200 replies to this topic

[DAEDALOS513]

Anomalocaris, on 24 April 2020 - 12:17 PM, said:

It's quite simple. You make sure that each team has as equal a level of _overall_ skill as possible. For example, let's say that you and I are in the matchmaking pool of 24 players and we happen to be the top ranked guys by average match score. The MM would look at that and place us on opposite teams. It would then look at the next 2 highest match score players and put them on opposite teams and so on. It doesn't matter if the bottom 2 players in the group of 24 are 150 MS score players. They'd be split as well. There is also the requirement to balance mech classes and tonnage, so you have to switch some guys around, but you end up with a pretty good balance of skill.

Balanced matches don't mean that only people like you and I end up in a match. It just means that the skill distribution is as equal as possible on each team. Yes, there will be low skill players on both sides, but that's the price of low population. But at least both sides get equal amounts. And if the top tier players neutralize each other, it can be the play of those low skill pilots that swing the outcome of the battle.

As the matchmaker works right now, you and I and another top 3% player could all end up on one side of the match in meta assaults while the other side gets 3 250 MS players in LRM Stalkers. And that's perfectly fine according the parameters the matchmaker uses.

You are forgetting about mechs.. the MM also has to deal with tonnage.. One side shouldn't have 5 assaults while the other has 2.. it's more complicated than you think and that is why it takes more than 24 players to make a good match..

PS. what if a skilled player takes a 'fun' or troll build out for a spin.. (like I do on occasion) and it does poorly.. is that MM's fault? A great MM will take your experience in each individual mech into consideration as well..

Edited by DAEDALOS513, 24 April 2020 - 05:58 PM.

[General Solo]

DAEDALOS513, on 24 April 2020 - 09:54 AM, said:

How on earth is the matchmaker going to work when there isn't a sufficient player base? Back in MWO's prime, matchmaker had a bigger pool to work with and did a decent job.. now matchmaker scrounges to find enough players to drop a single faction match..

It did a decent job because it takes time for upward bias to take effect.
Back then in the first year it was great.
After about a year you could already guess what was happening.aka I remember Tier1/TheSecretService being tougher. (RED quote)

[Anomalocaris]

DAEDALOS513, on 24 April 2020 - 03:56 PM, said:

You are forgetting about mechs.. the MM also has to deal with tonnage.. One side shouldn't have 5 assaults while the other has 2.. it's more complicated than you think and that is why it takes more than 24 players to make a good match..

PS. what if a skilled player takes a 'fun' or troll build out for a spin.. (like I do on occasion) and it does poorly.. is that MM's fault? A great MM will take your experience in each individual mech into consideration as well..

Yes to all that. But if you look back over the last year or two there have been a number of threads looking at end of match screens, pulling player data off of Jarl's and then demonstrating how just switching a few players between teams would have made for a much more balanced match than what PGI's matchmaker created. Because PGI considers nothing but Tier and tonnage.

It's ok to have higher aspirations for a matchmaker, but it's been shown that you can balance 24 players much, much better than what PGI does. It would be great to consider mech and build choices too, even player performance by mech class (you don't want me in an assault for example). But let's start with the easy stuff. Balance by average match score and matches will get better.

[Horseman]

DAEDALOS513, on 24 April 2020 - 03:56 PM, said:

You are forgetting about mechs.. the MM also has to deal with tonnage.. One side shouldn't have 5 assaults while the other has 2.. it's more complicated than you think and that is why it takes more than 24 players to make a good match..

You shouldn't balance by number of chassis within a category, as the value of these mechs depends entirely on the pilot. There are some terribads who have been turtling around in shittily built assaults for years and are not competent enough to make their tonnage, threat level and armor matter - they die repeatedly barely crossing 200 damage... if THAT much even. Hell, I've seen QP matches where a medium mech did more work than half the assault lance.

What can be done is use the player's recent (~3 mo) track history with that chassis and in absence of that - some kind of estimate based on an their overall recent performance. That way a cowardly LRM80 Fafnir won't be treated as interchangeable with a dakka Mk II B piloted by a competent pilot.
For extra big brain points, you could even use the global performance of that specific chassis on that specific map to correct for possible favorable/unfavorable environment.
And this is not rocket surgery.

Edited by Horseman, 25 April 2020 - 01:12 AM.

[Alreech]

DAEDALOS513, on 24 April 2020 - 03:56 PM, said:

You are forgetting about mechs.. the MM also has to deal with tonnage.. One side shouldn't have 5 assaults while the other has 2.. it's more complicated than you think and that is why it takes more than 24 players to make a good match..

