147 replies to this topic

[Vancer2]

Never, now shut up.

[Alan Hicks]

8v8, then 4v4, then 2v2, then bye-bye MWO.

[Imperius]

Just quit playing the game eventually the low population may finally make them pull the trigger on 8v8....

Oh wait no it it won’t

I forgot the original reason 12v12 was forced on us causing tons of imbalances and performance issues was to save SERVER COST.

No colorblind support on the paper doll still.

I’ve given up because I’m tired of fighting a fight shouldn’t even have to.

[Champion of Khorne Lord of Blood]

Mystere, on 31 October 2017 - 06:27 AM, said:

What about being bait? Vanguard? Distraction? Command? Leadership? Inspiration?

A great leader (whether or not a terrible fighter/shooter/whatever) who stays out of the fight while consistently giving out commands that make the team win will lose out on your "system".

As it should be in my opinion. A great leader who does all of what your hiding leader does while also leading by example in the front will move up while backseat armchair generals fall.

Players who play bait but also take someone down with them while they do it will move up, players who are on the vanguard and actually defended against an attack rather than hide in the back all the match move up, players that distracted the enemy by firing on them enough to pull their attention off other players move up, players who command and lead from the front move up, nothing is more inspiring than a teammate who runs up front and crushes all opposition, many times the Atlases who run up front to "inspire" their team only serve to crush morale when they get vaporized the second they step into the firing line.


I'm honestly really tired of the whole "but look at all these roles that help bring about victory that give bad match score" because the things that do give good match score tend to be even better at bringing about victory and are more solid than just the chances that running out front and "baiting" actually resulted in the enemy's defeat and not just that the rest of your team was good enough to carry while being a man down.

The whole point of match maker anyway was to put people with similar skill levels together, so all the people with great aim are together while the guys who have bad aim are down below. If someone is getting lots of kills I feel that they should move up, even if it is based on kill stealing or other forms of "gaming" the system they are still helping out the win by making sure that crippled player doesn't get another alpha strike off or run off to cap. If someone's putting out a load of damage but its spread, they're still weakening the enemy for the sake of all the other players on the team, resulting in much easier victory. Kills, assists, damage, all that really matters in the end, you can go and command, bait, and flank all day but if you didn't get any kills, assists, or damage then your team has no way to win.
The whole point of match maker anyway was to put people with similar skill levels together, so all the people with great aim

Edited by Dakota1000, 01 November 2017 - 12:39 PM.

[MischiefSC]

Dakota1000, on 01 November 2017 - 12:33 PM, said:

As it should be in my opinion. A great leader who does all of what your hiding leader does while also leading by example in the front will move up while backseat armchair generals fall.

Players who play bait but also take someone down with them while they do it will move up, players who are on the vanguard and actually defended against an attack rather than hide in the back all the match move up, players that distracted the enemy by firing on them enough to pull their attention off other players move up, players who command and lead from the front move up, nothing is more inspiring than a teammate who runs up front and crushes all opposition, many times the Atlases who run up front to "inspire" their team only serve to crush morale when they get vaporized the second they step into the firing line.


I'm honestly really tired of the whole "but look at all these roles that help bring about victory that give bad match score" because the things that do give good match score tend to be even better at bringing about victory and are more solid than just the chances that running out front and "baiting" actually resulted in the enemy's defeat and not just that the rest of your team was good enough to carry while being a man down.

The whole point of match maker anyway was to put people with similar skill levels together, so all the people with great aim are together while the guys who have bad aim are down below. If someone is getting lots of kills I feel that they should move up, even if it is based on kill stealing or other forms of "gaming" the system they are still helping out the win by making sure that crippled player doesn't get another alpha strike off or run off to cap. If someone's putting out a load of damage but its spread, they're still weakening the enemy for the sake of all the other players on the team, resulting in much easier victory. Kills, assists, damage, all that really matters in the end, you can go and command, bait, and flank all day but if you didn't get any kills, assists, or damage then your team has no way to win.
The whole point of match maker anyway was to put people with similar skill levels together, so all the people with great aim

The best way, hands down, for me to pad match score is run Clan streaks and hide/shut down/avoid direct odds of getting shot. Also LRMs but way more iffy. Boating UAC2s also works. And AMS.

