81 replies to this topic

[PhoenixFire55]

Nesutizale, on 25 April 2019 - 04:16 AM, said:

From what I gethered by what other game designers say its when in a team everyone is equal then each person has less impact on the game overall.

I'm not sure where you've heard that but it sounds like lame excuse tbh. If match is balanced it means everyone has a decisive impact on the outcome. Balance means really closely matched teams, which means that smallest decisions/mistakes influence the outcome, which in turn means that it matters what each and every player does. Now obviously it is far easier to balance a match when you have 24 identically skilled players, so even if your MM does frankly nothing, you'll still get a balanced match. Hence my "lame excuse" comment, coz grouping players of similar tiers simply hides the fact that MM fails to do its job. 12v12 hides it in a similar manner compared to 8v8, simply due to LoLN, even if you draw players randomly, you are more likely to make a balanced match the more people you have in each team, etc.

Nesutizale, on 25 April 2019 - 04:16 AM, said:

As you have the same skill as everyone around you, you don't matter that much anymore. If you fall, there is someone as good as you are to replace you. Thats what the statistics say.

Statistics say nothing of the like. Again, in a really balanced match every small thing every player does matters. Matches between bunches of equally skilled players are less likely to snowball, because for each fkup a player makes there is either a same probability fkup of the same magnitude on the other team, or equally probably epic heroics of the same magnitude on yours. On a team composed of players with greatly varying skill, a fkup of an elite level player who is expected to carry hard isn't likely to be compensated by a simultaneous heroics of the 11 potatoes on his team, and just as unlikely to be compensated by the similar fkup of an elite player from the other team, simply due to such events being rare.

But again, the fact that matches are more/less likely to snowball doesn't matter they aren't initially balanced. And in fact, actual team matches are supposed to snowball hard quite often, even when teams are perfectly balanced. And more so, the higher is the overall skill of players involved.

Nesutizale, on 25 April 2019 - 04:16 AM, said:

Yes that is what I said. It takes players skills and groups them together and to do that each players skill has to be looked at. With the limit of 24 players that you had given it would mean that the MM can not make a good match as the numbers of different Tiers would most likely never match to an equal state.
That is why you need a bigger group then 24 people that are searching for a match to have the equal teams you want.

Even if we assume we absolutely need to for whatever reason divide people into buckits (coz more buckits is good, amiright?), then the total population you need for balanced matches with five tiers is still ... 120 players. I somehow think, that MWO has way more active players than 120, and needless to say, it doesn't even need that many since T1's play with T2's and T3's. So, this number ... 120 people, is the highest estimate of the population you need for balanced matches. We have that. Way more than that in fact. So please stop with the nonsense about MM needing more population. Its just another lame excuse to cover the fact that a developer simply can't make a proper MM even tho they have all the tools and all the stats necessary to figure out a proper skill-related criteria for a MM to rank people with.

And for the record ... MM isn't supposed to be making your matches "fun". It is supposed to be making them balanced. If I'm asked to choose between "fun" 0-12 matches with different people vs "unfun" 11-12 matches with exact same people every damn time, I will without a doubt choose the latter.

[Eatit]

justcallme A S H, on 23 April 2019 - 07:35 PM, said:

FP will have nothing to do with QP PSR, thankfully.

Paul confirmed that earlier in the week.

PSR is not necceasrily bad. However because of WHAT goes into it within MWO - PSR is bad. PSR would be fine if wins didn't count for so much along with other things to affect rises and doesn't give you enough to go down/drop.

The only thing PSR/ELO/Skill will have to do with the new FW is the order in which solo's get added to matches. PSR/ELO/Skill will not be used in any manner to determine the composition of teams.

It's all still random. Groups will not have their ELO/PSR/Skill looked at to determine team composition. There is nothing else new here in regards to skill based match making.

Solo's getting added to teams will be placed by highest ELO/PSR/Skill first. That's it, nothing else related to skill will be used for match making.

[Nesutizale]

Quote

Hence my "lame excuse" comment, coz grouping players of similar tiers simply hides the fact that MM fails to do its job.

Then what is grouping of players by tiers if its not a match making?

Quote

even if you draw players randomly, you are more likely to make a balanced match the more people you have in each team, etc.

Yes statisticly it balances itself out at what team sizes?