Tonnage is only a problem if you play without respawns.
Instead of one Mech give every player 3 Mechs with a total tonnage, so he can choose how to divede it between Mechs.
Maybe the player could set multiples of those Mech selections, and choose before the match starts between the so he could use the fitting Mechs for the map.
Maybe PGI should try to build such a system for MWO, and called it Dropdecks, Dropship mode, or something...

24 Player is absolutly enough for a good match if the matchmaker only has to consider player skill.
Works fine in Quake, Unreal Tournament, CoD, Battlefield, Fortnite, all those great Games for Solo Players who don't need such things like factions, units, team tactics like spotter / artillery, different tech levels...

But the matchmaker for a Mechwarrior Game should also:
1st Put players of the same faction in one Team
2nd Should put players of the same unit in the same lance
3rd put players of the same tech level in the same match
4rd put players of the same skill in a match
5th consider tonnage / skill

Even with high player numbers this can't work, unless mutch or the work is done before the matchmaking:
Group players by Faction / Unit, adjusting tech level, skill & tonnage: this can all be done before matchmaking even starts.
But that can only be done if the players are forced to drop as group of a fixed size like 4.
Instead of 24 entites the matchmaker has now only to deal with 6 (3x4 vs 3x4). It would also be possible to set up smaller matches if needed (2x4 vs 2x4, 4 vs 4).

1st Faction: only players of the same faction in a group allowed.
2nd Unit: max group size = lance size
3/4/5: Tech Level, Skill Level, Tonnage: max Tonnage allowed is calculated by Skill & Techlevel. So a top tier player who want's to use a Clan Mech get less tonnage than a new Player in a IS (or Trial) Mech.

To be honest: if factions & team tactics don't matter, why playing a Mechwarrior Game?
For Big Stompy Robots?
For Competetive Play?
A Mechwarrior Game is a terrible plattform for competetive play (even Battlefield is it) because it is to complex.
There is a reason why Soccer / Football is the most popular competetive game, and not the very complex and slow Cricket...

[General Solo]

Since groups are formed on the fly and are not permanent with many combinations of players, faction members, unit members and mechs, player stats should form a skill base line (Lets say measured in avg Match score) for a group.

This baseline can be used to rank that group in the current group queue pool initially before they play their first game for that play session.

After which the groups performance (measured in win/Loss) in the current group queue pool is used to determine their rank within the current pool afterwards, for the remainder of that groups existence.

The next day when a new group is formed, identical or not, the process repeats.

I recommend for outliers were one team has a large skill gap delta over all other groups in the current group pool, use a round robin system, where all the groups play that skill gap outlier group once and then it repeats, to spread the pain.

At least till a competitive group joins the queue to match with said skill gap outlier group.

My thoughts

Edited by OZHomerOZ, 25 April 2020 - 04:23 AM.

[Horseman]

OZHomerOZ, on 25 April 2020 - 04:21 AM, said:

Since groups are formed on the fly and are not permanent with many combinations of players, faction members, unit members and mechs, player stats should form a skill base line (Lets say measured in avg Match score) for a group.

I believe the net WLR of the teams is what should be used to balance them TBH. Match score can be farmed by vomiting a lot of damage or running AMS boats, but without WLR it doesn't say whether that damage contribution has been effective.

[DAEDALOS513]

Horseman, on 25 April 2020 - 01:02 AM, said:

You shouldn't balance by number of chassis within a category, as the value of these mechs depends entirely on the pilot. There are some terribads who have been turtling around in shittily built assaults for years and are not competent enough to make their tonnage, threat level and armor matter - they die repeatedly barely crossing 200 damage... if THAT much even. Hell, I've seen QP matches where a medium mech did more work than half the assault lance. What can be done is use the player's recent (~3 mo) track history with that chassis and in absence of that - some kind of estimate based on an their overall recent performance. That way a cowardly LRM80 Fafnir won't be treated as interchangeable with a dakka Mk II B piloted by a competent pilot. For extra big brain points, you could even use the global performance of that specific chassis on that specific map to correct for possible favorable/unfavorable environment. And this is not rocket surgery.

So it would be fair if there were 5 assaults on one side and one assault on the other in a conquest match? Obviously not.. it's not just about damage.. matching by tonnage is also about team mobility, not damage done..

Besides, forgetting conquest matches... it's very hard being an assault pilot ever since 'the nascar' was born... players in faster mechs that abandon their assaults like to blame game losses on those same assaults.. story as old as time..

Edited by DAEDALOS513, 25 April 2020 - 07:36 AM.

[General Solo]

Horseman, on 25 April 2020 - 07:30 AM, said:

I believe the net WLR of the teams is what should be used to balance them TBH. Match score can be farmed by vomiting a lot of damage or running AMS boats, but without WLR it doesn't say whether that damage contribution has been effective.