None of which is good for winning matches. With most my MPL boats I can solo mechs with 100% CT damage. With my DS Lite 2xcerml, 2xhll, 2xgauss MC MKII I get a headshot kill about once every 2 or 3 matches - look for the LRM boat. You need to be on the high ground to pop a helmet on a Stalker but Highlanders, Maulers and BLRs are all easy.

All of which gets me mediocre damage. Same with pulling legs. I don't get the component destruction. It leaves the enemy in my teammates fire however and wins matches.

Match score is not about winning matches. It's about match score. You can use match score to drill down on WHY someone wins matches but your ability to win matches is only directly represented in....

How often you win matches. There's no magic here. If you win a lot of your matches you're... historically more likely to win matches. Damage, kills, AMS use, match score, KDR, comms use, mech choice, play schedule, there's tons of metrics that can be looked at for impact on driving wins. They are certainly potentially related and if I want to look at WHY someone wins (or loses) I can drill down on those but it absolutely does NOT replace your w/l as the most useful predictor of.... that's right. Winning.

[Champion of Khorne Lord of Blood]

MischiefSC, on 01 November 2017 - 12:58 PM, said:

The best way, hands down, for me to pad match score is run Clan streaks and hide/shut down/avoid direct odds of getting shot. Also LRMs but way more iffy. Boating UAC2s also works. And AMS.

None of which is good for winning matches. With most my MPL boats I can solo mechs with 100% CT damage. With my DS Lite 2xcerml, 2xhll, 2xgauss MC MKII I get a headshot kill about once every 2 or 3 matches - look for the LRM boat. You need to be on the high ground to pop a helmet on a Stalker but Highlanders, Maulers and BLRs are all easy.

All of which gets me mediocre damage. Same with pulling legs. I don't get the component destruction. It leaves the enemy in my teammates fire however and wins matches.

Match score is not about winning matches. It's about match score. You can use match score to drill down on WHY someone wins matches but your ability to win matches is only directly represented in....

How often you win matches. There's no magic here. If you win a lot of your matches you're... historically more likely to win matches. Damage, kills, AMS use, match score, KDR, comms use, mech choice, play schedule, there's tons of metrics that can be looked at for impact on driving wins. They are certainly potentially related and if I want to look at WHY someone wins (or loses) I can drill down on those but it absolutely does NOT replace your w/l as the most useful predictor of.... that's right. Winning.

If someone can maintain high KDR and high damage rates every game while just running a streak/LRM boat against players who are good and have good aim then maybe they deserve to move on up, as they're actually putting in work. That said, the players like you getting efficient kills have less chance of the kill being stolen from you, so you move up just as fast or faster than the guy spreading damage but putting out loads of it. Besides, it could be weighted so that kills matter the most, then damage, then assists, rewarding accurate players who cut down enemies more than people who cover by putting out high damage while still letting the people who put out high damage that lead to a kill move up by getting damage+assist.

I disagree that how often you win really should be a predictor of how much you will win, that's like going out in stocks and seeing that a price is going up so that is a predictor that it will just always go up, which is just wrong, there are actual factors as to why things rise and fall, in stocks its usually new advances, sales, and events, with MWO the reason people win is because they went out and got a lot of kills or put out a lot of damage or assisted their team in killing enemies. You don't just win a match because you won your last match. More often than not you can win a match while doing terrible in that match because other players carried, we can see the current heavily win biased PSR system isn't working just based off the low level of ability found in many T1 players.

[Mystere]

Dakota1000, on 01 November 2017 - 12:33 PM, said:

As it should be in my opinion. A great leader who does all of what your hiding leader does while also leading by example in the front will move up while backseat armchair generals fall.

Players who play bait but also take someone down with them while they do it will move up, players who are on the vanguard and actually defended against an attack rather than hide in the back all the match move up, players that distracted the enemy by firing on them enough to pull their attention off other players move up, players who command and lead from the front move up, nothing is more inspiring than a teammate who runs up front and crushes all opposition, many times the Atlases who run up front to "inspire" their team only serve to crush morale when they get vaporized the second they step into the firing line.


I'm honestly really tired of the whole "but look at all these roles that help bring about victory that give bad match score" because the things that do give good match score tend to be even better at bringing about victory and are more solid than just the chances that running out front and "baiting" actually resulted in the enemy's defeat and not just that the rest of your team was good enough to carry while being a man down.

