184 replies to this topic

[Mcgral18]

Wintersdark, on 29 November 2017 - 07:22 PM, said:

While I'm with you here, there are a lot of very important details that matter.

Do scores decay? Which players are counted? I haven't played in months, but say I was in the top 20%, am I still? This matters a great deal, or you have situations where players leave the game and basically lock out slots. Due to bell curve skill levels, they'd be predominantly in the middle ranks when they left the game, leaving fewer (say) T3 "slots".

The problem with these things is that guys always come out with "you just do this!" And wash their hands of it, because clearly PGI is just dumb for not doing that, but things are always more complicated.

Zero sum is perfect with a fixed pool of players, but is very problematic when players come and go... Particularly when players enter the system at one point (typically the bottom) and leave after getting higher.

Unfortunately, I'm not about all the little details
Especially not about how PSR is implemented.


I've barely played all month, but if I jump back, I won't have forgotten about how to shoot robots.
But the decay is for newcomers, not skill decay


Does it use the last 10 matches?
The average from all matches ever?

One makes changing easy, the other makes tanking hard, but more to keep on record



All I know is the current system is bad
It promotes Potato at the "Highest level of play"

[Wintersdark]

My point is that the little details aren't little.

If you stop playing, you're still occupying space in that zero sum system. It's what makes zero sum systems very difficult to actually use in practice.

I say this not to be contrary, but because there's a lot of people here who are all bandwagony about how wonderful zero sum would be, but it's not cut and dried, not simple.


A simple zero sum system breaks down severely over time as a result... Much like PSR as implemented breaks down over time.

[MischiefSC]

Mcgral18, on 29 November 2017 - 08:04 PM, said:

Unfortunately, I'm not about all the little details
Especially not about how PSR is implemented.


I've barely played all month, but if I jump back, I won't have forgotten about how to shoot robots.
But the decay is for newcomers, not skill decay


Does it use the last 10 matches?
The average from all matches ever?

One makes changing easy, the other makes tanking hard, but more to keep on record



All I know is the current system is bad
It promotes Potato at the "Highest level of play"

The concept wouldn't work. 5 top tier players could break that system with a few alts each. Also, 20% is way too huge. Scoring is also arbitrary - which is why any matchmaker needs based off w/l.

Which functionally does what you're wanting. A good Elo system for player, mech and loadout is going to break people into score tiers. You could even just have the MM break players into tiers by their score and take it as 5 increments of 20% of the active population in the last 180 days; with a decay on your score based on days since you played last.

There's a lot of ways to do it but zero sum doesn't work because player population isn't zero sum. Matches is, win/loss is, player population isn't.

[MrMadguy]

That's, what I call close and balanced match - match, where winning side is determined in 1vs1 fight. We had a lot of such matches back in Open Beta, but since release of Clans and forced 3/3/3/3 we had lost this. Reasons for this:
1) Terrible matchmaker, that should take personal skill into account, but it actually doesn't
2) Terrible balance: too overpowered Clans, too overpowered long range weapons, too overpowered Lights and Mediums, etc.
3) As result - terrible map design. Maps, that are unbalanced, too big, lack cover, etc.

[A Headless Chicken]

Well, even if they implemented "true skill" as a factor, such as by using average match score...

Judging from the distribution on Isengrim's Jarl's List (https://leaderboard.isengrim.org/), "good" players make up only the top 5% of the player base ("good" by my definition, average 300 match score, although adjusted to compensate weight class preferences on said site - that means usually around 500 damage, some kills).

Meanwhile a ~200 match score "average" player (around 300~400 damage in game, a few assists) is at the 55th percentile-ish.

Sure, you can rate by true skill but since the vast majority of players are "average", what difference is it going to make? Not enough people are going to be on at the same times, and I believe the most we will see is just a marginal improvement in game quality.

In fact, I think it just makes the better players lose out more because similar players are few and far between, while the middling players have more to complain about because they're always going to be grouped with others of similar caliber and potato-hood. Then someone'll say, there should be a failsafe in the matching system so everyone, good, average or bad will not have to wait too long for a game - and that margin will lead us right back where we are now.