Sounds even better

[Horseman]

DAEDALOS513, on 25 April 2020 - 07:33 AM, said:

So it would be fair if there were 5 assaults on one side and one assault on the other in a conquest match? Obviously not.. it's not just about damage.. matching by tonnage is also about team mobility, not damage done..

Good point, but that actually points to another issue: there's a disparity between the parameters even within a tonnage bracket. Take Annihilator and a Mauler on one side, Scorch and a Blood Asp on the other. It's still two assaults vs two assaults, but one side has a clear mobility advantage.

The key thing from my perspective is that the MM shouldn't try to slavishly match by class just to ensure 100% parity in both teams equipment (although obviously it should try for some broad equivalence if possible) but rather to ensure both teams as a whole are rougly equivalent in their ability to win the match.
I guess what I wrote in the following post makes even more sense if expanded - take the player's track record with the chassis and apply correction based on the difference between the chassis' global average WLR versus that it holds on the given map+mode combination (so chassis poorly suited for a mode are weighted lower whereas ones that tend to contribute to wins are weighted higher).
More finicky, but not unmanageable.
While this won't directly account for build variance, I think the player's history with the chassis should be close enough.

[Sniper09121986]

Horseman, on 25 April 2020 - 12:22 PM, said:

Good point, but that actually points to another issue: there's a disparity between the parameters even within a tonnage bracket. Take Annihilator and a Mauler on one side, Scorch and a Blood Asp on the other. It's still two assaults vs two assaults, but one side has a clear mobility advantage.

The key thing from my perspective is that the MM shouldn't try to slavishly match by class just to ensure 100% parity in both teams equipment (although obviously it should try for some broad equivalence if possible) but rather to ensure both teams as a whole are rougly equivalent in their ability to win the match.
I guess what I wrote in the following post makes even more sense if expanded - take the player's track record with the chassis and apply correction based on the difference between the chassis' global average WLR versus that it holds on the given map+mode combination (so chassis poorly suited for a mode are weighted lower whereas ones that tend to contribute to wins are weighted higher).
More finicky, but not unmanageable.
While this won't directly account for build variance, I think the player's history with the chassis should be close enough.

It is possible to dynamically calculate BV aka battle value for whatever mechs people assemble. Of course, given the nature of MWO, it would require custom multipliers to be assigned on a per-variant basis not unlike the battle rating in War Thunder, and the system will be gamed somehow, but at least there would be some way for the matchmaker to tell between a classic brawling Atlas and the one with lurms and stealth armour. Much more important is whether we can trust PGI with the implementation of such complex systems

[crazytimes]

You can do whatever maths you want to try and balance teams... the problem is at the current population size and the direction it's going, no amount of maths is going to make any real difference.

The other issue with trying to make each team 100% balanced, it means top notch players will get potato teams every game. How is that going to be fun for them?

[BALIander]

D U N E, on 22 April 2020 - 07:40 PM, said:

[...]
Many like A S H here aren't trying to stop it because they are going to be at the wrong end of the stick. They are saying it cause they are going to be the ones holding the stick and seal clubbing, and are trying to give hyper casual pugs a chance. Best case, MM can at least see the two highest rated teams should not be on the same side and the pugs are caught up between two really good teams. [...]

Oh nice, you also see the need and benefit of only allowing max. 4 players to play together for sure.
- no clubbing if your other 4 mates that also wait in cue come out in the opposing team, ;-)

And you are right because the 3 biggies that make people to never to come back are:
- long wait time for a match, (cannot waist my time looking at a circle spinning)
- beeing repededly clubbed for too long a time.
- cannot play casual with friends.


The lance sorting covers it all.

Afterthought, maybe it helps if sorting logic to my initial post is added here:

Spoiler

- With 60 lances available (or any other number) - you take the 6 highest ranked to match a game:</p><p>1. strongest lance to team blue | 2. 2nd and 3rd strongest to team red | 3. 4th and 5th lance to blue | 4. 6th "weakest" lance to red</p><p>- Repeat with the next 6 lances, proceed until no lance is left.</p><p>The fewer lances in pool - the wider the spread of skill - but that is something we will have to live with I guess..

[justcallme A S H]

DAEDALOS513, on 25 April 2020 - 07:33 AM, said:

So it would be fair if there were 5 assaults on one side and one assault on the other in a conquest match? Obviously not.. it's not just about damage.. matching by tonnage is also about team mobility, not damage done..

Besides, forgetting conquest matches... it's very hard being an assault pilot ever since 'the nascar' was born... players in faster mechs that abandon their assaults like to blame game losses on those same assaults.. story as old as time..

