125 replies to this topic

[Crunk Prime]

10-12 player groups can sometimes end up with a huge disadvantage.

Large groups are restricted to the 3/3/3/3 rule, while a team of smaller combined groups can easily end up with 4 assault mechs and 4 heavies. Ive even seen a team in group drops end up with 4 assaults and 5 heavies, while my team was stuck with 3/3/3/3

Ive seen units abuse that and hope to synchdrop together with their heavy and assault stacked teams.

[Mawai]

Wraeththix Constantine, on 21 September 2014 - 03:32 PM, said:

The MM works OK, other than it's a bit too eager to loosen restrictions to speed up matches so you can get to being rolled faster by 12-mans.

The larger issue has to do with how ELO is computed and/or used. I'm not sure what the calcuation does, but it appears to be based loosely off KDR and W/L and damage done. Which means if you're a newer player, not only do you have to deal with the "Tax" which is everyone else using strikes and UAVS every single match and having elited out mechs while you have some crap mech with no modules, in addition if you accidentally do well for a few games, you spend the next 2 weeks in hell.

Time played and total XP across all mechs/GXP should be the largest determinate. Newer players should be matched with newer players, and veterans should be matched with veterans.

As it is, right now I get stuck in matches with people playing for 2 years. Despite my founder's tag, I've only actually been playing for two weeks. So I basically get the "priviledge" of losing over and over and over and over again because a.) I'm not that good and b.) because the game is CONVINCED I'm awesome because the ELO calculations somehow place me on par with people from Lords, 226, ACES, etc.

And if losing repeatedly wasn't "fun" enough, I get the extra bonus of getting insulting comments directed at me from all the veterans. And honestly, they're probably justified in their anger (if not maybe the verbage), because why am I even in their games?

For a game that has on-boarding issues, let me suggest that getting yelled at every single match and losing repeatedly does NOT really entice someone to want to spend money on your game.

"I'm not sure what the calcuation does, but it appears to be based loosely off KDR and W/L and damage done."

There are a number of posts all over the forums explaining how Elo is calculated in MWO ... personally, I think the devs should have a FAQ for these sorts of things ... though there are some player resources with lots of useful information.

One example: http://mwomercs.com/...ost__p__3057938

In brief, the matchmaker takes the Elo scores of all players on both sides. The side with the higher Elo score is expected to win and the difference in Elo scores between the teams defines the probability of that happening.

The match is then played.

If the team that was expected to win actually wins then NO Elo scores are changed.

If the other team wins then ... the Elo of players on the losing team drops up to 20 points ... the actual number depends on how much they were expected to win by ... similarly the Elo of players on the winning team goes up in a similar fashion. The bigger the upset, the bigger the change in Elo.

So Elo scores do not depend at all on W/L ratio, KDR or damage done.

However, new players are seeded somewhere near the center of the range with a handicap that drops off the more games they play. This is part of why new players end up in groups with experienced players ... the bulk of players end up near the middle of the distribution ... that is the whole point ... but most of these players have been around for a lot of matches and even with the handicap new players will often find themselves in matches with folks from the center of the distribution.

The other reason that new players end up in groups with experienced players is that over time, in order to form a match, the matchmaker widens the Elo band if it can't find enough folks for the match it is making. This means that you can occasionally pull folks from the middle into high Elo range matches ... AND ... some of those middle of range folks could have played 50 or less matches (relative new player) ... since by that point their handicap will likely be gone but it can take quite a few games for Elo to settle down to an appropriate value. In addition, you have a different Elo for each weight class ... so when you start a new weight class as a newer player you will likely start near the middle of the range without a handicap until you get your Elo established in that weight class.

Finally, one more factor, the Elo calculation doesn't take into account the loadouts or experience you have in a chassis. It doesn't know that you are less effective when the skills haven't been unlocked ... particularly cooling, speed tweak and doubled basics. So most folks when they start leveling a new chassis will have their Elo drop until they at least Elite it ... and then start going back to where it should be after it is mastered.

There is a lot of contention about the use of a variant of Elo in a multi-player team game ... something like Microsoft's TrueSkill system might be better but it would cost money to license. However, over a large enough sample set Elo should converge to give a reasonable estimate of your average contribution to winning a match as part of a larger team since all of the outlier events that folks complain about should average out in the long run.