Quote

Matches between bunches of equally skilled players are less likely to snowball, because for each fkup a player makes there is either a same probability fkup of the same magnitude on the other team, or equally probably epic heroics of the same magnitude on yours

For every mistake I make someone else also does one and for every good action there is a good action on the other team, that is what you discribe.

So did it matter what I did? Everyone else could have done that. There is not one player where you could say "thats the guy that messed it up" or "that one did pretty good" as they would be equaly. Would it matter of I get better? I will be nothing special if I stay at my level and I will be nothing special if I go up or down the ladder. Is that what you want?

Do matches that allways run equaly make memorable matches or do you remember the matches that where good or bad?
Most people remember the good and the bad, not the normal things that happen. What games are you talking about or thinking about? Is it the game where you went through highs and lows or the okay ones? (I mean that in general not only MWO)

Quote

even if we assume we absolutely need to for whatever reason divide people into buckits (coz more buckits is good, amiright?), then the total population you need for balanced matches with five tiers is still ... 120 players

You mix something up. Buckets is for different gamemodes like QP, FP, Solaris.
What I discribed was what is neccessary to get equal teams as you want them. We have 5 Tiers. When you want a perfect equal team you need 24 people to form a T5 game. 24 people for a T4 game....and so on.
That totals into 120 people.

Realisticly you will most likely never have these exact number. That means you need more people of lower tiers then higher tiers but when more higher tier players are present you get unbalanced games, hance you would have to wait until more people show up to balance that out, in the worst case you need to wait for another 24 people group to form to have a balanced game.

Also looking at the Steam number for a rough estimation at peaks we have ~1300 people with an average of ~700 a day. Take the average and divide them by timezones. Since some overlapping will happen I did say we have at least 4 Zones at any given time, more likely more but lets go with 4. That would mean the matchmaker would have 175 people to choose from on average.
That is more then the 120 that I put up but its still a small number and when you read through the forums and I bet from your own experiance, you have seen a lot of these names again and again. So either MM works and you are put in the same group because of equal skill or its just that there aren't enough people and you will be thrown together anyway.

Edited by Nesutizale, 25 April 2019 - 10:57 AM.

[Eatit]

TLDR: Give specific details of your proposed MM. Give specific details of your player skill rating system. Explain why you think your system is better than the current system. Show examples of the award winning games you've created to back up your statements. Show actual player numbers to back up your statements. Soon there will only be enough players to support one bucket. Meaning all tiers play against each other.


Buckets are just a way of describing the groups of players that are required to make a match.

QP if divided into a tier system that only matches like tiers would be 5 buckets. One for each tier.

Right now there are only 3 buckets in QP. Tier 5-3 Tier 4-2 and Tier 3-1. It can also change that around for low pop times making it even fewer buckets. If we go to a strict tier only 5 bucket system the wait times will go up dramatically. The more buckets the longer the wait.

There is also another modifier that we haven't taken into account. The MM also looks at the weight class of the mech you pilot.

You need 6 of each weight class to make a match. Or you could open up the MM to more of any class based on percentage in que I guess. Although that makes the balance off.

Let's say there are 50 tier one people playing. That could be 2 simultaneous matches if everyone is driving the right size mechs.

What happens if 39 of them are in assaults, 3 are in lights, 5 are in mediums and 3 are in heavies.

The MM would be able to make one equal size match from that, 16 assault, 2 heavy, 4 medium, 2 light. Not that this is balanced for the medium mechs but it's the best the mm could do. If it had to wait for a balanced game it would need at least 6 of each weight class.

The second match would have to be created from 23 assault, 1 heavy, 1 medium and 1 light.

How will it make the 2nd match fair? It can't. The first match it made wasn't even fair.

Now the 26 remaining people in the que will have to wait for the first match to finish before another is created. That's a long wait. Much longer than the current 2 or so minute wait. The current MM is likely already creating matches with more than 3 tiers. I'm sure the valves open up wide to reduce wait times. Meaning tier 1 could be playing tier 5 if the population is low enough.

Consider that there may only be 23 tier 2 pilots on at a time and I think that is optimistic. Will those people just sit forever waiting for a match that will never happen? What is the alternative for them? The MM opens up and allows them to play with other tiers? Well, that's what we have right now.