The whole point of match maker anyway was to put people with similar skill levels together, so all the people with great aim are together while the guys who have bad aim are down below. If someone is getting lots of kills I feel that they should move up, even if it is based on kill stealing or other forms of "gaming" the system they are still helping out the win by making sure that crippled player doesn't get another alpha strike off or run off to cap. If someone's putting out a load of damage but its spread, they're still weakening the enemy for the sake of all the other players on the team, resulting in much easier victory. Kills, assists, damage, all that really matters in the end, you can go and command, bait, and flank all day but if you didn't get any kills, assists, or damage then your team has no way to win.
The whole point of match maker anyway was to put people with similar skill levels together, so all the people with great aim

How many generals these days lead an assault on the front tank/AFV? Did Napoleon ride at the apex of the cavalry charge?

How often did "Joe the sharpshooter" become Commanding General of the Army? Heck, how many World Cup or NBA coaches were former MVP players?

Commanding is a very different skill from shooting/killing/damaging. Why should the latter be rewarded more? Heck, again, why should a player dealing a whole lot of ineffective damage be given more c-bills and XP than the player who commanded the team to victory via sound tactical choices?

Edited by Mystere, 01 November 2017 - 02:06 PM.

[Khobai]

Quote

How many generals these days lead an assault on the front tank/AFV? Did Napoleon ride at the apex of the cavalry charge?

mech commander had the right idea for how commanding should work in the battletech universe.

commanders would be in a remote location away from the battlefield, and play the game like an RTS giving orders to everyone on the team. instead of actually piloting a mech themselves.

thats how MWO shouldve been. one person is the commander, and they just hang out inside the mobile field base and command the team from a bird's eye view of the battlefield. And maybe give them the option to hop into a mech if they absolutely need to, but theyd give up all their commander abilities.

Edited by Khobai, 01 November 2017 - 02:25 PM.

[El Bandito]

Mystere, on 01 November 2017 - 02:03 PM, said:

How many generals these days lead an assault on the front tank/AFV? Did Napoleon ride at the apex of the cavalry charge?

How often did "Joe the sharpshooter" become Commanding General of the Army? Heck, how many World Cup or NBA coaches were former MVP players?

Commanding is a very different skill from shooting/killing/damaging. Why should the latter be rewarded more? Heck, again, why should a player dealing a whole lot of ineffective damage be given more c-bills and XP than the player who commanded the team to victory via sound tactical choices?

Mate, this is a video game, not real life. I have seen many successful FP drop callers leading from the front.

[Champion of Khorne Lord of Blood]

Mystere, on 01 November 2017 - 02:03 PM, said:

How many generals these days lead an assault on the front tank/AFV? Did Napoleon ride at the apex of the cavalry charge?

How often did "Joe the sharpshooter" become Commanding General of the Army? Heck, how many World Cup or NBA coaches were former MVP players?

Commanding is a very different skill from shooting/killing/damaging. Why should the latter be rewarded more? Heck, again, why should a player dealing a whole lot of ineffective damage be given more c-bills and XP than the player who commanded the team to victory via sound tactical choices?

This is mechwarrior, we're practically ancient knights charging in at the enemy team, our leaders making glorious charges beside us into battle. The whole universe is based off stylized versions of warfare, glorifying mechwarriors as avatars of destruction with commanders leading by example. Even the previous games such as all of the Mechwarrior 4 series had you as a prince of Davion, or a mercenary commander leading your lance from the front.

If we're going to go and bring real life examples into this, how often do you see tanks drive out into positions with the sole purpose of dying for bait, how often do you see someone run out into enemy firing lines in an attempt to boost morale, how often does the commander not fighting reduce the army's strength by over 8%?

The reason shooting/killing/damaging should reward more than commanding is because this game is a shooter, not an RTS and if you're trying to play it like an RTS instead of tactical combat then you're a hindrance to the team.

[Akillius]

MWO is unstable with constant "re-balancing" for you water lilys and looks bad with massive changes done to it's fundamental structure like skills. Haven't been without a unit until 2 months after new skill tree (promised 100% cbills for module reneged to no cbills for 2016-or-older modules and double standard cbills for 2017 modules) and I saw the last hold outs were like ffs outta here that was last august. Tried chatting with some since in other games but they aren't returning because of ******** like that and 8v8 and other unstable ******** announced this past year...