P.S. It is never right for folks to yell at a person for their perceived performance in a game. I have had some matches where my performance was pitiful and it had nothing to do with how many games I have played, experience, or Elo ... it usually has to do with bad luck or a bad decision ... which everyone can make no matter how high or low their Elo. The folks who blame others for the failure of the team ... obviously don't know what it means to play on a team ... those were probably also games with very little useful team chat.

Edited by Mawai, 21 September 2014 - 05:23 PM.

[Kiiyor]

The devs have alraedy posted about this very problem. They cite the number of 'hard stops' in the MM, and are suggesting adding the game mode choice as a vote, rather than a hard stop. This means that 12 mans and the like will be more likely to meet another 12 man that may have selected a different mode.

[oldradagast]

Wraeththix Constantine, on 21 September 2014 - 03:32 PM, said:

The MM works OK, other than it's a bit too eager to loosen restrictions to speed up matches so you can get to being rolled faster by 12-mans.

The larger issue has to do with how ELO is computed and/or used. I'm not sure what the calcuation does, but it appears to be based loosely off KDR and W/L and damage done. Which means if you're a newer player, not only do you have to deal with the "Tax" which is everyone else using strikes and UAVS every single match and having elited out mechs while you have some crap mech with no modules, in addition if you accidentally do well for a few games, you spend the next 2 weeks in hell.

Time played and total XP across all mechs/GXP should be the largest determinate. Newer players should be matched with newer players, and veterans should be matched with veterans.

As it is, right now I get stuck in matches with people playing for 2 years. Despite my founder's tag, I've only actually been playing for two weeks. So I basically get the "priviledge" of losing over and over and over and over again because a.) I'm not that good and b.) because the game is CONVINCED I'm awesome because the ELO calculations somehow place me on par with people from Lords, 226, ACES, etc.

And if losing repeatedly wasn't "fun" enough, I get the extra bonus of getting insulting comments directed at me from all the veterans. And honestly, they're probably justified in their anger (if not maybe the verbage), because why am I even in their games?

For a game that has on-boarding issues, let me suggest that getting yelled at every single match and losing repeatedly does NOT really entice someone to want to spend money on your game.

While the matchmaker has issues, and the onboarding process in this game is rather awful (though the game itself is fun), there's just no excuse for treating another player that way, new or otherwise. It's not as if you're TRYING to kill your team-mates, throw the match, or whatever.

For what it's worth, I'm sorry you've had to put up with that, and there are plenty of folks out there in the game who don't behave in such a manner. I hope you run into more of them in your games vs. the "other kind" of people.

[FatYak]

with MWO's playerbase you can either have a great MM and long long wait times for a match, or, a not so good matchmaker and short wait times

We cant have both

[Kiiyor]

FatYak, on 21 September 2014 - 05:14 PM, said:

with MWO's playerbase you can either have a great MM and long long wait times for a match, or, a not so good matchmaker and short wait times

We cant have both

I think we can, if Russ goes ahead with having the game modes as a guideline instead of a hard stop.

This all depends on the number of people playing only one game mode, however.

[Project_Mercy]

Mawai, on 21 September 2014 - 04:14 PM, said:

"I'm not sure what the calcuation does, but it appears to be based loosely off KDR and W/L and damage done."

There are a number of posts all over the forums explaining how Elo is calculated in MWO ... personally, I think the devs should have a FAQ for these sorts of things ... though there are some player resources with lots of useful information.

One example: http://mwomercs.com/...ost__p__3057938

Thank you for the explanation, it makes more sense.

Or, at least I understand what you're saying, the design doesn't make much sense.

Solo queue really is like a wave. You win, you lose. If you start in the middle, and it keeps trying to adjust your ELO based on yoru win/loss, and your win/loss has little to no baring on what you actually accomplished, how exactly does that result in an interesting or fulfilling match? For me at least, it does not. It makes you feel like the whole process is a huge waste of time, and you might as well just YOLO charge in the beginning and then AFK waiting for the result.

I see people do that regularly, and I never understood why until now.

If they're going to have a ranking system, shouldn't it inherently be based on your personal ability and experience?

Edited by Wraeththix Constantine, 21 September 2014 - 05:42 PM.

[IceSerpent]

Mawai, on 21 September 2014 - 02:27 PM, said:

Interesting.

