins those. Rankings will even out long term. Once merc corps come in, it would be sort of nice to get a ranking, but then how is a single ranking reflective of any group of 8, if you have 200 members?

[HC Harlequin]

People who don't play in groups will tend to score lower on average than people who do. Since the ELO system is team based.. what's the issue?