Also, I'm using matchscore as a yardstick because I do think that is most telling of a player's performance in a game, though not the end-all.

Happy ever after? Yeahnahyeahnah, maybe.

Edited by A Headless Chicken, 29 November 2017 - 11:09 PM.

[Chef Kooker]

Actually more skilled a player is the more a possibility of a stomp. When 2 teams of about same skills meet, the team that makes the first mistake often lose in a game without comeback mechanic. This often turns into a snowball scenario. This is due to the other team recognizing the advantage they have and pushing it. There is always a possibility of a comeback, but if the game don't have comeback mechanics then it'll often be a stomp. In MWO you get close games if the leading team don't utilize their advantage and momentum and less skilled players will have problems with this.

Ofc weapon loadout, map and mech will influence this, but skilled players will recognize a small advantage and turn a game into a "stomp". While less skilled players will stay behind and snipe. That's why in t5 you find more games that are 11/12 10/12 than in higher tier were players got more experience.

So what is it people complain about? It's the feeling of despair when one of your team dies and you see the deathball rolling. It's possible to make a comeback but hard when you have less firepower. And since there are no comeback mechanics in the game it will increase the feeling of losses since you know that it is hard to stop the snowball. Comebacks do happen and in "high tier" it will be a rush since you know it almost never happens. But it was skills or luck that made it possible. If both teams had the same skills, the team that makes the first mistake will lose and mostly in a snowball fashion.

[Xavori]

MischiefSC, on 29 November 2017 - 08:58 PM, said:

The concept wouldn't work. 5 top tier players could break that system with a few alts each. Also, 20% is way too huge. Scoring is also arbitrary - which is why any matchmaker needs based off w/l.

Which functionally does what you're wanting. A good Elo system for player, mech and loadout is going to break people into score tiers. You could even just have the MM break players into tiers by their score and take it as 5 increments of 20% of the active population in the last 180 days; with a decay on your score based on days since you played last.

There's a lot of ways to do it but zero sum doesn't work because player population isn't zero sum. Matches is, win/loss is, player population isn't.

You don't need tiers at all. Both ELO and TrueSkill generate individual player ratings based on winning or losing versus the skill of the players you fought against. When you win, your rating goes up, and how much it goes up is based on the rating of your opponent compared to you. When you lose, your rating goes down, and again, the amount is based on the rating of your opponent compared to you. Over time, this rating will reach a point where two players with the same rating fighting each other would be expected to win 50% of the time.

The rating is also not "You are this skilled of a pilot." The rating is "Compared to other pilots, you are expected to win x% of the time based on your rating and the other pilot's rating.

The matchmaker can then take that rating and create two teams where the total player skill on both teams is roughly equivalent. The expectation is that this produces better matches because WE ALREADY KNOW IT DOES IN LOTS OF OTHER FORMS OF COMPETITIVE GAMING.

The rating also lets you create actual leaderboards because the rating is entirely comparison based built on quality, rather than random, matches.

[MischiefSC]

Xavori, on 29 November 2017 - 11:34 PM, said:

You don't need tiers at all. Both ELO and TrueSkill generate individual player ratings based on winning or losing versus the skill of the players you fought against. When you win, your rating goes up, and how much it goes up is based on the rating of your opponent compared to you. When you lose, your rating goes down, and again, the amount is based on the rating of your opponent compared to you. Over time, this rating will reach a point where two players with the same rating fighting each other would be expected to win 50% of the time.

The rating is also not "You are this skilled of a pilot." The rating is "Compared to other pilots, you are expected to win x% of the time based on your rating and the other pilot's rating.

The matchmaker can then take that rating and create two teams where the total player skill on both teams is roughly equivalent. The expectation is that this produces better matches because WE ALREADY KNOW IT DOES IN LOTS OF OTHER FORMS OF COMPETITIVE GAMING.

The rating also lets you create actual leaderboards because the rating is entirely comparison based built on quality, rather than random, matches.

Well, the tiers are fluid but essentially the goal is not to match high/low to an average but if the match goal is, say, 1600 score for each team then both teams need made of 1400 to 1800. So you've got 'tiers' in as much as matches are built in a range not to an average.

