271 replies to this topic

[Sandpit]

MischiefSC, on 10 July 2014 - 06:37 PM, said:

Yes, but here's the thing - the causes of those issues are not the matchmaker.

Better new player experience is a lobby or outreach system, or a multiplayer training ground.

Stomps have little to do with the matchmaker. They are an inherent aspect of team games like this. You've played 12mans, you know this. The most competitive matches between the most competitive teams still often end up as 12-0, 12-1, 12-2.

Better communication tool? I dunno, better school education for western civilizations? I will never use a VOIP system for this game - most people won't. If someone wants that we already have TS and the like available. Some easier and more intuitive targeting tools would be great. I've discussed NARC improvements for, essentially, information sharing/information warfare. Nothing MM related though.

More mechs on the field? Would freaking LOVE a complete balance pass - though sweeping and comprehensive changes in mech, weapon and equipment balances are a completely different topic.

The matchmaker though? Quite pleased with it. It does extremely well now with the resources available to it.

regardless of what you think is the root cause of these issues the whole idea is to make a better MM as a whole. Even if you personally don't think it's the MM system itself what ideas do you ahve to help improve matches? That's the entire point.

I like you Mischief but jesus is it really that difficult to post a thread and ask people to just give feedback and kick around ideas on how to improve the game?

GRiPSViGiL, on 10 July 2014 - 05:41 PM, said:

ROLFStomps will never be eliminated. The nature of this game isn't one where games can be turned around much although rarely can happen. Once the first mech goes down it just becomes a domino effect. Not only do the numbers advantage extrapolate that but the psychological effect further knocks down more domino's.

We should not be focusing on the elimination of Stomps in my opinion. Even in the competitive scene you often see high level teams get stomped. Such is life in this kind of a game. One life means you have to think about what you are doing because one mistake can start the domino's falling in the wrong direction. I think the MM changes have been outstanding and the game is in a much better state currently.

I think I've stated exactly the same thing
several times
repeatedly
roflstomps will never be eliminated but they can be mitigated along with helping ensure new players don't get into a run of them which turns them off of the game which kills retention which eventually would kill the game

I've posted several ideas on how I think PGI could fix the stuff I've listed, now I'm asking everyone else to HELP by either pointing out flaws in my ideas so they can be refined or offer up alternative ideas.

MischiefSC, on 10 July 2014 - 06:37 PM, said:

Better communication tool? I dunno, better school education for western civilizations? I will never use a VOIP system for this game - most people won't. If someone wants that we already have TS and the like available. Some easier and more intuitive targeting tools would be great. I've discussed NARC improvements for, essentially, information sharing/information warfare. Nothing MM related though..

I never said voip specifically or only
I asked about better communication tools

Everyone seems to be missing the entire point

In order to be able to help balance out the complex MM system that MWO uses there's going to have be more than one factor used to help balance matches.
Come off the MM exclusive thoughts

Yes, the MM is a piece of the puzzle. There's got to be MORE to it if you want a more balanced system overall.

[BLOOD WOLF]

When talking about the MM. 3's was a solution to only one problem. There are still many variables that come into play. Also the MM is not perfect nor can it be;However it can be tuned.

They got a ways to go but I have been having a swell time with the MM. It can be frustrating when I go into a match with 2 warriors and 4 sibkos and we run into aces or some other team like that. We always put up a good fight and it is never a shut out game if we lose. It is in part due to teamwork and leadership.

as for MM adjustments, yea it can be improved but that does not mean we should scrap the current one seeing it does a well enough job. Like i said it is not a perfect MM that is why we always get matches better than others.

I have had way better games than I have ever had and I can thank the New MM for that. Whenever we get rolled I can now say that the enemy team was playing better than I was, but I can not say the match was ever uneven.

That is how it is like for most of my matches.

[Sandpit]

BLOOD WOLF, on 10 July 2014 - 08:35 PM, said:

When talking about the MM. 3's was a solution to only one problem. There are still many variables that come into play. Also the MM is not perfect nor can it be;However it can be tuned.

