17 replies to this topic

[JC Daxion]

I see lots of new players ask how it works.. Play well you go up... play bad you go down... But one thing that everyone says... the longer it takes, the higher you are...


Thoughts on this?

[Thorqemada]

If the search für Public Matches takes long you are either very good or very bad bcs the MM cant find Players close to your Elo Level and expands the search over the time more and more until it gets a Patchworkteam of Elos together that matches the average Level of your Opponent but not necessarily the competitiveness.

[Gamuray]

Each team is composed in such a way that the two are roughly equal in average ELO. (or calculated skill level) So you can have a few really high ELO's and a bunch of low ELO's, against a bunch of middle ELO's. Which generally means a bad game, but not always.
Your ELO doesn't determine your opponent levels really.. you team's average ELO determines your opponent's average ELO, but not the specific ELO of each player you face... which can vary wildly so long as the average is still somewhat equal.

Your ELO changes depending on whether or not you were facing a team of higher or lower average ELO. Higher, you should have lost. If you did, no ELO change, if you won, ELO increase (for your whole team). Lower ELO enemies? You should win. If you did, not much change. If you lost, ELO decrease.

Additional Note: The match makers also attempts to make your win/loss ratio about 50/50. So if you have won a bunch, and your ELO goes up, it will start giving you (and your teams) tougher opposition, therefore making your win/loss ratio more likely to fall back in line. <-really stinks... decides ahead of time whether your team should be winning...

[Mawai]

The matchmaker does NOT try to make your win/loss ratio 50/50 ... it could care less.

Elo starts off at a particular value near the middle of the distribution.

Matchmaker forms a match by grabbing players in the queue ... it looks for a 3/3/3/3 weight class distribution (modified depending on queue size and wait times). It also looks to grab players for both sides with comparable Elo. The initial "Elo" search value is taken from the "seed" or initial player used to start the match. This player is the one who has been queued the longest. If it can't find enough players in that Elo bracket within a specified amount of time (something like 2 minutes) it starts to relax the Elo requirements until it can fill the match. For this reason, people with very high or very low Elo can end up waiting a longer time for a match since there just aren't that many players in the same Elo bracket. For the majority of players this is not an issue.

The other reason a match can take a while to form is if you drop in a popular mech class ... like heavies. Heavy mechs can be up to 50% of the queue at time but the matchmaker will tend not to put more than 4 (rarely 5) of one weight class into a match. This means that if you drop in a heavy mech when the queue is busy you can easily have a 5 minute wait for a match while dropping in a light at the same time might be typically under 30 seconds.

Anyway, to get back to Elo and the matchmaker.

If your team is predicted to lose and you win, your Elo goes up a lot, If you are predicted to win and you win then your Elo goes up a little. If you are predicted to lose and you lose then your Elo drops a little while if you are predicted to win and you lose your Elo drops more. In addition, how MUCH you are predicted to win or lose by also affects the magnitude of change in your Elo. As a result, your Elo and your win/loss ratio are NOT the same thing. Sometimes a win can result in only a small change in Elo and other times it will be a large one and there is no way for the player to know.

Finally, as your Elo rises or falls to a value that reflects your overall average contribution to how the team wins or loses ... you are placed with other players and against other players with similar team contributions. This means that the teams should be evenly matched. What do you think your chances are of winning or losing an evenly setup match? About 50/50. This has NOTHING to do with the matchmaker ... it has to do with making even and fair matches. However, any even and fair match has a 50/50 chance of either team winning. This means that as long as you do NOT improve ... you play the same ... your Elo stays the same and you are matched against comparable people ... you end up with a win/loss that is somewhere around 1. The matchmaker does NOT care what your results are .. it just happens because you are matched against equally skilled people. If your W/L is greater than 1.0 then it means you are probably getting better ... learning how to play better ... learning to avoid dumb moves ... your Elo is going up. Eventually though, you will find yourself matched against comparably skilled players, your Elo will settle down and your W/L will tend toward 1.0 BECAUSE you have found your correct level of play.

[Khaze]

Mawai, on 01 February 2015 - 08:07 AM, said:

Spoiler

The matchmaker does NOT try to make your win/loss ratio 50/50 ... it could care less.

Elo starts off at a particular value near the middle of the distribution.

Matchmaker forms a match by grabbing players in the queue ... it looks for a 3/3/3/3 weight class distribution (modified depending on queue size and wait times). It also looks to grab players for both sides with comparable Elo. The initial "Elo" search value is taken from the "seed" or initial player used to start the match. This player is the one who has been queued the longest. If it can't find enough players in that Elo bracket within a specified amount of time (something like 2 minutes) it starts to relax the Elo requirements until it can fill the match. For this reason, people with very high or very low Elo can end up waiting a longer time for a match since there just aren't that many players in the same Elo bracket. For the majority of players this is not an issue.