I haven't seen any numbers on group drop rate
I haven't seen any numbers on the number of groups of various sizes dropping at different times of the day
I haven't seen any numbers on selected game modes for groups
I haven't seen any numbers on large group drop rates
I haven't seen any numbers on the mech class distribution in groups
I haven't seen any numbers on the ELO distribution in groups
I don't have any idea how the release valves work in group match making ... whether it can relax mech composition, ELO, or groups size for opposing teams, which of these gets released and how fast.

However ... this is all the information needed to make a judgement about whether the current matchmaker "is doing the best it can".

Your sarcasm is misplaced - while one would need those numbers in order to fine tune the MM, we don't need them for determining that MM is not doing the best job.

A situation where full 12-man group gets matched against 3 x 4-man groups is only justfiable when no other matching can be done. In other words, there are no groups in queue larger than 4 that can be used and there are not enough groups of 4 and smaller to make another team of 12. More than that, it implies that situation doesn't change (no more groups join the queue) during a fairly long period of time, i.e. at least tens of seconds. This is an extremely improbable situation.

Mawai, on 21 September 2014 - 02:35 PM, said:

LOL

You know ... this is pretty much exactly how the group matchmaker already works

http://mwomercs.com/...ost__p__3696769

When you finish LOLing, try come up with an explanation how you can possibly get a 12-man vs. 6 x 2-man in a MM that "alternates adding to each team, as long as the size of the groups being added are roughly equivalent", unless you think that size of 2 is roughly equivalent to size of 12.

Quote

However, the group matchmaker is not just doing group size but mech weight class balancing and ELO as well. When these other factors are included the process is not nearly as simple in terms of what groups are available in the queue. Do you make matching large groups the priority and ignore ELO and mech class?

No, I would do the exact opposite - only fetch players (groups) from a proper Elo bracket, but fixing Elo is a different topic, so for now I'd settle for at least fixing the group sizes.

[Mystere]

Kiiyor, on 21 September 2014 - 05:25 PM, said:

I think we can, if Russ goes ahead with having the game modes as a guideline instead of a hard stop.

This all depends on the number of people playing only one game mode, however.

Forcing people to play skirmish might not be a good idea (to put things ever so politely).

[Mawai]

"More than that, it implies that situation doesn't change (no more groups join the queue) during a fairly long period of time, i.e. at least tens of seconds. This is an extremely improbable situation."

All sarcasm and joking aside, I have no idea whether this is an extremely improbable situation or not.

I don't know how many large groups try to drop in MWO, with what regularity, into what Elo brackets and with what mech distributions.

Without that information, I have no way to say whether the matchmaker is working well or not. IF large group drops were frequent across all Elo brackets and all game modes ... then I don't think the matchmaker would have any trouble matching large groups against each other (you seem to think otherwise ... and instead seem to think that there must be sufficient players in all bins that any failure to match large group vs large group is a matchmaker issue).

The problem is that without the analytics there is no way to tell who is correct and whether the matchmaker is doing the best it can or not. That is really all I am saying (though I admit sarcastically )

[White Bear 84]

Jackpoint, on 21 September 2014 - 12:35 PM, said:

Anyone actually thought it might be the player base that's shrinking, ergo you will get 12man drops versus pugs or 2-3 player unit drops. That's my theory anyway after todays few games, 12 games 1 win.

Have considered this.

Rather than deal with the QQ of 'your team should have played better' blah blah blah I am actually collating screenshots from nights of play in groups of different sizes, from pug rounds to 4 man groups. Much better than just shouting 'MM is broken'. I will profoundly show that it is not balanced.

After the first night of capture the ration went something along the lines of 12-0, 12-3, 12-1, 12-4, 12-2, 12-2, 12-8.

If it is due to low player numbers spreading the elo, then maybe the MM needs to be tweaked to accomodate for low player numbers?

[IceSerpent]

Mawai, on 21 September 2014 - 07:12 PM, said:

Without that information, I have no way to say whether the matchmaker is working well or not. IF large group drops were frequent across all Elo brackets and all game modes ... then I don't think the matchmaker would have any trouble matching large groups against each other (you seem to think otherwise ... and instead seem to think that there must be sufficient players in all bins that any failure to match large group vs large group is a matchmaker issue).

Mawai, the part you're missing that number of large groups should only be relevant to how fast those large groups get a match. To put it simple, a large group might need to wait for another large group to queue up. This has very little bearing on what happens to small groups, as they can be matched as fillers for larger groups and they can be "large" groups themselves. So, all we really need is sufficient number of players in groups of 6 and smaller. Larger groups having to wait a bit can't be helped - you can't conjure another 12-man (or 10-man) out of thin air. On the other hand, smaller groups getting a fair match can be helped, and should be.

