37 replies to this topic

[Screech]

Maybe the game Is not for you.

[Catra Lanis]

CocoaJin, on 31 January 2015 - 07:56 AM, said:

Devs said that ELO changes are based on the outcome of the match, win/lose. After the teams are made, the MM makes an assumption of the outcome based on the two team's ELOs. If the outcome is different from the presumed outcome, as in the lower ELO team wins, then the ELOs of players change, with the underdogs getting an increase in ELO and the previously assumed winning team players getting a decrease.

If the presumed winning team wins, there is no ELO change for anybody...or at best there is a very small change, but honestly, I'm 99% sure I read that there would be no change.

The MM does the best it can to always create a balanced matchup, but it's already accepted there will be some acceptable deviation in the two team's ELOs...that is, unless there is a short fall of acceptable players in the queue to match, in this case, the MM does the best it can with what it has.

Is that a good thing? Seems to me that it could happen by chance, a one off. I think you should have at least a few "unexpected" victories before it goes up.

[CocoaJin]

Catra Lanis, on 31 January 2015 - 09:49 AM, said:

Is that a good thing? Seems to me that it could happen by chance, a one off. I think you should have at least a few "unexpected" victories before it goes up.

You are right, but the increase is small, that one win is unlikely to significantly change your position much. Plus, you have to realize that the ELO process doesn't expect to accurately determine your appropriate ELO early on or in the short term, it's assumed that your ELO will more accurately portray your appropriate level over time. So even if you have a string of fortunate wins, the ELO benefits of those wins will be short lived as more and more games played after those wins will naturally settle you back toward where you "ought" to be.

In the long run, you'll tend to hover in the vicinity of your appropriate ELO, but the nature of the gaming environment means it happen in a never ending oscillation of over shoots above and below. Occasionally those overshoots might put you in a new "bracket", but I can't imagine it would happen frequently to most of us, unless your appropriate ELO was somewhere in the vicinity of the bracket threshold.

Edited by CocoaJin, 31 January 2015 - 10:59 AM.

[SolCrusher]

Lowering your elo is easy. First if you want to lower your elo buy the smallest mech in each class. So get a locust for lights, get a cicada for mediums, get a dragon for heavies, and get a victor for assaults.

Load up the crappiest load out you can think on each one. Pack the biggest engine in so you can go fast, make sure it's an XL. I'd go with all small pulses on laser spots, all srm2s on missile spots and machine guns on ballistic spots. Basically you are a crap build. Run around "trying hard" and die. Don't go off not supporting your team, make your death "believable" so noone's feelings gets hurt. Rinse repeat have fun croaking.

Oh for get your w/l ratio and k/d ratio too.

Edited by SolCrusher, 31 January 2015 - 12:20 PM.

[Agent 0 Fortune]

CocoaJin, on 31 January 2015 - 12:01 AM, said:

Reference source plz

Ghogiel, on 31 January 2015 - 01:52 AM, said:

It never used to. Sauce?

Joseph Mallan, on 31 January 2015 - 02:40 AM, said:

I agree. This is the first I read that Agent.

Made some errors in original post.... so many errors. I think CocoaJin has it right.
I was mistaking reality with some convoluted system I had been thinking about.

[Felicitatem Parco]

"Elo" is a guy's last name. It says much about your discussion piece if you spell it in all caps...

[Scyther]

I am not a very good player, and this morning I was in a Spider that is new, un-elited, and I am also not very good in.
(So it's not likely my ELO that is a factor here)

The game gave me 8 out of 9 matches where my team killed 4 or less (usually only 2) enemy mechs.

One match, a TKer killed 2 enemy, then got killed by his team (so they were 3 mechs down and at least 1 wounded). My team still only killed 3 mechs total (6-12 final score).

Fortunately, my (literally) tens of thousands of matches in World of Tanks taught me that these things happen, they tend to come in streaks, no biggie.

Take a break, take a walk, change your style, change your mech, play CW or something different, or heck, go read a book.

Edited by MadBadger, 31 January 2015 - 01:03 PM.

[Wintersdark]

CocoaJin, on 31 January 2015 - 07:56 AM, said:

Devs said that ELO changes are based on the outcome of the match, win/lose. After the teams are made, the MM makes an assumption of the outcome based on the two team's ELOs. If the outcome is different from the presumed outcome, as in the lower ELO team wins, then the ELOs of players change, with the underdogs getting an increase in ELO and the previously assumed winning team players getting a decrease.

