50 replies to this topic

[Dracol]

I've noticed a trend in PuG games a lot related to high damage dealers. They get focused on getting a kill and leaving their team mates to more prominent threats.


It has become fairly easy to get 4 mechs on one team chasing 1 or 2 friendly mechs. The left over enemy gets crushed by the 10/11 to 8 odds left behind by those looking to get damage/kills.

[Fastwind]

WarHippy, on 15 July 2014 - 05:37 AM, said:

Perhaps you should have been more concerned with the map objectives instead of pewpewing?

Right as a heavy mech
L2P

[Fastwind]

Zerberus, on 15 July 2014 - 03:45 AM, said:

OP: You do understand that "Victory" or "Defeat" only appear at the end of hte game, right? No need to qualify every single screenie as "at the end" when it`s that obvious with huge white text over the top of the screenie, wastes your time and annoys people that are actually capable of applying logic.


This essentially nails it.. it`s a whine about constantly losing and having bad teammates, while completely disregarding teh only common factor in all of those matches as a possible issue, while displaying a complete and total ignorance of how the matchmaker actually works. Ironically , the screenshots even help to pinpoint the OP`s performance as a possible issue: in all of the losses, his personal teamplay appears to be lacking when taken at the same face value that he expects everyone here to take his statements at. Very few, all the way down to zero assists on the losses, and very many on the victories.

Taken at face value, that says that in the losses the OP either went off chasing squirrels or chose personal targets instead of focusing fire, and actually played with his team properly on the wins.

So how exactly is it NOT your fault, OP?

TL,DR: It`s just another uninformed whine by someone that thinks he should have 12-kill victories handed to him on a golden platter.

I mean, come on, "ONE GOOD MATCH"?? In all 3 victories out of the seven matches you posted you easily made well in excess of 150k c-bills and 1,5k XP even without premium or a hero mech, so what exactly are you whining about?

You rly haven't looked at ALL the screen shots do you!?!
He did work!
He did his part!
He is 1 (ONE) of 12,how much more should he have to carry?

You need to look on the diffrence between how many on his team didn't break 100/200 dmg and how many on the enemy team didn't break 100/200 dmg,that tells you there is an ELO fail on the matchmaker side.

Oh and you can only get assists if your team actually kills stuff.

Edited by Fastwind, 15 July 2014 - 10:11 AM.

[WarHippy]

Fastwind, on 15 July 2014 - 09:57 AM, said:

Right as a heavy mech
L2P

Are you implying that the Timber Wolf is somehow too slow to move around the map? If that is your assertion that would be silly. Of course even an Atlas can pay attention and help with map objectives while pewpewing, but he was complaining that he lost when it was clear all he was doing was looking for his next target; objectives be damned.

[Koniks]

I'll leave these here. Post-patch solo queue.





I didn't grab a screenshot of the one with Lincoln getting teamed with scrubs.

[Naduk]

the quality of the match maker has NO bearing on the chances of a 12-0 game
a match maker can only assess a few key things and all of those things are numbers

numbers cannot account for the human element
elo cannot account for that time you coughed and accidentally fired your ppc's off into the distance and killed a randomly passing enemy via head shot

it cannot account for the time those three enemies just stopped dead in the middle of a massive fire fight because their dinner was ready or the cat unplugged their modem and you scored your self 3 fat juicy kills

it cannot account for the time mr pro player was sick and could hardly see straight for a week and is now back and in top from smashing anyone in his path

its not the match makers fault when half the team that should of won according to ELO chases a jenner gets killed

it cannot account for a squad having a great elo but only if player X is alive

as i said the human element
but most importantly it cannot account for scaling attrition and HUMANS bad decisions

[Lincoln Kotare]

That's the problem is it not? Numbers do not lie its all in how you view the numbers. If you used the match score of the individual player to adjust ELO rank then the match maker would be able to put quality match's together. When you base it off a win/lose players can be carried through the match and still get a higher ELO Rank because of that. I'm not saying to use the current match score because its base off more damage then anything else but if it was reworked then you could. So if you use a game mode like conquest capping a point would give lets say 10 point in a players match score and Kills/damage/assist/spot assist/attacking/defending points would be a % of a match score. Then I think that would give a pretty good idea of how good a player is.

[Naduk]

an elo system that factors contribution to a win instead of just win/loss might be very nice
however 12-0 games would still occur
they would also still occur at the same rate

with the introduction of 3x4 the quality of games increased beyond measure compared to any of past tests/tweaks to elo

that said none of it really effects the 12-0 situation to a large degree other than making it a nicer experience

the biggest factor effecting match quality is players lack of understanding the consequences of their actions and how that fits into scaling attrition

the easiest place to find this lack of understanding is the player who
does 1000+ damage gets 4+ kills and cries because he loses most of his matches anyway
"i cant carry these scrubs he will cry"
it will make him very upset because he believes that he is an outstanding player and there is something wrong with
the game because he does so well but loses anyway

the truth is this player is so concerned with getting high damage and kills he plays to the detriment of his team
i have seen this countless times
instead of changing targets to those of higher threat values as they appear he peruses his kills and personal glory relentlessly
this player will keep shooting the gimped atlas/direwolf until it is dead instead of turning his superior gunnery towards a new target that is a real threat
what this does is give him nice kills and huge damage at the end of a match but while he was getting this the enemy team just took apart his team and laughed the whole time as a mech with no weapons tanked most of their team

hardest thing for these players is they think they are good because of their stats
and because they think they are good or doing it right they do not look for what they are doing wrong and blame something else instead (elo,team mates,matchmaker,lag)
its just like bronze players in league or starcraft
your not in bronze because you got cheated by elo
your in bronze because your bad learn to live with it

as i said before, scaling attrition and peoples lack of understanding

[Koniks]

Elo does its job. Wins and losses are the appropriate metrics to use for it because they are the most strongly correlated with a player's probability of winning a match. This isn't the win statistic for pitchers in baseball where it's subject to arbitrary restrictions. If a team wins a game, every player gets a win added. If a team loses a game, every player gets a loss added. The effect on a player's Elo rating is determined by how likely they were to win or lose that given match based on their team's Elo compared to the opponent's Elo. And Elo algorithms are good at using those to assess win probability.