Otherwise you get a match build around a 1600 average with 1200 to 2000 or worse and it's pretty much like what we have now.

So you want 'tiers' but you want them built off actual player skill not an XP bar and you want it built around a set deviation from the target to build the match, not just to an average.

[Navid A1]

Two things:

1- Matchmaking MUST be based on weight class. What it means is that when the current MM grabs 24 players to form a match, it should count the number of Lights, meds, heavies and assaults and distribute pilots so that each team has equal rating in individual weight classes.

The way match making is done right now is lazy and dumb, since you are worth the same no matter what mech you are in. A spider is no equal to a Deathstrike with the same pilot.



2- Since PSR is heavily affected by match score, then the score calculation should be adjusted to rely less on damage... or PSR should take more things into account for each weight class.
Right now, you can just take an LRM boat and power trough max T1 in no time!



But, all of that takes effort... and it can not be monetized... so... whatever.

Edited by Navid A1, 29 November 2017 - 11:46 PM.

[A Headless Chicken]

Xavori, on 29 November 2017 - 11:34 PM, said:

The matchmaker can then take that rating and create two teams where the total player skill on both teams is roughly equivalent. The expectation is that this produces better matches because WE ALREADY KNOW IT DOES IN LOTS OF OTHER FORMS OF COMPETITIVE GAMING.

Given how games such as Overwatch, LoL, CSGO have daily rants about matchmaking being borked on Youtube, Reddit, and in screenshot form, with the devs continually trying to improve said systems, I find it hard to believe that...

1. PGI is willing/able make a proper matchmaking system.
2. A matchmaker is the end all to make everything alright.

Lets face it, there is always going to massive variance in skill levels between a bunch of people on at the same time. Beyond everyone suddenly becoming godtier pilots, there is no matchmaking system that will completely solve all this. All it can do is make "massive variance" become "some variance" and then people complain that they aren't getting games, or maybe that their teams are potating all over again after awhile because we are all headless chickens.

Edited by A Headless Chicken, 29 November 2017 - 11:48 PM.

[Navid A1]

In the current match maker you are worth the same no matter if you are in a weapon-less Spider or in a meta deathstrike!



If that is not a lazy MM algorithm, then I don't know what is!

[A Headless Chicken]

I'm not too worried about ratings for individual weight classes because last I remember, pilots make the 'Mech work for them and not vice versa. If you can aim and shoot, plus position well, you're already a top player.

Maybe I'll do a weaponless Spider for one whole day. Pray I don't get reported.

Maybe I'm just not thinking enough.

[Ignatius Audene]

Elo will not change the result guys. To small playerbase. Under Elo wait time in Solo was ground 20 min. They could prevent the MM from using different Tiers for one match. Bug again looooooooong wait time.

P.s. I am ranked in the Top 800? So much from ELO /avg. Match score /what ever

Edited by Ignatius Audene, 30 November 2017 - 12:35 AM.

[Xavori]

Ignatius Audene, on 30 November 2017 - 12:32 AM, said:

Elo will not change the result guys. To small playerbase. Under Elo wait time in Solo was ground 20 min. They could prevent the MM from using different Tiers for one match. Bug again looooooooong wait time.

P.s. I am ranked in the Top 800? So much from ELO /avg. Match score /what ever

Elo does not change wait time. It has nothing to do with wait times. All it changes is how the players who are in queue get arranged into teams.

Each player will have an Elo (or TrueSkill...this would likely be better as it's better built for teams, including unbalanced teams which would help with disconnects) rating. When the matchmaker is assembling the teams, it does so trying to get the total rating on both sides as close as possible.

It does not wait until it has equal rated players. It does not need tiers of players. It just takes the players it has waiting and mixes them as best it can into equivalently rated teams.