The other reason a match can take a while to form is if you drop in a popular mech class ... like heavies. Heavy mechs can be up to 50% of the queue at time but the matchmaker will tend not to put more than 4 (rarely 5) of one weight class into a match. This means that if you drop in a heavy mech when the queue is busy you can easily have a 5 minute wait for a match while dropping in a light at the same time might be typically under 30 seconds.

Anyway, to get back to Elo and the matchmaker.

If your team is predicted to lose and you win, your Elo goes up a lot, If you are predicted to win and you win then your Elo goes up a little. If you are predicted to lose and you lose then your Elo drops a little while if you are predicted to win and you lose your Elo drops more. In addition, how MUCH you are predicted to win or lose by also affects the magnitude of change in your Elo. As a result, your Elo and your win/loss ratio are NOT the same thing. Sometimes a win can result in only a small change in Elo and other times it will be a large one and there is no way for the player to know.

Finally, as your Elo rises or falls to a value that reflects your overall average contribution to how the team wins or loses ... you are placed with other players and against other players with similar team contributions. This means that the teams should be evenly matched. What do you think your chances are of winning or losing an evenly setup match? About 50/50. This has NOTHING to do with the matchmaker ... it has to do with making even and fair matches. However, any even and fair match has a 50/50 chance of either team winning. This means that as long as you do NOT improve ... you play the same ... your Elo stays the same and you are matched against comparable people ... you end up with a win/loss that is somewhere around 1. The matchmaker does NOT care what your results are .. it just happens because you are matched against equally skilled people. If your W/L is greater than 1.0 then it means you are probably getting better ... learning how to play better ... learning to avoid dumb moves ... your Elo is going up. Eventually though, you will find yourself matched against comparably skilled players, your Elo will settle down and your W/L will tend toward 1.0 BECAUSE you have found your correct level of play.

But in all seriousness, really, who cares how it works? It is in the game's nature for things to snowball very rapidly and it may not have anything to do with the ELO difference of teams. Someone ***** up at some point in the match and your defence crumbles like a deck of cards.

People should try recording some of their games for shits and giggles. Going through the footage at a later time you can usually very quickly identify the point where it all went wrong. But we're all piloting slow as balls battlemechs with very little capability to hurt the opponent if you're facing more than 2:1 odds - you just die too quickly before you've done any significant damage. In contrast you take a twitch FPS shooter where a single teenager juiced up on Red Bull can turn the entire match around by popping 4 headshots in half as many seconds - that clever 'flanking' manouver suddenly stands for nothing when said teenager is teabagging your corpses.

[LordKnightFandragon]

ELO: How it works

Are you good? Here, carry these 11 terribads

Are you bad? Here, carry there 11 terribads

What? you won 4 games in a row? Here, have 10 more losses

LOL, you actually thought you were getting above 50% win rate, silly you.

[Apocryph0n]

LordKnightFandragon, on 01 February 2015 - 03:20 PM, said:

LOL, you actually thought you were getting above 50% win rate, silly you.

But... you can :-S

[LordKnightFandragon]

Apocryph0n, on 02 February 2015 - 01:28 AM, said:

But... you can :-S

Yeah, if you carry the other 11 tards on your side......

Cuz sure as ****, you cant rely on them to atleast pull their own weight in the match. 11 other mechs, out of 100% of the match, they each only need to do like 9% of the weight in the match, and yet they cant be bothered to do that. If you want to have a winning record in this game, or any game like it, you must be a god player who is capable and willing to Carry every match.

Edited by LordKnightFandragon, 02 February 2015 - 01:33 AM.

[Ens]

LordKnightFandragon, on 01 February 2015 - 03:20 PM, said:

ELO: How it works

Are you good? Here, carry these 11 terribads

Are you bad? Here, carry there 11 terribads

What? you won 4 games in a row? Here, have 10 more losses

LOL, you actually thought you were getting above 50% win rate, silly you.

That´s EXACTLY how it works.

[Apocryph0n]

LordKnightFandragon, on 02 February 2015 - 01:32 AM, said:

you must be a god player who is capable and willing to Carry every match.

Willing, yes, everyone should be. Capable? I dunno, about 85% of my matches in the last few months were in Locusts, Mist Lynx and now Novas.

Consensus: The ways of the MatchMaker are mysterious.

[ThisMachineKillsFascists]

If we would have the non elo system back, we would most likely have better matches. Random team vs random team

You are too good, now we put noobs into your team to make you loose bs is beyond annoying.

Sometimes iam like: Lets do it slowly, we have alot members without a faction and without unit. After 2 mins i look at the kill counter and my team is already short one hole lance :C

Edited by ThisMachineKillsFascists, 02 February 2015 - 02:00 AM.