I would love to see a single tier only MM, but the number of players just doesn't support that. The MM is good, the metric they use for player skill is bad. We all know that PSR/ELO/Tier is more like an XP bar. If you've played long enough you are in Tier 1. That doesn't mean you have the same skill as everyone in tier one. I'm sure there are people in Tier 1 that are in less than 80th percentile according to Jarl's.

If we broke up the MM by tier and we used Jarl's to create the tiers based on 20% for each tier. Would we have matches that are more likely fair?

Does Jarls look at XP in its decision for %? Nope I don't think so. It just looks at average match score for the season. What if a tier 5 player is good enough to get high match scores against other tier 5 players? Would his match score be that high against tier 1 players? No because the tier 1 players understand the game more.

Jarl's would have to be reset and it would take some time to sort the good from the bad. All matches would be unbalanced while it tried to create a tier system. what if you were on an all bad player team because nobody is ranked yet. The other team is all good players. How high is your match score going to be? It won't reflect your skill it will reflect that you got shafted by the MM. How long will this unbalanced match making go on before we say current percentile score will be used to determine tiers?

How fun will it be during that time?

I think if you want to propose an alternative system for the match maker you should lay it out in detail. That way people that understand it more than you do will be able to explain to you why it will or will not work.

If you just say "My way is better" without explaining why then it clearly isn't better. PGI has all the metrics for the player population and has created the MM to create matches as quickly as possible. That's because we as a player base have stated that shorter wait times are important to us. Some people only have a few hours to play and I'm sure they don't want to spend it staring at the spinning wheel of doom. I'll take a lopsided match over no match any day of the week.

We need to get a grip as we aren't always as smart or knowledgeable as we think we are.

[justcallme A S H]

Wil McCullough, on 25 April 2019 - 02:11 AM, said:

Imbalanced matches don't cause stomps. There's an argument to be made that they're correlated but the relationship isn't causal. Stomps are caused by snowball effect. And when both teams start with such limited and non-renewable resources, the snowball effect is ridiculous. If differences in skill level caused stomps, there wouldn't be many stomps in the world championships. And the stompers and stompees keep alternating too.

Stomps happen when one team, by accident or design, is able to punish early mistakes from the other team. That's why close matches are more rare. Either nobody made a mistake throughout the match or neither team is good enough to roll an advantage into a win. That's one reason why close matches are more common in the lower tiers where it's mostly tater on tater action.

Yes and no.

I've carried many - no exaggeration - MANY games from 0-3 to a 12-10. There are many factors involved but it comes down to largely skill.

Yes there is snowball but it can be stopped with good game-play. That is the issue however, good game-play is something only a small handful of this player-base has. For whatever reason the vast majority of players in this game just aren't any good. You're talking 100 or so top tier and maybe a couple 1,000 of decent. The rest are all just in the bucket of 35,000 of average to low skill. And not being rude, just the fact of the matter.

Agree on the point of lower tier match being more even because it contains less top shelf players. Top shelf capitalising is what turns the tide of any game. More so MWO because of the skill disparity in a T1-T3 game because of how many people in T1-T3 only play at a T4 level but have "grinded" the EXP bar to end up in higher places than they should ever be.

Being in a higher tier than you should be also gives you a false sense of skill. I've met many a "T1" player who is outright garbage. They have skills barely above what I would class as basic yet they have played 5,000 games. This is the problem with constantly rewarding people for playing poorly which MWO does.

[MrMadguy]

Yeah, whole PSR's idea is better, than ELO. ELO is good for 1vs1 and may be pre-made groups, i.e. where skill value is static. Random groups suffer from "superposition principle". I.e. you just can't distingush personal skill on a background of other 23 players' skills. 1 + 2 + 3 = 2 + 2 + 2. At the same time PSR is cheap version of TrueSkill, cuz many things are just thrown out. Prediction, uncertainty of skill, distribution coefficients, etc.

But biggest problem is - any MM should allow player to drop his rating as fast, as gain it. This two should be balanced, if developers don't want rating to be biased. But it isn't the case in MWO. Mostly because "desired target skill value" is soooooooo low, that "bad performance" is almost non-existent in this game. You either play good and your rating stays the same. Or you play very good and your rating rises.