What currently seems to be happening in a situation where you have a 12-man and (let's say) 12 x 2-man in the queue (assuming same game mode and Elo bracket for all those players) is that 12-man goes into team A and gets matched against 6 x 2-man on team B. The other 6 2-man groups are still waiting in the queue.
What should be happening in this scenario is 6 x 2-man go into team A and are matched against the other 6 x 2-man on team B, while 12-man remains in the queue due to the fact that it's the group that doesn't have a proper opponent at the moment.

[JP Josh]

Einaescherin, on 21 September 2014 - 12:08 PM, said:

3*4 player grps vs 12 player grp
or
10player grp + 2 player grp vs 4 *3player grps

Is it to difficult to make the teams even in the matchmaker?
Then don't implement futures your not able to handle!
4 times in a row get rapped by a 12 player grp cuz of low developer skill is disgusting!!!

im gald you wernt a part of closed beta that being said iv had pugs win out against fully formed 12 man death squads.


also fun to fight the 12 man squads that are messing around.... 11 wangs and 1 awsome. 12 brawler atlases with machien guns and big engines. ac/20 raven swarm

[Russ Bullock]

IceSerpent, on 21 September 2014 - 01:16 PM, said:

No, it's not difficult. The way it should work is to only match large team vs. large team within a given threshold. For example (assuming threshold of 2):

Groups in the queue: 12-man, 10-man, 2 x 6-man, 2 x 4-man, 2 x 2-man
Matching:
- take the largest group, check if we have a match (second largest one) within threshold - we do (12 and 10)
- put largest group into team A and matching one into team B - we have 12 v 10 at the moment
- fill second group with the first 2-man and we're done (12 vs. 10 + 2).

What if we don't have a matching second largest group? I.e. no 10-man in the previous example.

Groups in the queue: 12-man, 2 x 6-man, 1 x 4-man, 4 x 2-man
Matching:
- take the largest group, check if we have a match (second largest one) within threshold - we don't (12 - 6 > 2)
- skip the the largest group (leave it in the queue), start with second largest (I'll call it "largest" from now on) - we have a match (6 and 6)
- put largest group into team A and matching one into team B - we have 6 v 6 at the moment
- fill team A with a next largest group - we have 4-man, now we're at 10 (6+4) v 6
- fill team B with a next largest group , we have 2-man - now we're at 10 (6+4) v 8 (6 + 2)
- fill team A with a next largest group - we have another 2-man, now we're at 12 (6+4+2) v 8 (6+2) and team A is good to go
- fill team B with a next largest group - we have another 2-man, now we're at 12 (6+4+2) v 10 (6+2+2)
- fill team B with a next largest group - we have another 2-man, now we're at 12 (6+4+2) v 12 (6+2+2+2), we're done.

To summarize the algorithm:

The "core" (largest) teams on each side are always within the given threshold.
The remaining slots on both teams are filled with the largest group that fits there, alternating between teams - add group to team A, then add group to Team B, the n add group to team A, etc.
Smaller teams get the fastest MM, as they can be added to larger groups and can be the "core" themselves (if the largest team in the queue is 4-man, it becomes a "core").
Larger teams may have to wait for another sufficiently large team to get into the queue.
Scenario where 12-man is matched against 6 x 2-man can never happen, the worst discrepancy can be the "filler" part ending up with 6-man on one side and 3 x 2-man on the other when "core" is 6-man vs. 6-man.

The difficult part is to convince PGI to implement someting like this.

It's a good set of thoughts but most players have over simplified the process in their heads to a large degree. Unfortunately the Match Maker is keeping track of way more things then is listed in your post.

You are not taking into account ELO - even if it wants to match certain people together as you suggest it is not allowed to because of an ELO range that is far to separated to be fair.

Next is it does not take the 3,3,3,3 aspect into account, although the MM does allow some flexibility in this and ultimately has to at least match weight class for weight class in a match.

Your example doesn't even take weight class into account. What if that 10 man just simply can't fit with the 2 man you have because the 2 man is two Heavy mech's and the 10 man group already has all the Heavies allowed?

Next is game mode selector - say the 2 man says no to the game modes that the 10 man has selected.