Well you have a map vote screen. You have a team, you have organisation. If you have mass assaults you also have lights.

So yes it's fair. It then comes down to coordination and team play.

Matching tonnage is never going to be possible if Groups are allowed to drop with Solos. Let along GroupQ itself. There has always been huge tonnage disparity between large/small groups in GroupQ and that will never change.

You are saying that there should be an additional condition to the MM that is impossible to max work, ever.

BALIander, on 25 April 2020 - 04:59 PM, said:

Oh nice, you also see the need and benefit of only allowing max. 4 players to play together for sure.
- no clubbing if your other 4 mates that also wait in cue come out in the opposing team, ;-)

Realistically that'd going to do nothing. A limit of 4 down from 8 or 12 makes no actual difference because the less mechs you have the more tonnage available.

[DAEDALOS513]

justcallme A S H, on 25 April 2020 - 06:35 PM, said:

Well you have a map vote screen. You have a team, you have organisation. If you have mass assaults you also have lights.

So yes it's fair. It then comes down to coordination and team play.

Matching tonnage is never going to be possible if Groups are allowed to drop with Solos. Let along GroupQ itself. There has always been huge tonnage disparity between large/small groups in GroupQ and that will never change.

You are saying that there should be an additional condition to the MM that is impossible to max work, ever.



Realistically that'd going to do nothing. A limit of 4 down from 8 or 12 makes no actual difference because the less mechs you have the more tonnage available.

All i'm saying is MM has a complicated job to do and you need more than just the 24 players in queue as was stated earlier by a few guys.. MM can not do it's job with the current population numbers..

Edited by DAEDALOS513, 25 April 2020 - 08:17 PM.

[justcallme A S H]

It can do a better job of balancing out players.

That is its major flaw. Because it just sees "Tier 1-3". Tiers mean nothing. Tier 1 should be reserved for high skill players. Then the MM could make sure there are 1 x T1 player per side if there are 2 in the Queue. Right now it often dumps 2-3 high skill players on the same team when it should never do that.

All because PSR is garbage in, garbage out.

So no you do not need more than 24 players at all. The 24 just need to be distributed better, which currently it cannot ever/possibly do.

[DAEDALOS513]

justcallme A S H, on 25 April 2020 - 07:08 PM, said:

It can do a better job of balancing out players.

That is its major flaw. Because it just sees "Tier 1-3". Tiers mean nothing. Tier 1 should be reserved for high skill players. Then the MM could make sure there are 1 x T1 player per side if there are 2 in the Queue. Right now it often dumps 2-3 high skill players on the same team when it should never do that.

All because PSR is garbage in, garbage out.

So no you do not need more than 24 players at all. The 24 just need to be distributed better, which currently it cannot ever/possibly do.

I feel like we're goin in circles..

[Horseman]

crazytimes, on 25 April 2020 - 02:18 PM, said:

You can do whatever maths you want to try and balance teams... the problem is at the current population size and the direction it's going, no amount of maths is going to make any real difference.

The other issue with trying to make each team 100% balanced, it means top notch players will get potato teams every game. How is that going to be fun for them?

Not... quite.
First, expecting to get perfectly balanced teams is unrealistic, but a proper matchmaker would produce teams with less disparity in their overall performance. Get it close enough and whatever advantages one team has versus the other would be small enough that the other side has a decent fighting chance. Stomps aren't really fun, except for overgrown playground bullies. Matches where it could have gone one way or the other until almost the last minute? That's FUN.

Same goes for who top notch players are grouped with - overall they should get just enough semi-competent teammates to work with rather than be randomly stuck with full teams of deaf-mute potatoes. Again, it's not about ideal team composition for one side (because then WHO THE HELL DO YOU THINK IS GOING TO BE PUT ON THE OTHER TEAM?) but about bringing them closer to an equilibrium.

[justcallme A S H]

DAEDALOS513, on 25 April 2020 - 08:18 PM, said:

I feel like we're goin in circles..

Not really. I am explaining the flaws in what you are saying and explaining the reasons why.

You are just ignoring them and sticking to your point. A point that is factually incorrect.

It is incorrect because it is absolutely possible to balance matches better, right here / right now, in MWO. It just requires a bit more 'smarts'.

Edited by justcallme A S H, 26 April 2020 - 01:24 AM.

[Bistrorider]

I agree that separate 8vs8 and 12vs12 groups is a bad idea. This way we have 12vs12 in QP, 8vs8 in groups. So imo two different things, like two game mods, (and something like new game mode at last!) and that is good.

Edited by Bistrorider, 26 April 2020 - 02:29 AM.