I'm not sure, if such design is intended. But I suspect, that it is, because what PGI was trying to achieve with PSR system - to lower queue times for higher Tiers.

And about balance. As I've already said many times, balance wouldn't be needed, if this game would have proper matchmaking. All players would just play in their league. And balance isn't actually possible in this game. Mostly due to large amount of customization and different variables, that are needed to be balanced. Also due to lore reasons. Yeah, some 'Mechs are OP and should play in their own "divisions". For example my MKII can't do anything against Annihilators and Dire Wolfs in S7. And there is no D0, we could move this guys to. And for example my only D7 'Mech, some loyalty Cicada, can't do anything against SRM24 guys. And there is no D8 to move not OP Light/Medium 'Mechs to.

[Nesutizale]

@Eatit

I think Phoenix and I are even before the part of the specific workings of a MM but what it should achive.

As far as I understand him he wants perfect balanced teams that will result in close games because he says every player then is meaningfull to the team effort.

I think that perfect balanced teams aren't even possible or desirable as the games would be kinda dull cause the individual can't hardly make an impact.
Like in a footballgame with two equal teams. The ball goes back and furth but nothing realy happens.

[FRAGTAST1C]

Nesutizale, on 25 April 2019 - 11:30 PM, said:

Like in a footballgame with two equal teams. The ball goes back and furth but nothing realy happens.

This analogy is considering two equal teams with zero flair players. Like two teams battling relegation that are too afraid to lose.

What is better is to have is two equally strong teams with flair players. Then it's entertaining to be involved in.

[MrMadguy]

Nesutizale, on 25 April 2019 - 11:30 PM, said:

@Eatit

I think Phoenix and I are even before the part of the specific workings of a MM but what it should achive.

As far as I understand him he wants perfect balanced teams that will result in close games because he says every player then is meaningfull to the team effort.

I think that perfect balanced teams aren't even possible or desirable as the games would be kinda dull cause the individual can't hardly make an impact.
Like in a footballgame with two equal teams. The ball goes back and furth but nothing realy happens.

If we can't achieve fair 50/50 chances to win in every single match, then at least there should be fair 50/50 chance to get advantage/disadvantage. There is nothing bad, if in some match you lose with 0 chances to prevent it, but only if you don't do it in tens and hundreds of matches in a row.

[Nesutizale]

What is a "flair player" ?

Quote

If we can't achieve fair 50/50 chances to win in every single match, then at least there should be fair 50/50 chance to get advantage/disadvantage. There is nothing bad, if in some match you lose with 0 chances to prevent it, but only if you don't do it in tens and hundreds of matches in a row.

And that is the part where, in case of MWO, it becomes realy complex. If everyone would bring the same mech, everyone would have the same advantages/disadvanteges and only player skill on what to do with what is given would count...like Chess.
All figures have the same potential so its up to the player.

With MWO how do you archive that with all the mechs and configurations? Tonnage alone isn't enough. A good Prianha can shreet an assault in no time. So we would need a system like Battlevalue (BV) from the TT to make mechs even compareable.

Lets say we have a good BV System that gives us a way to compare mechs. Then we should be able to create teams where each side has roughly the same potential in what mechs are on the field to tip the balance in their favor. Still the differance in configurations can open up different playstyles for players to exploid.

Next we have to take the players skills into account. A player can make even a bad mech config work because he "trained" with it and perfected his playstyle.
So I did say that player skill would outweight the BV but what about beginners?
Maybe we would need a "dynamic" mixed calculation that the goes like "The higher the PSR the lower the BV matters".
Also the BV will never be ignored.

There is more to think about like how PSR would be calculated but I gota go now...

[Burning2nd]

justcallme A S H, on 24 April 2019 - 07:51 PM, said:

Yeah it's again the win condition abd the fact you can basically suicide and still go up that's the problem.

If you're match score is below 210 and you're in a win. It should be no change.

Below 180 - small drop.
Below 150 - medium drop.
Below 125 - large drop.

Or something like that. It would then stop all the players with 190 average match score making it into T1. 190 being below the overall player base average

the problem is this doesn't account for all of the other issues that PGI has (or has had) I took a year off when they introduced this crazy psr system.. when they decided to change the skill tree and change the game completely

they are not judging on skill that they should be, accuracy, doesn't count like it should.. Win / lost doesn't count as it should participation counts: BUT complete wrong.... things like formation and communication stuff that makes mechwarrior.... mechwarrior...

what we have is a glorified first person shooter...