Summary:
8vs8 is fine in CW/FP or for "competitive".

But quickplay is fine in 12vs12 solo/team but it really should be 16vs16.
Change QP to 8v8 and it'll be the prime reason for me to finally cancel ALL orders and leave this game too.

[Mystere]

El Bandito, on 01 November 2017 - 02:22 PM, said:

Mate, this is a video game, not real life. I have seen many successful FP drop callers leading from the front.

What can I say? I want some of my games to have some semblance of reality. <shrugs>

[Mystere]

Dakota1000, on 01 November 2017 - 02:51 PM, said:

The reason shooting/killing/damaging should reward more than commanding is because this game is a shooter, not an RTS and if you're trying to play it like an RTS instead of tactical combat then you're a hindrance to the team.

WUT? A commander who consistently wins battles by leading without being directly involved in the actual fighting is a hindrance to the team?

Edited by Mystere, 01 November 2017 - 06:48 PM.

[El Bandito]

Mystere, on 01 November 2017 - 06:43 PM, said:

What can I say? I want some of my games to have some semblance of reality. <shrugs>

Then you are playing the wrong game.

[MischiefSC]

Dakota1000, on 01 November 2017 - 01:12 PM, said:

If someone can maintain high KDR and high damage rates every game while just running a streak/LRM boat against players who are good and have good aim then maybe they deserve to move on up, as they're actually putting in work. That said, the players like you getting efficient kills have less chance of the kill being stolen from you, so you move up just as fast or faster than the guy spreading damage but putting out loads of it. Besides, it could be weighted so that kills matter the most, then damage, then assists, rewarding accurate players who cut down enemies more than people who cover by putting out high damage while still letting the people who put out high damage that lead to a kill move up by getting damage+assist.

I disagree that how often you win really should be a predictor of how much you will win, that's like going out in stocks and seeing that a price is going up so that is a predictor that it will just always go up, which is just wrong, there are actual factors as to why things rise and fall, in stocks its usually new advances, sales, and events, with MWO the reason people win is because they went out and got a lot of kills or put out a lot of damage or assisted their team in killing enemies. You don't just win a match because you won your last match. More often than not you can win a match while doing terrible in that match because other players carried, we can see the current heavily win biased PSR system isn't working just based off the low level of ability found in many T1 players.

You're trying to equate stock market analysis with matchmaking predictive analysis. There is no correlation except that both have math in them. Process, purpose and every single analytical tool, criteria, equation and formula you use for them is different.

The only thing you're trying to do with a matchmaker, the only thing at all, is match up the teams odds of winning and establish which has better odds of winning while trying to keep them as close as possible. You're not trying to match damage, or kills, or match score. Just who is more likely to win and try to keep comparable odds of winning people on both sides.

Odds of winning. That's it. Not score. You're not building matches to make sure both teams get comparable damage. Just odds of winning.

Odds.

Of winning.

Winning.

A lot goes into winning at anything. Numerous factors and you can study those factors to see why this particular subject (be in a player or a dice or a car or whatever) wins at whatever it's trying to do more or less often but if you're not looking at why, just win/lose, all that matters is the winning.

Suppose I have 24 quarters and all of them are slightly unbalanced. I want to get 2 sets of 12 quarters that each average 50% heads and 50% tails. They could be unbalanced because sides are clipped or shaved or notches cut out or bent so as to make them aerodynamically unstable on a particular side or dozens of other factors. You could make a study of those factors and determine their relative impacts on each quarters odds of coming up heads or tails -

however all you're trying to do is balance each set of 12 to be 50% of the time heads, 50% tails, so why doesn't ******* matter. At all. Even a little bit. Once, just for special occasion. You flip each coin 100 times the same way by the same person in the same environment and you get the most accurate results.

Everyone plays in pug queue with 11 people selected by the same matchmaker. All those odds wash out. Law of large numbers. All that you're left with is your specific playing impacting your team each match to influence its odds of winning or losing. We're all being flipped by the same guy in the same setting the exact same way. Sure, sometimes it's harder than others but because we're all getting flipped a lot of times by the same guy the same quirks balance out to give a consistent relative value.

Why you win or lose doesn't matter for matchmaking. It's irrelevant. It's like taking your favorite color into account. You're not matching teams based on WHY they win or lose or HOW. Just their odds of winning or losing.