[LordKnightFandragon]

ThisMachineKillsFascists, on 02 February 2015 - 01:58 AM, said:

If we would have the non elo system back, we would most likely have better matches. Random team vs random team

You are too good, now we put noobs into your team to make you loose bs is beyond annoying.

Sometimes iam like: Lets do it slowly, we have alot members without a faction and without unit. After 2 mins i look at the kill counter and my team is already short one hole lance :C

Yeah, its also why I think there should be a 100 game "Cadet Q"......all the new players against the new players. let the players with a clue get in against other players wiht a clue....or you hope after the 100 games they get one....

[ThisMachineKillsFascists]

LordKnightFandragon, on 02 February 2015 - 02:02 AM, said:

Yeah, its also why I think there should be a 100 game "Cadet Q"......all the new players against the new players. let the players with a clue get in against other players wiht a clue....or you hope after the 100 games they get one....

Ideally we would have ranks. Dependend on how your average matchscore is. 3-4 Rank. You will only play within your rank in solo matches. In Teammatches your average Rank of your premade counts.

Not sure how big the population in this game is to have a splitted MM

Edited by ThisMachineKillsFascists, 02 February 2015 - 02:06 AM.

[LordKnightFandragon]

ThisMachineKillsFascists, on 02 February 2015 - 02:05 AM, said:

Ideally we would have ranks. Dependend on how your average matchscore is. 3-4 Rank. You will only play within your rank in solo matches. In Teammatches your average Rank of your premade counts.

Not sure how big the population in this game is to have a splitted MM

With the current rate and pace of this game, we will never have a big playerbase. I dont even think once it goes out on steam it will get that big a playerbase. I mean, it might get a few more, but prolly most will come on, see the grind, see how fast stuff dies, and just be like...I just spent 300 games and I can afford 1 mech.....and now i need 2 more to get this one mastered? WTF?

Then those who want PVE in a serious manner, they will be sorely disappointed. Then there are those who will just want to come on and dakka something in the face with a troll build, but I am pretty sure that will be the playerbase. Die Hards in CW and the troll bads in Pub Q....any of the moderately serious casuals who are not quite die hard and are better then scrub, will be like...meh.....and play rarely or quit.....

But a split Q MM would be grand, only getting your general rank and skill lvl in the MM, would be amazing.

Edited by LordKnightFandragon, 02 February 2015 - 02:15 AM.

[ThisMachineKillsFascists]

LordKnightFandragon, on 02 February 2015 - 02:14 AM, said:

With the current rate and pace of this game, we will never have a big playerbase. I dont even think once it goes out on steam it will get that big a playerbase. I mean, it might get a few more, but prolly most will come on, see the grind, see how fast stuff dies, and just be like...I just spent 300 games and I can afford 1 mech.....and now i need 2 more to get this one mastered? WTF?

Then those who want PVE in a serious manner, they will be sorely disappointed. Then there are those who will just want to come on and dakka something in the face with a troll build, but I am pretty sure that will be the playerbase. Die Hards in CW and the troll bads in Pub Q....any of the moderately serious casuals who are not quite die hard and are better then scrub, will be like...meh.....and play rarely or quit.....

But a split Q MM would be grand, only getting your general rank and skill lvl in the MM, would be amazing.

Agreed. The grind is really terrible.

Edited by ThisMachineKillsFascists, 02 February 2015 - 02:34 AM.

[kapusta11]

It doesn't work the way it is supposed to:
Elo is used for 1vs1 matches, even in football it's 1 TEAM vs 1 TEAM with CONSISTENT team line-up, they are treated as a whole.
Elo has leagues/categories/brackets, MWO mix those because of low population.
All above creates so many convoluted relations that instead of Elo you might as well just assign any random number to evaluate player performance.

Want to avoid PGI's borked Elo? Find 4+ people and play in group queue.

Edited by kapusta11, 02 February 2015 - 02:36 AM.

[Jeffrey Wilder]

LordKnightFandragon, on 01 February 2015 - 03:20 PM, said:

ELO: How it works

Are you good? Here, carry these 11 terribads

Are you bad? Here, carry there 11 terribads

What? you won 4 games in a row? Here, have 10 more losses

LOL, you actually thought you were getting above 50% win rate, silly you.

What amazes me is how the system keeps your ELO level when you lose more than you win MOST of the time.

Edited by Jeffrey Wilder, 02 February 2015 - 04:03 AM.

[Joseph Mallan]

JC Daxion, on 31 January 2015 - 10:48 PM, said:

I see lots of new players ask how it works.. Play well you go up... play bad you go down... But one thing that everyone says... the longer it takes, the higher you are...


Thoughts on this?

Not the better or worse you play. The more your team wins loses.