If the presumed winning team wins, there is no ELO change for anybody...or at best there is a very small change, but honestly, I'm 99% sure I read that there would be no change.

The MM does the best it can to always create a balanced matchup, but it's already accepted there will be some acceptable deviation in the two team's ELOs...that is, unless there is a short fall of acceptable players in the queue to match, in this case, the MM does the best it can with what it has.

This is correct. However, the bold part isn't exactly right - Elo's do change in almost every instance, but by a small amount. Karl discusses this in his epic thread, though I'm not going to find the particular post. Basically, Elo's won't change only when the variance in team Elo ratings is very large - and that's something that only rarely happens, despite what players think.

Basically, average deviation between teams is significantly less than 100pts (ref: One of the linked posts in Karl's Megathread, referenced in the OP). At 100pts, the likelyhood of a win is 64:36; which is still a pretty close match.

---

There's so much misinformation about how Elo and the Matchmaker works in the game, it's unreal. Mostly, it's fueled by players having a couple bad games (in most cases, fewer than they really think - confirmation bias is a *****) and basically assuming the MM is "trying to make them lose" which is of course ludicrous.

See: the classic theory: I won a bunch of games, so the MM wants to make me lose to bring me back to 1:1 WLR. The supposes that the poster is a Special Snowflake, and the MM is building a match all about him. The reality of course is that the MM attempts to match you as evenly as possible all the time, though it doesn't always work. If it hasn't matched you fairly, however, it takes that into account in scoring as CocoaJin said above... but it's a coin toss, basically, which side of the imbalance your on.

Still, Karl also referenced the "within team" variance being 150 pts or less. That's not a huge difference, so even if you are the highest Elo player on the team, the other players are not going to be drooling idiots in comparison.

What people need to understand is atrociously bad performance happens for a number of reasons, often totally unrelated to skill.

Maybe you're trying something new, and it doesn't work. Maybe you're in a new mech, without a good loadout/skills. Maybe you're just unlucky (round a corner and OH HAI WHOLE ENEMY TEAM!). Maybe you were just distracted at a bad moment, or maybe you just missed a really critical shot, or misjudged what you'd need to kill a target before dying yourself.

Maybe the above happened to a couple players at the start of the match, ensuring your team was grossly outnumbered. In that case, the early disparity of numbers will ensure if the opposing team is deathballed and focus fires (either intentionally or just because they encounter your team one by one) it'll just roll over your team and the only people who do any real damage will be the last couple alive, who had the opportunity to fire for a while unopposed while the enemy team crushed their allies.

The above will result in a team with typically 2 lances who "did nothing at all" and one lance with decent scores. This doesn't mean the high scoring lance was better, just that it had the fortune to be on the other side of the map when the enemy team rolled along.

MadBadger, on 31 January 2015 - 01:02 PM, said:

I am not a very good player, and this morning I was in a Spider that is new, un-elited, and I am also not very good in.
(So it's not likely my ELO that is a factor here)

The game gave me 8 out of 9 matches where my team killed 4 or less (usually only 2) enemy mechs.

One match, a TKer killed 2 enemy, then got killed by his team (so they were 3 mechs down and at least 1 wounded). My team still only killed 3 mechs total (6-12 final score).

Fortunately, my (literally) tens of thousands of matches in World of Tanks taught me that these things happen, they tend to come in streaks, no biggie.

Take a break, take a walk, change your style, change your mech, play CW or something different, or heck, go read a book.

They do happen. Even if the MM calculation is a flat 50:50, sometimes the coin comes up tails many times in a row. And losses between very evenly matched teams are often steamrolls - see the various tournaments. One team gets an advantage, then exploits that advantage to crush their opposition.

[Agent 0 Fortune]

Wintersdark, on 31 January 2015 - 01:16 PM, said:

Basically, average deviation between teams is significantly less than 100pts (ref: One of the linked posts in Karl's Megathread, referenced in the OP). At 100pts, the likelyhood of a win is 64:36; which is still a pretty close match.

Still, Karl also referenced the "within team" variance being 150 pts or less. That's not a huge difference, so even if you are the highest Elo player on the team, the other players are not going to be drooling idiots in comparison.

I recall him talking about time in queue and how the MM will grab the oldest person to make a team, and how teams are built one then the other (not simultaneously). So you end up with most of outliers (and variance) on team 1. But the outliers are the statistical annomoly (not actually paying players) so they can be ignored.