I don't know how else to put this. Please google statistical analysis, I put law of large numbers up because most people don't understand that and it's important. Posting links to stuff like statistics and predictive analysis feels like it would come across really snotty and that's not my intent. I want a better matchmaker but you can't get that with garbage data. That's going to be any data other than win/loss for predicting pug matches. You can drill down by having a win/loss for mechs and loadouts, that would be useful. Even a player specific breakdown of mechs and loadouts. That would probably be more depth and precision than is useful with our low population. However match score, damage, kills, whatever is a HOW and a WHY. Any attempt to include those in the win/loss skews your numbers by double-dipping their impact or misrepresenting.

Just win/loss. That's it. No more, no less. You can get super detailed by historical data about chassis and builds and how each player plays them and if you really, really want to get crafty how they play them on specific maps and against specific enemy chassis/builds but again, more detail than we probably need. It's all going to be around win/loss though. Nothing else has a place in an accurate predictive model.

[SFC174]

MischiefSC, on 01 November 2017 - 07:45 PM, said:

Just win/loss. That's it. No more, no less. You can get super detailed by historical data about chassis and builds and how each player plays them and if you really, really want to get crafty how they play them on specific maps and against specific enemy chassis/builds but again, more detail than we probably need. It's all going to be around win/loss though. Nothing else has a place in an accurate predictive model.

The only caveat to this is how many games do you need to have a big enough sample that ensures your teammate selection is sufficiently randomized? I've started this season 2-10 with a 300 avg match score and 1.5 KDR. It's been potatoes vs. high level players for most of those matches and very imbalanced. I'm sure it will balance out over the season, because I'm not that bad a player. So you may need to give some weight to other stats beyond W/L with their weight decreasing as number of matches increase.

[Champion of Khorne Lord of Blood]

MischiefSC, on 01 November 2017 - 07:45 PM, said:

You're trying to equate stock market analysis with matchmaking predictive analysis. There is no correlation except that both have math in them. Process, purpose and every single analytical tool, criteria, equation and formula you use for them is different.

The only thing you're trying to do with a matchmaker, the only thing at all, is match up the teams odds of winning and establish which has better odds of winning while trying to keep them as close as possible. You're not trying to match damage, or kills, or match score. Just who is more likely to win and try to keep comparable odds of winning people on both sides.

Odds of winning. That's it. Not score. You're not building matches to make sure both teams get comparable damage. Just odds of winning.

Odds.

Of winning.

Winning.

A lot goes into winning at anything. Numerous factors and you can study those factors to see why this particular subject (be in a player or a dice or a car or whatever) wins at whatever it's trying to do more or less often but if you're not looking at why, just win/lose, all that matters is the winning.

Suppose I have 24 quarters and all of them are slightly unbalanced. I want to get 2 sets of 12 quarters that each average 50% heads and 50% tails. They could be unbalanced because sides are clipped or shaved or notches cut out or bent so as to make them aerodynamically unstable on a particular side or dozens of other factors. You could make a study of those factors and determine their relative impacts on each quarters odds of coming up heads or tails -

however all you're trying to do is balance each set of 12 to be 50% of the time heads, 50% tails, so why doesn't ******* matter. At all. Even a little bit. Once, just for special occasion. You flip each coin 100 times the same way by the same person in the same environment and you get the most accurate results.

Everyone plays in pug queue with 11 people selected by the same matchmaker. All those odds wash out. Law of large numbers. All that you're left with is your specific playing impacting your team each match to influence its odds of winning or losing. We're all being flipped by the same guy in the same setting the exact same way. Sure, sometimes it's harder than others but because we're all getting flipped a lot of times by the same guy the same quirks balance out to give a consistent relative value.

Why you win or lose doesn't matter for matchmaking. It's irrelevant. It's like taking your favorite color into account. You're not matching teams based on WHY they win or lose or HOW. Just their odds of winning or losing.

I don't know how else to put this. Please google statistical analysis, I put law of large numbers up because most people don't understand that and it's important. Posting links to stuff like statistics and predictive analysis feels like it would come across really snotty and that's not my intent. I want a better matchmaker but you can't get that with garbage data. That's going to be any data other than win/loss for predicting pug matches. You can drill down by having a win/loss for mechs and loadouts, that would be useful. Even a player specific breakdown of mechs and loadouts. That would probably be more depth and precision than is useful with our low population. However match score, damage, kills, whatever is a HOW and a WHY. Any attempt to include those in the win/loss skews your numbers by double-dipping their impact or misrepresenting.