[Burning2nd]

and being that there is no reward other then making money and getting MC... there is really no point to anything other then that

[FRAGTAST1C]

Nesutizale, on 26 April 2019 - 12:45 AM, said:

What is a "flair player" ?

A player who can do something that can alter the course of the game in favour of their team. So, if a couple of players who are really good based on the proper ranking system are distributed evenly, then individual skill at particular moments in a match will matter and it won't boil down to boring matches.

I'm just referring to what you suggested that perfectly balanced teams would result in boring matches.

[Wil McCullough]

justcallme A S H, on 25 April 2019 - 08:30 PM, said:

Yes and no.

I've carried many - no exaggeration - MANY games from 0-3 to a 12-10. There are many factors involved but it comes down to largely skill.

Yes there is snowball but it can be stopped with good game-play. That is the issue however, good game-play is something only a small handful of this player-base has. For whatever reason the vast majority of players in this game just aren't any good. You're talking 100 or so top tier and maybe a couple 1,000 of decent. The rest are all just in the bucket of 35,000 of average to low skill. And not being rude, just the fact of the matter.

Agree on the point of lower tier match being more even because it contains less top shelf players. Top shelf capitalising is what turns the tide of any game. More so MWO because of the skill disparity in a T1-T3 game because of how many people in T1-T3 only play at a T4 level but have "grinded" the EXP bar to end up in higher places than they should ever be.

Being in a higher tier than you should be also gives you a false sense of skill. I've met many a "T1" player who is outright garbage. They have skills barely above what I would class as basic yet they have played 5,000 games. This is the problem with constantly rewarding people for playing poorly which MWO does.

Let's not be coy haha. You're one of the top players in this game. Your ability to carry potatoes in qp is far greater than most. I was kinda top 1000 for a while and even i couldn't really carry 3 potatoes. Going down by 3 right from the start of a match may be something you can recover from personally but for most players, it's almost a.guaranteed sign of an incoming goon stomp.

[PhoenixFire55]

Nesutizale, on 25 April 2019 - 11:30 PM, said:

I think Phoenix and I are even before the part of the specific workings of a MM but what it should achive.

I'm actually way past that part. Like 4 years past. I've no illusions whatsoever regarding this Minimally Viable Product^TM.

Nesutizale, on 25 April 2019 - 11:30 PM, said:

Like in a footballgame with two equal teams. The ball goes back and furth but nothing realy happens.

This is a totally ridiculous and untrue statement based on nothing.

[PhoenixFire55]

MrMadguy, on 26 April 2019 - 12:11 AM, said:

If we can't achieve fair 50/50 chances to win in every single match, then at least there should be fair 50/50 chance to get advantage/disadvantage. There is nothing bad ...

Wait wait wait ... But thats exactly what current MM does. It puts you on a pre-determined 12-0/12-3-ish and 0-12/3-12-ish stomps with roughly the same chance. Are you telling us thats "fun" and "nothing bad" now?

[FRAGTAST1C]

PhoenixFire55, on 26 April 2019 - 02:29 AM, said:

Wait wait wait ... But thats exactly what current MM does. It puts you on a pre-determined 12-0/12-3-ish and 0-12/3-12-ish stomps with roughly the same chance. Are you telling us thats "fun" and "nothing bad" now?

No, if you didn't misquote him, you'd understand that he's saying losing in a match 'cause of MM setting it up that way isn't bad AS LONG AS IT DOESN'T HAPPEN MULTIPLE TIMES IN A ROW. Nowhere did he say that it's fun.

[PhoenixFire55]

Nesutizale, on 25 April 2019 - 10:46 AM, said:

Then what is grouping of players by tiers if its not a match making?

Useless bandaids.

Nesutizale, on 25 April 2019 - 10:46 AM, said:

Yes statisticly it balances itself out at what team sizes?

It depends on the margin in the score difference you consider "balacned" as well as the probability to get inside said margin you use as a threshold. But quite frankly even 8v8 match is very balanced if you randomly draw players from a skill pool with a normal-distribution. 12v12 would be well balanced by nearly every reasonable criteria.