The issue is with the sensitivity of the matchmaker. It only accounts for a player's Elo and mech class. It doesn't differentiate between strong and weak mechs in a class. It doesn't account for loadout or efficiencies unlocked. It also doesn't account for the delta between the highest and lowest Elo on a team, only the delta between each team's average Elo. And because players and PGI's partners don't want 2 minute searches, it releases the valves on both Elo and mech class very quickly.

Edited by Mizeur, 17 July 2014 - 09:43 AM.

[Void Angel]

Any attempt to modify Elo (which is not an acronym) to reflect additional factors is either:

• No longer an Elo system
  
• More complex and prone to bugs and gaming the system
  
• Produces more unintended or undesired consequences.
  
• Any or all of these things.

For example, currently, MWO's Elo system uses the average of your team's Elo ratings v. the enemy team's average to determine the win/loss probability. When the match is over, Elo changes (or not) based on the probability of the result. If you were to add or change a criterion, this gets trickier.

For example, let's say that the matchmaker instead compares your personal Elo to the enemy team's average. Hey, now higher-Elo teammates gain less from a win and lose more from a loss, while lower-ranked teammates benefit more and lose less - despite the fact that higher-ranked players are going to tend to contribute more. Wait, why not apply a modifier to your personal Elo change based on the ratio of your Elo to the average - so if you were 20% higher than your team's average Elo, you'd gain or lose 20% more ranking? Well, that seems more fair, but it could screw up the Elo curve and catapult people around the rankings like ping-pong balls. Say you're 20% above-average for your team. If you were expected to win, you don't get much Elo for a win, but you get hit really hard for a loss; if the teams are relatively even, you've got the same potential win/loss, but fasterI; and then if you were expected to lose, you're going to go soaring into the wild blue Elo yonder. This causes two problems: it'll be hard to balance how fast the Elo system should adjust players; and it will cause players who experience streaks of wins/losses to be slung up and down the rankings at an accelerated rate - encountering teams of radically varying average skill and making self-evaluation and improvement difficult.

Well, okay. That was awkward, but hey, why not use that higher (or lower) performance itself to affect rankings? You'd avoid the problems from the first paragraph, and we've already got a performance-based reward system in the experience bonuses for kills, assists, spotting, etc! Well, that's true in a way, but it opens up even more worms than monkeying with the Elo system. First, the existing reward system has a recognized weakness in that it rewards firepower far more than role warfare, supporting play, and scouting. Now, you can get a lot of damage and kills in a match with any chassis - the solo tournament scores demonstrate that - but you cannot get as much consistent damage and kills with a Light as you can with a Heavy or Assault chassis. The difference in firepower is too great. So using the exp system would reward the heavier chassis and slow your Light and Medium Elo progression, causing mismatches in the relative "buying power" of Elo between the lighter and heavier weight classes - almost exactly like spending your nation's currency in another country. You also have the problem not only of whether your reward system favors one play style over others, but the weightier problem of how would you know? Then there's the question of whether the Exp rewards are really balanced against each other in relation to how they reflect skill, whether all measurable aspects of "skill" are accounted for in the system, etc. Then there are the immeasurable aspects of skill - leadership, tactical acumen, the ability to take one for the team and make it count. While the effects of some immeasurables can be extrapolated from kills, assists, spotting, and so forth, others can only be inferred from the player's track record of results over time...

Which brings us back to Elo.

Elo isn't a perfectly inclusive system; it cannot and does not track "skill." What it does do is allow relative levels of skill to be extrapolated from results - which is really what skill estimates should be based on. Elo does what it says it does within the limits of its assumptions - this cannot be rationally disputed - and there is no available system shown to do it better. Hypothetical ranking systems are just that - unproven hypotheses that have not been shown to work. Elo isn't a "bad" system because of its simplifying assumptions; rather, it works because of them. The systems I've seen proposed to account for measurable performance in answer to this "problem" fail their own test: the assumptions are still there, in the assertion that this or that set of individual stats can be correllated more closely with skill than win/loss results.

In the end, this is a problem of systems engineering - and engineering is the science of trade-offs. Elo trades detailed analysis of stats for freedom from subjective valuation of those stats, and it is right to do so.

[Void Angel]

Naduk, on 16 July 2014 - 03:38 PM, said:

The easiest place to find this lack of understanding is the player who does 1000+ damage, gets 4+ kills, and cries because he loses most of his matches anyway. "I cant carry these scrubs," he will cry! It will make him very upset, because he believes that he is an outstanding player and that there is something wrong with the game since he does so well, but loses anyway.
[grammar and composition edited by quoter, because pedantry!]

I've seen fairly accomplished players fall into this same trap. It's bias confirmation, and no one is immune, myself included. The pernicious thing is that by definition you tend not to notice bias confirmation. Players who operate well as part of a premade team, for example, will sometimes respond to the solo drop environment by trying to "use the PuGs" in various ways - such as you described elsewhere in the post I quoted. In reality, their focus on hanging back and protecting themselves often means that their team is fighting at a disadvantage in local superiority. Once the fight is done, they pick up kills and extra damage mopping up wounded enemies, but the battle was lost long before they achieved their 1337 score.