As long as you are an average player in the middle of the Elo bell curve statistics are in your favor. If you are in outlier range (high or low), your typical experience will be dramatically different. I also thought that at one time they put in hard Elo boundaries, which means you if you are at the bottom end of a boundary, you will always be facing better palyers, and it will be much harder to drop Elo (because you are always predicted to lose, so no adjustment, another acceptable statistical casualty).

Edited by Agent 0 Fortune, 31 January 2015 - 01:39 PM.

[Wintersdark]

Agent 0 Fortune, on 31 January 2015 - 01:39 PM, said:

I recall him talking about time in queue and how the MM will grab the oldest person to make a team, and how teams are built one then the other (not simultaneously). So you end up with most of outliers (and variance) on team 1. But the outliers are the statistical annomoly (not actually paying players) so they can be ignored.

This depends on player counts.

Worst case scenario(this also applies at the other end, of course): Seed player is 2800 Elo (max). MM builds his team, trying to get players at the same Elo rating. As there are few people at this point, it's going to have to climb back down the ladder. It grabs then the next 11 best players in the queue... because it's goal at this point is to build a team with as small a variance within the team as possible. It then determines the Team Average Elo rating, and builds the opposing team targeting that value. This means it grabs players at that value, or as close to it, as possible.

Thus, team one still has minimal variance within it. Team 2 is going to be lower, because it must be (all the best players are already taken) but is still going to have very good players as a rule. These matches will show the highest delta between the teams, though, because player counts are lowest at the extremes. How big the variance is within the team depends on the number of players at the seed players' Elo. If there were very few high-Elo players, then that's that. This results in a more "above average" team with a couple ringers(but still no window-lickers), against a slightly lower ranked team - basically what you say above.

However, if there are that few high (and by high, I mean the very top) ranked players? Then that match eats them all, and all the matches over the next few minutes is all very well balanced. In short, the match type you present above is extremely rare by nature.

Everyone likes to think they're high Elo. Few people are - very, very few. By the Elo distribution charts they showed back in the day, the number of players at very high and very low Elo ratings are extremely low in comparison to the bulk of players.

Quote

As long as you are an average player in the middle of the Elo bell curve statistics are in your favor. If you are in outlier range (high or low), your typical experience will be dramatically different. I also thought that at one time they put in hard Elo boundaries, which means you if you are at the bottom end of a boundary, you will always be facing better palyers, and it will be much harder to drop Elo (because you are always predicted to lose, so no adjustment, another acceptable statistical casualty).

Yeah, the buckets. I'm unsure if they stuck with the buckets or not, but the thread progress there indicates that they did not.

If you are very high or very low Elo, however (and this is VERY few people) then yeah, the MM can only do so well for you. This is unavoidable, there just aren't many people at that level. But that also means this problem only exists for a tiny group of players, too. How many would be tied to play time - playing at peak time would mean only the very highest ranked players (less than 1%) would have an issue, while that number would grow dramatically at off peak times. When I play at 3AM Pacific time, for example, teams are basically entirely random. Of course, that's by necessity.


The reality though is that under all circumstances, the situation you complain about affects very few players... and I'd warrant that most who THINK it affects them are simply wrong, they're just drawing incorrect conclusions about player Elo from what happened in the match/EOR screen.

The VAST majority of matches are working out really well, Elo-wise.

[Agent 0 Fortune]

Wintersdark, on 31 January 2015 - 02:16 PM, said:

*** good information ***

I go on rants during the week nights (especially Friday), but game play is really good on Saturday. It gives me hope that with sufficient population, all matches can be well balanced.
Unfortunately right now we don't have the population to keep the engine turning on the "off-hours".

[Agent 0 Fortune]

Agent 0 Fortune, on 31 January 2015 - 02:25 PM, said:

I go on rants during the week nights (especially Friday), but game play is really good on Saturday. It gives me hope that with sufficient population, all matches can be well balanced.
Unfortunately right now we don't have the population to keep the engine turning on the "off-hours".

And as if on queue, at 4pm PST Saturday, we are back to our regularly scheduled stomp game.
See you guys next week.

[CocoaJin]

This example is why we need to stop coming to forums and bad mouthing the MM and spreading misinformation and vitriol. All it does is promote and justify the lack of personal responsibility for our contributions to loses. What's worse, it fans the flames of harmful, destructive, contemptuous comments and discontent in the community that eventually finds itself plastered on the various media formats we frequent. This unwarranted war on the MM is only hurting our community and the game...that is all, it produces no good fruit.