For example, let's say we had a quickplay 4v4 (cuz I'm not writing out 12v12 for an example) and there were 8 people in queue with the following ratings: 1700,760,900,2000,1200,1420, 890, and 630. The matchmaker would just take those 8 people and divide them into two teams:
Team A: 2000, 1200, 900, 630 (4730 total)
Team B: 1700, 1420, 890, 760 (4770 total)

The two teams aren't perfectly equal, but they're pretty close, and since that's what the matchmaker had to work with at that moment, that's what it's going with. And the algorithm for putting the two teams together is stupidly easy and fast to run. And for group play, you just treat each group as a single entity for rating with the only added difficulty being that the total number of players needs to end up at 12.

So no, no increased wait times. Just higher probability that the two teams are evenly matched.

[Ignatius Audene]

Mixing tier5 to 1. Simply no.

There is a reason for leagues in sport. The only way to use elo is after MM put together 24 poeple with the current system, afterwards control ELO and switch players from one team to the other.

Edited by Ignatius Audene, 30 November 2017 - 05:06 AM.

[Xavori]

Ignatius Audene, on 30 November 2017 - 04:53 AM, said:

Mixing tier5 to 1. Simply no.

There is a reason for leagues in sport. The only way to use elo is after MM put together 24 poeple with the current system, afterwards control ELO and switch players from one team to the other.

It'd be Johnson and Johnson's Baby Shampoo. No more tiers.

Tiers don't work because they are too broad and cannot account for player population without just giving up on the tiers during matchmaking (which is what we have now). If there aren't enough T5 players, it looks at T4, still not enough, go to T3. Blech.

With a player rating built on ELO, WHR, TrueSkill, etc., each player has their own individual rating. That's what gets used to put together the team. It's also how you rank people on a leaderboard (well, not purely in TrueSkill, but even there it's the start). A player with a higher ELO rating is expected to beat a player with a lower ELO more than half the time.

And no, ELO doesn't shuffle afterwards. That's wasted time and processor cycles. You can really for reals tell a computer to put together two equivalently rated teams on the fly. Computers are smart like that.

[Bigbacon]

it wouldn't really solve anytihng. you'd still cry and ***** and moan about crap games.

[Weeny Machine]

Sometimes all you can do is just shrug and accept that stomps are there. Life gets suddenly a lot easier and I doubt any system could prevent it.


Sidenote: What is more unnerving but also somehow amusing when someone attacks you for having played bad after a stomp (I did 430 dmg in a medium) while most others, that particular dude as well, did 100 damage. That is the moment when you wonder wtf is going on hahaha

Edited by Bush Hopper, 30 November 2017 - 06:50 AM.

[Mole]

Bush Hopper, on 30 November 2017 - 06:42 AM, said:

Sometimes all you can do is just shrug and accept that stomps are there. Life gets suddenly a lot easier and I doubt any system could prevent it.

Pretty much this. I've found the game to feel much more rewarding and less frustrating if I just focus on doing well and if I lose, I lose. If I win, cool. Either way, I don't get bent out of shape about it. I have a ton more fun this way.

[Asym]

Sorry, a little rant: I just discovered my MW3 green battlefield reference manual and the memories of such a good time.....

Again, the player base is less than a 1,000 people at anyone time.... Days without being able to even find FP matches..... Long waits for QP. Virtually no one in off periods unless there is another everyday event.

What in the world does this very good discussion, by some very smart players, do to improve or to make "vale added" what we have? We, as a system, are so broken we will not recover back to what was....... It's now three days w/o being able to find more than 3 FP matches a day !!! The last event just ended........can you hear the silence?

What was said in Arpil of this year and confirmed in May, with the new skill tree, is the "new reality": PGI de-evolved the entire platform to create a space for a FPS arcade platform: Solaris. Hairbrain has the Battle tech side of the MW universe so lore goes there. PGI has the FPS side of the Universe and a dedicated, very small, lucrative and always buying extreme niche that is profitable. 1x1, 2x2 and e-Sports is the new reality. 4x4 and 8x8 they'll keep to farm new players into Solaris or whatever PGI creates for the Arcade gameplay platform....

Sorry, the writing is on the wall. Look at the advertising list, the BT announcements and displays at MECON and I'd love to hear how elo or any "just" MM system will help an arcade, FPS'ing platform, that buys, no matter what happens, everything.....

Sorry, I love this game since inception years and year ago.