Nesutizale, on 25 April 2019 - 10:46 AM, said:

For every mistake I make someone else also does one and for every good action there is a good action on the other team, that is what you discribe.

So did it matter what I did? Everyone else could have done that. There is not one player where you could say "thats the guy that messed it up" or "that one did pretty good" as they would be equaly. Would it matter of I get better? I will be nothing special if I stay at my level and I will be nothing special if I go up or down the ladder. Is that what you want?

You are thinking in terms of probabilities of things happening, which is quite frankly irrelevant in any particular match, where those probabilities transform into actual events, that will in turn be different from the scenario that is most likely based on probabilities, with 100% probability.

Think of every player having a normal distribution of his impact on a match vs probability of it happening centered around his average performance. In a rough model a team performance is a sum of these distribution, obviously a normal distribution as well. Now, if teams are balanced, then the sum distributions of them lay basically in the same spot, and hence every player underperforming/overperforming can cause the team impact/performance to deviate significantly from an opposing team, which means he is directly affecting the outcome. And in contrast, if teams aren't balanced, and thus their normal distributions are much further apart from one another compared to the probability swing of a single player, it means that any actions of a single player are quite frankly irrelevant towards the outcome of the match given all other players perform close to their average expectations (which overall happens with almost 100% probability due to LoLN again).

Nesutizale, on 25 April 2019 - 10:46 AM, said:

Do matches that allways run equaly make memorable matches or do you remember the matches that where good or bad?
Most people remember the good and the bad, not the normal things that happen. What games are you talking about or thinking about? Is it the game where you went through highs and lows or the okay ones? (I mean that in general not only MWO)

Again, this is some sort of BS kindergarden evaluation criteria regarding matches. Memorable? ... Like wtf does this have to do with the question at hand ... quality of matchmaking? You want memorable roflstomps of baby seals or better yet, getting roflstomped like a baby seal you are? Thats not memorable, thats just dumb.

Nesutizale, on 25 April 2019 - 10:46 AM, said:

You mix something up. Buckets is for different gamemodes like QP, FP, Solaris.

No. You just don't understand what "bucket" is. It is any division of players into separate groups in terms of MM, i.e. groups from which people can't end up in a same match, be it different game modes, different servers or different PSR tiers.

Nesutizale, on 25 April 2019 - 10:46 AM, said:

Realisticly you will most likely never have these exact number. That means you need more people of lower tiers then higher tiers but when more higher tier players are present you get unbalanced games, hance you would have to wait until more people show up to balance that out, in the worst case you need to wait for another 24 people group to form to have a balanced game.

Again, this is utter BS. Skill distribution is symmetrical and there is an equal amount of high and low skill players.

Nesutizale, on 25 April 2019 - 10:46 AM, said:

Also looking at the Steam number for a rough estimation at peaks we have ~1300 people with an average of ~700 a day. Take the average and divide them by timezones. Since some overlapping will happen I did say we have at least 4 Zones at any given time, more likely more but lets go with 4. That would mean the matchmaker would have 175 people to choose from on average.

And again, clueless BS assumptions. You have no stats on how many people are playing in each timezone, no stats on how long is the time window during the day in which they play is. I find it hilarious that you are trying to discuss MM without a simple understanding of the fact that the distribution of players along the timezones can very well be non-uniform.

[PhoenixFire55]

FRAGTAST1C, on 26 April 2019 - 02:37 AM, said:

No, if you didn't misquote him, you'd understand that he's saying losing in a match 'cause of MM setting it up that way isn't bad AS LONG AS IT DOESN'T HAPPEN MULTIPLE TIMES IN A ROW. Nowhere did he say that it's fun.

Well, I got bad news for you and him ... In any kind of 50/50 experiment, be it a coin toss or a MM setting you up for a guaranteed win/loss, there is a nearly 100% probability that the amount of same events happening in a row will be equal to square root of the number of experiments done. I.e. you play 100 matches, be prepared to have a 10 match long win/loss series with near 100% chance.

[Knight Captain Morgan]

If nothing else, matchmaker could at least start putting some of the discos and assaults who go AFK for the 1st 1/2 of the match on red team instead of dumping them all on blue team.
/sigh only 137 days to go.

Edited by Andres Gomez, 26 April 2019 - 04:22 AM.