They got a ways to go but I have been having a swell time with the MM. It can be frustrating when I go into a match with 2 warriors and 4 sibkos and we run into aces or some other team like that. We always put up a good fight and it is never a shut out game if we lose. It is in part due to teamwork and leadership.

as for MM adjustments, yea it can be improved but that does not mean we should scrap the current one seeing it does a well enough job. Like i said it is not a perfect MM that is why we always get matches better than others.

I have had way better games than I have ever had and I can thank the New MM for that. Whenever we get rolled I can now say that the enemy team was playing better than I was, but I can not say the match was ever uneven.

That is how it is like for most of my matches.

which offers ideas on how to improve it how....?
you've already stated you like the new MM. That's great. Now ideas on how to improve the rest of the stuff it didn't fix maybe?

[BLOOD WOLF]

What can we improve? From what I gather 3's was a sucess and matches are really benefiting from that, seeing Clan mechs balanced well. I do not know what else the MM could do, besides improving functionality to where it can perfectly match things up. there is not much else

What I mean is like this, take a unskilled worker and he will do the job but at first he might not do it well but over time he will get better. The MM can be tweaked here and there to improve functionality over time but as far as that It is doing good.

So asking form improvements is gonna lead to subjective ideas of how the MM functions at this point. unless there is something really wrong with the MM as to where it is acting very wacky and starts putting 8 assaults constantly in a succession of 20 matches.

[AntiCitizenJuan]

I am completely satisfied with solo queue and 4x3.

Good work PGI, keep it up

[MischiefSC]

Sandpit, on 10 July 2014 - 08:24 PM, said:

regardless of what you think is the root cause of these issues the whole idea is to make a better MM as a whole. Even if you personally don't think it's the MM system itself what ideas do you ahve to help improve matches? That's the entire point.

I like you Mischief but jesus is it really that difficult to post a thread and ask people to just give feedback and kick around ideas on how to improve the game?


I think I've stated exactly the same thing
several times
repeatedly
roflstomps will never be eliminated but they can be mitigated along with helping ensure new players don't get into a run of them which turns them off of the game which kills retention which eventually would kill the game

I've posted several ideas on how I think PGI could fix the stuff I've listed, now I'm asking everyone else to HELP by either pointing out flaws in my ideas so they can be refined or offer up alternative ideas.


I never said voip specifically or only
I asked about better communication tools

Everyone seems to be missing the entire point

In order to be able to help balance out the complex MM system that MWO uses there's going to have be more than one factor used to help balance matches.
Come off the MM exclusive thoughts

Yes, the MM is a piece of the puzzle. There's got to be MORE to it if you want a more balanced system overall.

My fundamental issue is that focusing on the MM is essentially focusing on 'it's not my fault, it's my team/the other team'. I'm not griping about the thread but that if you look at the majority of the responses they tend to move towards 'the MM should just sorta KNOW and only match up fights where I've got good odds of winning no matter what mech I bring or how I play'. That's not going to happen.

I am strongly of the opinion that the fix for these things is totally unrelated to the matchmaker, but instead mech/chassis/weapon balancing. Move more weapons away from PPFLD (this strongly favors veterans over nubs, the Dire Wolf 50pt alpha design being an extreme but valid example) and more towards DoT, as this rewards *consistent* skill performance but is at the same time more forgiving of simple mistakes and brief periods in poor positioning.

As an example I've taken to loading my banshee with a big XL, Gauss and 3xPPCs. It's a 45pt alpha but way more maneuverable than the Dire Wolf.

It strongly, *strongly* favors luck. Some poor ******* makes one mistake, runs up the wrong alley where I'm aiming and I just killed him at worst or pulled half the armor off one location at best. You know what? I miss a lot - slow cooldown.

I only have to get lucky once per enemy though. Which pays off on average; pays off significantly.