So here is what I saw today...River City, Water vs Bridge start, we were water. We lose 6-12. Some yahoo starts barking about the MM fail, I had to remind him of the play by play that led to our lose and how it was out fail.

Our Assaults started near the cargo docks and moved counter-clockwise toward the citadel and through the urban area toward the Dropship platform. Our light and medium/heavy lance quickly move toward the Dropship platform with total disregard for the Assaults(we've seen this a million times), the lights and mediums see what appears to be a lone Atlas and smell blood in the water as he slowly scrambles to get across the bridge to his teammates. All along, our Assaults are trudging through around the citadel and through the cramped urban area trying to get into the reaction, pausing to fire off LRMs and take shots at targets poking out in the urban area across the water. No one pays attention to my call that we are leaving the Assaults behind...no, they want to beat on the Atlas.

Unbeknown to our lights and mediums, there are ECM enemy in the urban area across the bridge, as their bloodlust catapults them headlong into the attack on the Atlas and one other light, none of them recognize the trap they just set off, nor does anyone convey the ECM trap after they die. So what was initially a early 1-0, rapidly turns into a 1-6.

I see the test as if unfolds and yell out they have ECM across the bridge, but it's too late, the team is already in the jaws of the beast and they are just thrashing about in desperation. Me and the Assaults(I'm a habitual Assault escort in my Wuberine) can only watch from the citadel/urban area (prior to the Dropship platform) as our blue Doritos pop.

Needless to say, we got overrun...but how anyone could blame that of the MM is beyond me...but because he has been encouraged to transfer the blame for a lose to the MM he will yell to the highest mountain and the lowest forum how bad the MM is and how fail Paul is and inept PGI is with religious zeal...and all it does it hurt us as a community and keep him from getting better as a player.

Edited by CocoaJin, 31 January 2015 - 05:40 PM.

[Wintersdark]

I wish I could like that post more than once, CocoaJin.

The lack of personal responsibility really annoys me. Sure, its not always a given players fault that they lost, but you can be outplayed by an equally or even lower skilled team. When you blame the MM for simply being outplayed, you downplay the achievement of the opposing team and insult your teammates.

[BellatorMonk]

Mycrus, on 30 January 2015 - 08:11 PM, said:

Use a new userid... problem solved

This just puts your Elo back to the mid average and in Elo hell with all the rest of the nubs..how would that be better?

[Metafox]

One temporary way to get a lower elo is to group up with players who have a much lower elo than you do. Your average team elo will be lower than you're used to, so you handicap the enemy team at the cost of playing with a handicapped team. From there it's a matter of luck. With some teams, you'll have a turkey shoot. With other teams, your teammates will get you killed every time.

[Wintersdark]

BellatorMonk, on 31 January 2015 - 06:24 PM, said:

This just puts your Elo back to the mid average and in Elo hell with all the rest of the nubs..how would that be better?

Mid-Elo isn't "Elo hell" - it's the land where, new players notwithstanding, you're playing against and with equally skilledsuccessful players.

The VAST majority of players are relatively middle ranked. That happens automatically. New User Elo is a fair bit below median, but you'd progress up to middle very quickly.

Middle Elo is where the best match quality is, where you're almost always garaunteed to have players of equal successfulness. You'll still get just as many stomps as everywhere else, because that's an inevitability of this kind of game. But it does mean you've got to take more personal responsibility for wins and losses, as has been discussed in this thread. Because all the players in the match are of very similar Elo rankings.

(Also: If you are indeed a solid "high elo" player, being in the new user to middle ranked Elo range should make you a god amongst mice. If you're not mopping the floor with everyone..... you're not as good as you thought you were.)

[Smith Gibson]

If you want to lower your Elo rating, just buy a few new 'Mechs, and then don't upgrade them to Endo Steel, Ferro Fibrous Armor, Double Heat Sinks, or Artemis. After long enough playing with effectively crippled 'Mechs, your losing streak will lower your Elo rating.

Then again, if Elo rating is based only on the win/loss ratio, you will be lowering the Elo rating of everyone on your team as well.

By the way, you don't have to capitalize the whole word Elo. Just the first letter will do. It's named after it's creator: Arpad Elo. It's not an acronym or anything.
http://en.wikipedia....o_rating_system