Just win/loss. That's it. No more, no less. You can get super detailed by historical data about chassis and builds and how each player plays them and if you really, really want to get crafty how they play them on specific maps and against specific enemy chassis/builds but again, more detail than we probably need. It's all going to be around win/loss though. Nothing else has a place in an accurate predictive model.

Law of large numbers would lead to matches being balanced on average just through sheer brute force, but it still lets the matches be entirely unbalanced, just that you have around the same number of wins as losses. In short, reasoning such as yours is why we have so many stomps and a matchmaker that so many people dislike.

I'd rather raise the average quality of matches rather than leave it as is by actually taking a look at the how and why of winning and putting like-minded players together, resulting in more balanced teams, more back and fourth on matches, and generally less flat out stomping that results from a purely win/loss based system weeding out those who have been carried. Leaving it up to law of large numbers seems no better than just having no matchmaker at all, because if we use purely random matchmaker then over an infinite amount of trials you'll have balance just the same as this.

I think the issue here is that I fundamentally disagree with you on the purpose of a matchmaker. You say that a MM is only supposed to set you up to have a 50% win rate while I say that a match maker should pair you up with players at the same skill level as you, which would eventually result in you having a 50% win rate as you settle in. I also feel that the teams should have put out similar damage, kills, and assists on each team rather than just having similar odds of winning each match, that would get rid of the main complaint that players have of stomps being so common.

Due to the low population, and the extremely low skilled population (as seen by the Jarl's List percentiles) I feel that we need a matchmaker that has a bit more depth and precision if we want to have any semblance of balanced matches, otherwise we may as well just remove matchmaking entirely and leave it random.

I've been through statistical analysis classes and used it in my profession and hobbies, I'm very inclined to look into the reasons why on any stats I go through for the sake of balance, which leads me to find a purely win/loss based matchmaker entirely unacceptable, as it doesn't balance anything quickly or directly, it just lets averages wash out over time, resulting in a bad play experience for people throughout that entire period of averaging that never ends.

Mystere, on 01 November 2017 - 06:47 PM, said:

WUT? A commander who consistently wins battles by leading without being directly involved in the actual fighting is a hindrance to the team?

Find me a commander who consistently wins battles by leading in MWO while sitting in spawn, then compare his win rate to the average unit dropcaller. If that guy is consistently winning then the statistics demand that the guy doing the same quality of leading while also shooting is consistently winning more.

[Skipmagnet]

8v8 will be the nail in the coffin for me. I can contribute in 12v12 cuz there are enough bodies for me not to be vaporized first some of the time. With 4 fewer players, that is more attention on me, and I know I'm not good enough in my current tier, play pool and mech choice to withstand that. An 8 man with me on it might as well be a 7 man for all the good I will do at that point.

So, yeah. 8v8 is when I bow out and stop ******** up your matches. Yay, I guess?

[MischiefSC]

SFC174, on 01 November 2017 - 08:20 PM, said:

The only caveat to this is how many games do you need to have a big enough sample that ensures your teammate selection is sufficiently randomized? I've started this season 2-10 with a 300 avg match score and 1.5 KDR. It's been potatoes vs. high level players for most of those matches and very imbalanced. I'm sure it will balance out over the season, because I'm not that bad a player. So you may need to give some weight to other stats beyond W/L with their weight decreasing as number of matches increase.

About 40 will seat you pretty accurately, 80 is tight, for 12 v 12 you really want about 160-180 total matches (not a season, but as recently as possible) to seat someone within a percent or two.

Side note -

Flip a coin 1,024 times and at one point you'll get heads 10 times in a row.

However you'll end with 512 heads, 512 tails, +/- 2 over 99.9% certainty.

Tide comes in, tide rolls out, ocean stays the same.

Dakota1000, on 01 November 2017 - 08:23 PM, said:

Law of large numbers would lead to matches being balanced on average just through sheer brute force, but it still lets the matches be entirely unbalanced, just that you have around the same number of wins as losses. In short, reasoning such as yours is why we have so many stomps and a matchmaker that so many people dislike.