It is however a stupidly expensive build and works best with the right modules and to play it heat efficient and as maneuverable as a Banshee needs to be it needs double basics unlocked -

Which is to say it was an investment of, in the end, about 40k GXP, 100k XP and about 60 million cbills. Vastly out of scope for new players.

THAT is the biggest issue. Newer players dropping in matches with more experienced players isn't that terrible. It's 12 v 12, there really are no better ways for new players to learn without a huge commensurate investment. A newbie queue for example just makes for a great troll fishing pond for experienced players running new accounts and the like. There is no 'safe' way in a F2P to shepherd new players away from vets completely.

Better weapon balancing more towards DoT and away from PPFLD is probably the single biggest benefit the game could have for all these factors. Take mechs gimped by poor hardpoint design and hitboxes. Fixing those is a huge investment with no payoff. Chassis perks like what clan omnipods have could make a big difference here.

Those are the biggest things, no question.

[BLOOD WOLF]

well said.

[MischiefSC]

Working on inputting all the data from all the screencaps I took. It'll be a little bit - for reference though, here's the stats from the first ten -

All matches in Skirmish to keep consistency. This also means it's more likely to skew a little heavier, but going to do 20 drops in Conquest for comparison.

Total mech populations -

45/55/73/67 -

Average of team breakdown of per team 2.5/2.75/3.65/3.35.

All very close. Lights are a bit under-represented but not dramatically so.

surviving mechs on winning team:
12/5/7/7/10/2/3/8/2/11
Average - 6.7 mechs on winning team. Only 1 in 6 matches would be considered a 'stomp'.

Match lengths -
6:20/11:16/9:29/6:49/8:22/8:12/7:17/7:10/10:55/7:55

An average of ~8:35/match.

Additional data of interest -

7 out of 10 matches a Clan mech held the high score.

65.8% of all mechs were IS mechs.

This means that while Clan mechs were only 34.2% of the mech population they were the top-performing mech 70% of the time. Literally representing themselves at 2x their population weight.

I've got a ton more matches to spec out. Going to collect damage and score plus kills and assists by mech. May as well gather it all in, right?

Admittedly from what I've seen though, the 3/3/3/3 is working pretty well. Match length is around 28% higher than it used to be pre-Clans - that's good. Number of lights/mediums total - also up by almost 25%. Scores are, strangely, higher *across the board*.

Conversely, as I'm digging into the data....

Clan mechs seem to have a distinct, discernible and measurable advantage. Higher score, more kills for their weight class and total representation in the drop over-all, not just by team.

Having my first full weekend. I may get it done and up to post, I may not. This was my third week of 80-90 hours and I'm pretty f**king done with statistical analysis for a bit I'll get it out soon enough though.

[BLOOD WOLF]

I am loving the data, It was a nice read and I am a bit surprised by some of the findings.

[Sandpit]

I understand what you're saying. I like the data. It's still not offering anything to improve it. So what can we do to improve the game quality for everyone.?

PappySmurf, on 10 July 2014 - 06:56 PM, said:

I think right now the pug/solo/casual MM FFA 12v12 matches are the best they have ever been so I say leave it alone.If anything just add more game mode options and maps.

3/3/3/3 should actually be just a game mode plus they should just add a stock trial mode also for fun and a IS VS Clan game mode as well.It would spice up the game quite a bit from the only 3 game modes we have now.

Game mode list
Assault
Capture
Skirmish
3/3/3/3
Stock/trial
Clan Vs IS

The group queue in the recent announcement looks like it is being worked on so 1v1-12v12 match ups are going to be a viable option. Which should be better for balancing out groups.


If Cw is going to be like say the old NBT or MWL leagues werre I think it will work out just fine for MWO.

http://www.netbattletech.com/nbt-hc/

I actually like the idea of having it as a game mode.

I don't care what pgi says or does, no amount of restrictions are going to make people want to take a better mech variety.

