69 replies to this topic

[GARION26]

VonBruinwald, on 25 January 2021 - 06:33 AM, said:

That's the biggest problem, groups break the MM as it cannot balance group PSR within teams. A 4-man of potatoes will always drag the team down and a 4-man of try-hards will always have an advantage unless there happens to be an equivalently skilled group on the opposite team.

So to my understanding from what they said about matchmaker changes when they introduced groups - matchmaker tries to match on the following criteria
1) Balance number of players in groups on both sides of the match
2) PSR
3) Tonnage slots
4) Wait time

When they introduced groups they made item #1 (equal groups on both sides) the first thing matchmaker looks at.

That's in my opinion the problem as you note the system weighs a group of potatoes the same as a group of elite players.
Their effect on the match is entirely opposite in direction however.

Move PSR up to #1 in the matchmaker algorithm would fix that. But you might as a result sometimes have 2 groups on one side and none on the other which would be unpalatable to folks who feel all groups are an advantage. An additional option would be to give people in groups a small PSR boost temporarily while in a group.

I think however in terms of maximizing match quality PSR is more likely to predict outcome then group vs non group and by making it first in the algorithm we'd have better outcomes.

I'm a tier 4 who likes to play group games with my ten year old tier 5 son. We are clearly not the same as even a group of two tier 4's in terms of positive match impact.

Edited by GARION26, 25 January 2021 - 09:45 AM.

[Vlad Ward]

GARION26, on 25 January 2021 - 09:44 AM, said:

So to my understanding from what they said about matchmaker changes when they introduced groups - matchmaker tries to match on the following criteria
1) Balance number of players in groups on both sides of the match
2) PSR
3) Tonnage slots
4) Wait time

When they introduced groups they made item #1 (equal groups on both sides) the first thing matchmaker looks at.

That's in my opinion the problem as you note the system weighs a group of potatoes the same as a group of elite players.

Can you back this up with anything? I hear it all the time but it defies any and all semblance of reason.

The only piece of evidence pointing towards anything resembling this is a post from Paul saying the MM will 'Try to balance groups between teams.' I can't find anywhere where a Dev or Design team member has said that the MM will prioritize groups at the expense of everything else.

[GARION26]

Paul references it here when they introduced group queue


https://mwomercs.com...ue-update-2020/

" Group team balancing

The match maker will put groups on opposing teams as much as possible. If there's only 1 group in queue, there's only so much the match maker can do. However, if there are two groups in queue (no matter what the size... and another reason why 4-man max makes a lot more sense) it will put one group on team 1, and the other group on team 2."


You can see it doesn't appear to pay attention to tonnage slot or PSR when looking at the groups- just lets get a group on both sides. There may be a more nuanced background sort in the matchmaker but it isn't in what he's written that I've seen.


If the population is high enough there are likely many groups in the queue so they can be sorted to better balance on the PSR, tonnage slot and wait time criteria, but those are secondary, tertiary and quaternary criteria it appears. So it's not that they can't use them to balance teams - but it looks group first and then tries to balance everything else.

The historic 3 metric's being balanced by the MWO matchmaker schema shown nicely here as well as how they can change the weighting of different factors. It really is a nice visualization of how they manage the matchmaker and how they can tweak it.
https://mwomercs.com...67#entry6334567

But that graphic doesn't reference his previous group vs non group matchmaker criteria addition which I linked to above. I suspect that's because it's the standard visualization tool they have used historically and they didn't think to add a fourth corner to the diagram based on matchmaking on group vs non group.

Edited by GARION26, 25 January 2021 - 11:01 AM.

[VonBruinwald]

GARION26, on 25 January 2021 - 09:44 AM, said:

So to my understanding from what they said about matchmaker changes when they introduced groups - matchmaker tries to match on the following criteria
1) Balance number of players in groups on both sides of the match
2) PSR
3) Tonnage slots
4) Wait time

When they introduced groups they made item #1 (equal groups on both sides) the first thing matchmaker looks at.

That's in my opinion the problem as you note the system weighs a group of potatoes the same as a group of elite players.
Their effect on the match is entirely opposite in direction however.

Another thing of note is 2 and 3 only apply to the solo players. The only "balance" criteria for groups is size.

Effectively the MM is making a 4v4 group match and a separate 8v8 pug match, mashing them together and hoping for the best.

[Elizander]

I think all this group talk is a bit off topic for the post though. I personally don't mind having groups in the queue. It's no worse than the MM opening up and putting better T1s on one team compared to mine.

I've had the top-top players on Jarl's list on my team in a group of 4 and we still lost games so whatever. Being 0.001% on Jarl's + having a 4 man team doesn't guarantee victory.

I suppose the only annoying thing is that losing PSR will upset most people. It's just compounded by the fact that there is no modes where we can just level mechs and not worry about being rated by a system after every game. You know, gamers don't like it when their rating goes down, even if it might not be important. That's just how things are. People who are upset by it should turn it off, but I just want to say that I don't particularly enjoy killing cadets which is why I still crawl up to T3 now and then to avoid hitting T5. There's no satisfaction fighting someone who gives me 8 seconds to line up a headshot without shooting at me.

Edited by Elizander, 25 January 2021 - 11:27 AM.

[GARION26]

VonBruinwald, on 25 January 2021 - 10:54 AM, said:

Another thing of note is 2 and 3 only apply to the solo players. The only "balance" criteria for groups is size.

Effectively the MM is making a 4v4 group match and a separate 8v8 pug match, mashing them together and hoping for the best.

It's staged criteria based on the links I gave. If the population is big enough they can get all the criteria right in a drop, if the population is low then it makes sure group vs non group is factored in first and then does it's best on the later criteria.