I'd rather raise the average quality of matches rather than leave it as is by actually taking a look at the how and why of winning and putting like-minded players together, resulting in more balanced teams, more back and fourth on matches, and generally less flat out stomping that results from a purely win/loss based system weeding out those who have been carried. Leaving it up to law of large numbers seems no better than just having no matchmaker at all, because if we use purely random matchmaker then over an infinite amount of trials you'll have balance just the same as this.

I think the issue here is that I fundamentally disagree with you on the purpose of a matchmaker. You say that a MM is only supposed to set you up to have a 50% win rate while I say that a match maker should pair you up with players at the same skill level as you, which would eventually result in you having a 50% win rate as you settle in. I also feel that the teams should have put out similar damage, kills, and assists on each team rather than just having similar odds of winning each match, that would get rid of the main complaint that players have of stomps being so common.

Due to the low population, and the extremely low skilled population (as seen by the Jarl's List percentiles) I feel that we need a matchmaker that has a bit more depth and precision if we want to have any semblance of balanced matches, otherwise we may as well just remove matchmaking entirely and leave it random.

I've been through statistical analysis classes and used it in my profession and hobbies, I'm very inclined to look into the reasons why on any stats I go through for the sake of balance, which leads me to find a purely win/loss based matchmaker entirely unacceptable, as it doesn't balance anything quickly or directly, it just lets averages wash out over time, resulting in a bad play experience for people throughout that entire period of averaging that never ends.

All win/loss is averages. You want to drill down do it by mech and loadout. Functionally player Elo, mech Elo, then Elo modifier for different weapons/equipment. So a good player in a good mech with a good loadout will rate higher than a bad player in a good mech with a bad loadout, etc.

There is no good way to get useful data out of a QP environment than law of large numbers. That's it. You either have to go insanely granular (each player had a ton of entries relating both to stat performance in each situation and relative to every other player) or you LLN. Given that we have a limited population and some other significant criteria for the matchmaker (tonnage and always 12 v 12 within a given timeframe) LLN is the only good approach.

How/Why can not be a modifier for a matchmaker because of double-dipping and wrong attribution. Your damage, match score and KDR all factors into your W/L. You can't then double back to use them to augment win/loss because you're stacking their value. Also they may NOT always be a factor in win loss. You can say that 'a high average match score is usually a factor in winning'. That may be. However it is not always.

You know what is always a factor in winning? If you won or not.

Any factor in a matchmaker other than win/loss is literally just dirtying your data source. It's skewing your own numbers.

The beset, most complex matchmaking tool in the world is called TrueSkill. It's exemplary because it is used by Microsoft for XBox to seed your ranking in one game based off your performance in another. So your CoD stats can be used to rank you with Titanfall 2 players, or even checkers. It's a double-value system, it tracks both estimated skill level and confidence to give a more accurate belief rating for the players ranking.

However it's based entirely around w/l. All the other metrics and modifications are used to help translate your win/loss in one game to another.

WHY someone wins or loses and be interesting. HOW certainly has value for understanding what the rankings mean. However to create the matchmaker itself using anything *except* win/loss isn't going to give you an accurate result. Not for pug queue. Group queue, FW, comp queue? Sure, you're affecting the variability of both teams. Pug queue though? The consistent variability of teams is what gives it accuracy.

eyeballs, on 01 November 2017 - 08:39 PM, said:

8v8 will be the nail in the coffin for me. I can contribute in 12v12 cuz there are enough bodies for me not to be vaporized first some of the time. With 4 fewer players, that is more attention on me, and I know I'm not good enough in my current tier, play pool and mech choice to withstand that. An 8 man with me on it might as well be a 7 man for all the good I will do at that point.

So, yeah. 8v8 is when I bow out and stop ******** up your matches. Yay, I guess?

Except there's 4 less people to shoot at you.

We had 8 v 8 before, comp play has 8 v 8. It raises TTK, no question. The odds of you only getting shot by 1 or 2 people is way higher. 12 v 12 dropped TTK like a rock because it significantly increased concentration of firepower.

If it doesn't work for you, I get it. I'd recommend giving it a try though. Statistically what you'll find is fewer total guns on the other side. Sure, your side has fewer too - however in the context of your survival the total firepower arrayed against you dropped 33%.

Edited by MischiefSC, 01 November 2017 - 08:51 PM.

[Gierling]

8v8v8 game mode would be very interesting.

You could call it "Rumble", it would also work with Conquest and Domination.