MischiefSC, on 10 July 2014 - 09:48 PM, said:

My fundamental issue is that focusing on the MM is essentially focusing on 'it's not my fault, it's my team/the other team'. I'm not griping about the thread but that if you look at the majority of the responses they tend to move towards 'the MM should just sorta KNOW and only match up fights where I've got good odds of winning no matter what mech I bring or how I play'. That's not going to happen.

I am strongly of the opinion that the fix for these things is totally unrelated to the matchmaker, but instead mech/chassis/weapon balancing. Move more weapons away from PPFLD (this strongly favors veterans over nubs, the Dire Wolf 50pt alpha design being an extreme but valid example) and more towards DoT, as this rewards *consistent* skill performance but is at the same time more forgiving of simple mistakes and brief periods in poor positioning.

As an example I've taken to loading my banshee with a big XL, Gauss and 3xPPCs. It's a 45pt alpha but way more maneuverable than the Dire Wolf.

It strongly, *strongly* favors luck. Some poor ******* makes one mistake, runs up the wrong alley where I'm aiming and I just killed him at worst or pulled half the armor off one location at best. You know what? I miss a lot - slow cooldown.

I only have to get lucky once per enemy though. Which pays off on average; pays off significantly.

It is however a stupidly expensive build and works best with the right modules and to play it heat efficient and as maneuverable as a Banshee needs to be it needs double basics unlocked -

Which is to say it was an investment of, in the end, about 40k GXP, 100k XP and about 60 million cbills. Vastly out of scope for new players.

THAT is the biggest issue. Newer players dropping in matches with more experienced players isn't that terrible. It's 12 v 12, there really are no better ways for new players to learn without a huge commensurate investment. A newbie queue for example just makes for a great troll fishing pond for experienced players running new accounts and the like. There is no 'safe' way in a F2P to shepherd new players away from vets completely.

Better weapon balancing more towards DoT and away from PPFLD is probably the single biggest benefit the game could have for all these factors. Take mechs gimped by poor hardpoint design and hitboxes. Fixing those is a huge investment with no payoff. Chassis perks like what clan omnipods have could make a big difference here.

Those are the biggest things, no question.

Yes it's a 12v12. 6 new players on a team against 12 vets is not a balanced match. Or 8 or 9
New players need a queue for themselves that prevents things like that
At the very least limit the number of new players per team.

[MischiefSC]

Sandpit, on 11 July 2014 - 06:05 AM, said:

I understand what you're saying. I like the data. It's still not offering anything to improve it. So what can we do to improve the game quality for everyone.?


I actually like the idea of having it as a game mode.

I don't care what pgi says or does, no amount of restrictions are going to make people want to take a better mech variety.


Yes it's a 12v12. 6 new players on a team against 12 vets is not a balanced match. Or 8 or 9
New players need a queue for themselves that prevents things like that
At the very least limit the number of new players per team.

Population is a complex topic when talking about something like a matchmaker. It's not about how many total players - it's about how many total players available RIGHT NOW.

Think of it in the same context as CLG, from that thread where everyone was taking turns slapping Mudhutwarrior around like that hysterical woman on Airplane.

New players die very quickly in matches. More so than vets. They're also less invested in sticking around - they're the ones who die and quickly drop for another match.

This means that they are statistically more likely to be dropping in another match than a Vet. No friends list stuff to chat about - they don't have mechs to sort through and choose from between matches, bunches of loadouts to play with. Take most vets and most newbies on average (always exceptions of course) and the newbie will be playing more total matches in the same time-frame.

This gives him an out-proportioned presence in the matchmaker. So if you have 50 newbies and 50 vets available in a given hour, in any 30 second interval there might be 20 newbies and 7 vets available to drop into a match.

Next, filling matches themselves is like dropping nuts from a conveyor into bags. You can't let the nuts pile up at the end, you need to put them in the right bag but as fast as possible. You don't get to take them back out after you put them in so you can then put them in another bag. While ideally each bag should have 3 of each kind of nut and they should be of a similar size, sometimes you just don't have that at hand. You can't just stop filling bags or taking nuts off the conveyor. So you stick what you got where you can.