So I don't think it's that groups don't count towards PSR or tonnage slots balancing at all - it's just treating those two issues as of lesser importance then getting at least the same number of grouped players on both teams.

See my previous post on my opinion on how to deal with this (make groups vs non group lower priority then other items in the matchmaker.)

[Vlad Ward]

GARION26, on 25 January 2021 - 10:23 AM, said:

Paul references it here when they introduced group queue


https://mwomercs.com...ue-update-2020/

" Group team balancing

The match maker will put groups on opposing teams as much as possible. If there's only 1 group in queue, there's only so much the match maker can do. However, if there are two groups in queue (no matter what the size... and another reason why 4-man max makes a lot more sense) it will put one group on team 1, and the other group on team 2."


You can see it doesn't appear to pay attention to tonnage slot or PSR when looking at the groups- just lets get a group on both sides. There may be a more nuanced background sort in the matchmaker but it isn't in what he's written that I've seen.


If the population is high enough there are likely many groups in the queue so they can be sorted to better balance on the PSR, tonnage slot and wait time criteria, but those are secondary, tertiary and quaternary criteria it appears. So it's not that they can't use them to balance teams - but it looks group first and then tries to balance everything else.

The historic 3 metric's being balanced by the MWO matchmaker schema shown nicely here as well as how they can change the weighting of different factors. It really is a nice visualization of how they manage the matchmaker and how they can tweak it.
https://mwomercs.com...67#entry6334567

But that graphic doesn't reference his previous group vs non group matchmaker criteria addition which I linked to above. I suspect that's because it's the standard visualization tool they have used historically and they didn't think to add a fourth corner to the diagram based on matchmaking on group vs non group.

See. I knew about that quote you referenced, but I think you have (and everyone else has) wildly misconstrued it.

Setting aside the dumb crap people repost 24/7 on this board for a minute, in what sane realm does "Yeah we'll try to keep a group on each team when we can" translate to "As long as there are two groups in the queue, they will be matched against each other always period without any other considerations at all" ? That's insane. We all know it's insane. Paul is not that insane.

That post was also the first draft/proposal of the system. Not the outline of the system in practice. For that, see this post: https://mwomercs.com...es-4-week-test/

Where Paul says: "How the combined public queue will work:
- All matches will remain 12v12.
- Max group size will be restricted to 4 players. All group sizes of 2,3 and 4 are valid group sizes.
- When the Match Maker builds a 12-player team, it will ensure that there is no more than one 4 player group or as close as possible (2 player + 2 player for example) per team and all remaining slots will be filled with solo players."

This is a smart approach and does not suggest anything remotely like "T5 groups match against T1 groups because they're groups"

VonBruinwald, on 25 January 2021 - 10:54 AM, said:

Another thing of note is 2 and 3 only apply to the solo players. The only "balance" criteria for groups is size.

Effectively the MM is making a 4v4 group match and a separate 8v8 pug match, mashing them together and hoping for the best.

According to whom??????

Edited by Vlad Ward, 25 January 2021 - 11:26 AM.

[VaudeVillain]

I don't play this account as often as my main account, but before the PSR reset, I managed to get to tier 2 and every 'Mech is a "meme" 'Mech because I only use Level 1 tech 'Mechs, with no 'Mechs built after 3030. Don't confuse this with stock 'Mechs as I will definitely modify them, just no new tech and only Arty or Air strikes, since any other consumable would be beyond the technology capabilities back then. So, even with crappy 'Mechs, you can improve your PSR, but it takes a while.

[GARION26]

Vlad Ward, on 25 January 2021 - 11:17 AM, said:

See. I knew about that quote you referenced, but I think you have (and everyone else has) wildly misconstrued it.

That post was also the first draft/proposal of the system. Not the outline of the system in practice. For that, see this post: https://mwomercs.com...es-4-week-test/

Where Paul says: "How the combined public queue will work:
- All matches will remain 12v12.
- Max group size will be restricted to 4 players. All group sizes of 2,3 and 4 are valid group sizes.
- When the Match Maker builds a 12-player team, it will ensure that there is no more than one 4 player group or as close as possible (2 player + 2 player for example) per team and all remaining slots will be filled with solo players."

This is a smart approach and does not suggest anything remotely like "T5 groups match against T1 groups because they're groups"

Appreciate the critique on my reading of Paul's post. You certainly may be right. It's clearly weighted criteria for the matchmaker with various weights given to different components that they can tweak based on their other other posts.

The April 20th post implies group vs non group is the primary weighted item. You are right subsequent posts don't mention that (but they don't tell us where it is in the criteria) so it may be lower down.

Irrespective of where it's in the weight I don't think the crazy stuff happens like tier 5 groups in a tier 1 game often (if at all) because the population is big enough that there are more group options to matchmake then just two on the opposite of the PSR spectrum.

[Vlad Ward]

GARION26, on 25 January 2021 - 11:47 AM, said:

The April 20th post implies group vs non group is the primary weighted item.

I think even this requires a very strict reading of the phrase "whenever possible." It requires you to interpret this to include matching a T1 group with a T5 group when they are the only groups in the MM. An alternative reading would be "When all other conditions are met," which would instead imply that a group of 24 players that have been chosen by the MM to form a match because they have tight PSR grouping would then have the groups allocated to different teams.

GARION26, on 25 January 2021 - 11:47 AM, said:

Irrespective of where it's in the weight I don't think the crazy stuff happens like tier 5 groups in a tier 1 game often (if at all) because the population is big enough that there are more group options to matchmake then just two on the opposite of the PSR spectrum.

You might not, but plenty of people around here do and it's disruptive.

Edited by Vlad Ward, 25 January 2021 - 12:01 PM.