This is a complex problem - I don't expect our customers to understand all of the complexities but please don't make posts explaining how easy it is unless you understand all of the layers of complexity at hand. I am willing to bet many of my engineers are about as intelligent of people as you will find on this earth.

We can implement the suggestion if you do not care if you end up with 12 heavies against 12 mediums or any other extreme combination. Or if you don't mind if one team is Comp team good and the other full of noobs and so on.

Trust me if the problem were as simple as you describe it would be working as described.

[Russ Bullock]

FatYak, on 21 September 2014 - 05:14 PM, said:

with MWO's playerbase you can either have a great MM and long long wait times for a match, or, a not so good matchmaker and short wait times

We cant have both

This is a valid point - although player counts have been consistent pretty much since the beginning. MechWarrior has a LOT of factors to account for.

Players want exact tonnage matching plus a variety of mechs on the battlefield and to play against players of similar skill. It is a lot to take into account and unfortunately for a niche product to have some of the more strict requirements of any game. It isn't a great combination.

[SaltBeef]

I figured there were Many, many variables, and if the Player Queue is slow building, less players, (Slow Night) I gets more difficult. Nice again to see the Owners and Devs responding.

Edited by SaltBeef, 21 September 2014 - 10:15 PM.

[Mystere]

Russ Bullock, on 21 September 2014 - 10:07 PM, said:

This is a valid point - although player counts have been consistent pretty much since the beginning. MechWarrior has a LOT of factors to account for.

Players want exact tonnage matching plus a variety of mechs on the battlefield and to play against players of similar skill. It is a lot to take into account and unfortunately for a niche product to have some of the more strict requirements of any game. It isn't a great combination.

Since we're on the topic of the matchmaker, I really need to ask: Will merc base defense in Community Warfare be about "bringing my best team to defend my merc home base", or will it be subject to the same Elo, 3/3/3/3, and other rules that govern the solo and group queues?

Edited by Mystere, 21 September 2014 - 10:19 PM.

[BLOOD WOLF]

Russ Bullock, on 21 September 2014 - 10:07 PM, said:

This is a valid point - although player counts have been consistent pretty much since the beginning. MechWarrior has a LOT of factors to account for.

Players want exact tonnage matching plus a variety of mechs on the battlefield and to play against players of similar skill. It is a lot to take into account and unfortunately for a niche product to have some of the more strict requirements of any game. It isn't a great combination.

I have been saying that since The launch module. Player really simplify the MM and make these huge ideal MM's and neglect certain aspects. really grinds my gears.

The 3's fail threads were always a good laugh though. as well as MM fail threads and the like.

Dont get me wrong some players have very good ideas.

Edited by BLOOD WOLF, 21 September 2014 - 10:21 PM.

[Russ Bullock]

Mystere, on 21 September 2014 - 10:17 PM, said:

Since we're on the topic of the matchmaker, I really need to ask: Will merc base defense in Community Warfare be about "bringing my best team to defend my merc home base", or will it be subject to the same Elo, 3/3/3/3, and other rules that govern the solo and group queues?

There will be more posts on this soon but ELO doesnt really have a home in CW nor a way we can really bring it over. Also 3/3/3/3 really doesn't exist in CW any longer either since it will be balanced out by the dropship which in a way recreates 3/3/3/3 but the is more flexibility on how you drop in etc

BLOOD WOLF, on 21 September 2014 - 10:19 PM, said:

I have been saying that since The launch module. Player really simplify the MM and make these huge ideal MM's and neglect certain aspects. really grinds my gears.

The 3's fail threads were always a good laugh though. as well as MM fail threads and the like.

Dont get me wrong some players have very good ideas.

Honestly we can make the MM amazing if players will accept the trade offs. For instance allow us to put 1 mech per weight class into even smaller groups so for instance a group of two can't be two heavies etc. This alone would mean the MM could fit the jig saw pieces together really easy. Put that on top of the game mode selector being a vote and were in business.

But it comes with trade offs most wouldn't accept....but we can keep talking this through. I just ask that we spread the word to the player base on exactly what is going on with the MM and that it isn't that we can't program a simple method of putting groups together.

[TygerLily]

My dream matchmaker:

Ditch game mode selection. 33% chance of each game mode...done.

Super loose 3/3/3/3 valves.

Super tight Elo and Weight MATCHING (ton for ton).

I'd love to drop in to a game knowing our tonnage is equal and out skill is equal. Just need to get Clan tech balanced and IS vs Clan buckets instituted and I'd be a happy cat.