6 newbies plus 6 vets vs 12 vets is going to be pretty rare though. They're normally going to be spread pretty evenly. Also, from a statistical perspective, most players are not a lot better in any given match than a newbie - they may handle better and work the map better but the performance percentage increase is embarrassingly small. Even good players have bad matches or may run a trial mech. I do sometimes. I'm not that good in them. So in those matches I'm not carrying anything close to my Elo.

Ideally your Elo should take into account the specific chassis and build you're taking. Even better would be adjusting the relative Elo values in the team-building tally by your historical synergistic successes - simplest example is, if I'm in my Jenner (S) with NARC and UAV, the more LRM boats on my team the higher my relative Elo is going to be. The reality is even more complex; when I'm in my D-DC I get exponentially more dangerous with people who are good brawling flankers who do well playing close to me. I'll soak fire for them, I'm very conscious of teammates taking LRM fire and use my ECM to cover them when they drift far enough away to be vulnerable, I'll fake a charge at people who are trying to run down a damaged teammate. I'm like a mama elephant protecting my herd and all modesty aside I'm good at it. If I have a team willing to exploit that to win I do really well. My KDR is only moderate but I pug close to a 2.0 win/loss in it. How far do you drill down on player behaviors to accurately reflect that?

You can extract the data to drill down as granular as you want is the answer. You get increasingly complex and swiftly run into cost/reward limitations but you truly could fine-tune the Elo aspect of the matchmaker tight enough to rank every player in the game sequentially by mech, loadout, even success with a given weapon. It wouldn't really be an 'Elo' score at that point but much like Gliko or Trueskill more of a system unto its own but MW:O as an environment would cater to that really well.

Currently though it would make only a minute difference. With the group/pug queue split you've pulled a huge swath of weight off the top of the Elo bucket at any given time. I wouldn't be surprised to find out the pug queue population is over 90% people with high-average to sub-average Elo scores - almost all under 1800.

So.... time is the fix for that, honestly. While having new players on your team isn't ideal it's not enough to drive people away from the game. It is likely to make more of those people stay. We need another 18 months of population growth and I think the current MM will seriously help with that. Customer retention isn't so much about reality as perception. Premade teams can *feel* unfair to a newbie pug. Elite players and mechs... those too, but you'll see them on your own team too. You'll spectate them and be motivated by them. Even getting beaten by one can be motivating. You can get everything they have. Teams though feel more like 'picking on' than 'out-performing'. There's actually a lot of brain chemistry involved. If it's 3 or more people together and you're alone and they're hostile, they trigger your 'Other' response. They are Other, enemy. Hostile, dangerous, threatening. That's why premades dropping in the pug queue, kicking behind and acting like jerks was always so toxic to the game - it's kicking off all these chemical responses of hostility, the last thing most those people wanted was to join them or even stay in the game.

So long winded, I know. Sorry if I came across as short before; it's been a long two months and I've been tired and fussy. It's a good subject and important and worth discussing. For the time being I think the matchmaker is as good as it can get without moving into large investments with minimal financial return for PGI, which means 'unlikely'. From here and for the next year or so I'd say weapon rebalancing and mech rebalancing (there are so many crap chassis - newbies get into them, invest their precious cash and find out.... whoops! Crap! Now grind that pile of crap for a couple months and try again! That's a *horrible* experience) will get a much bigger bang for the buck.

[Joseph Mallan]

Sandpit, on 11 July 2014 - 06:05 AM, said:

I understand what you're saying. I like the data. It's still not offering anything to improve it. So what can we do to improve the game quality for everyone.?

I have to stop right here. You cannot improve something for Everyone.
A Trades teacher I once had broke it down like this

Some=All-Most
Most=All-some
All=Most+some

There will always be Some that will not like an improvement for whatever reason, so shoot to get it right for Most, and hope the Some is a very small number.