47 replies to this topic

[AjerWerklWerkl]

Joseph Mallan, on 10 November 2014 - 06:40 AM, said:

Does that group Elo match up factor attrition into the equation??? Cause THAT is what throws Elo out the window. With every death on the field, Elo changes in this game.

This is the big question, yes! When a game ends, what factors affect the change in ELO per person?

If it's only Win/Loss, then we have a problem... I would think it could be a weighted formula of Match Score (which already includes everything else, like kills, assists, TAG, etc...), Win/Loss, and average opponent ELO.

[Joseph Mallan]

AjerWerklWerkl, on 10 November 2014 - 06:28 AM, said:

Well, just like any other game or sport, you should generally be playing with people of similar (and ideally, slightly higher) skill level... you LEARN by watching videos on YouTube, or spectating the best folks at your level after you die.

Creating wide-disparity-in-ELO matchups is not the solution to learning.

Egomane: thank you for your informative response! That is exactly the point I was trying to bring to light when this thread started. I hope everyone reads what you said!!!

I always learned best by doing. And that meant doing my best against the hardest toughest job/fighter/team I could find.

[AjerWerklWerkl]

Joseph Mallan, on 10 November 2014 - 07:23 AM, said:

I always learned best by doing. And that meant doing my best against the hardest toughest job/fighter/team I could find.

BTW, that is one of the best features of ELO in chess. You go on a server and you can see the range of ELOs available. You can then challenge whatever level you want... the larger the rating gap, the more points are gained/lost (capped at 32, so you can't have a giant swing all at once). It's a great system. If you can find someone strong who is willing to play you, you can do it.

In MWO, it's a little trickier because it's team based. Maybe the analog is making friends and making a private match so you can learn?

[Egomane]

Joseph Mallan, on 10 November 2014 - 06:40 AM, said:

Does that group Elo match up factor attrition into the equation??? Cause THAT is what throws Elo out the window. With every death on the field, Elo changes in this game.

Why should it? Elo is only applied during match making and for the end result adjustment. Are you attempting to move the goal post?

Both teams were near equal at the beginning. Both had near equal chances to win. There should be no further responsibility for the Elo formula.

Edit:
Your argument brought back to chess...
As you lose towers and knights, how does the chess Elo factor that in? It's totally throwing the result of, if one player loses more pawns then the other.

Edited by Egomane, 10 November 2014 - 10:51 AM.

[Abivard]

Egomane, on 10 November 2014 - 04:58 AM, said:

Elo was initially designed for chess. Since then Elo has been adjusted for many, many tasks, including group matchups.

There are various Elo formulas, adjusted for the tasks they are used for. It is not one formula to fit them all.

Repeating that this system should not work, because it was initially invented for chess, just shows, that you do not understand the concept of it at all. Ballpens where invented for space missions, they still work in a lot of different environments and for multiple tasks.



They did not change it. This was always the case, but you still misunderstand the implemented system completly.

If a group drops, in the group queue, the average Elo of the group is used, to match this group with other groups of similar Elo values. In the single player queue, every player is his own group and the matchmaker will try to match him with players with similar values.

The matchmaker will, for two minutes, try to create a match with all playing instances (24 players in the single player queue or a varying number in the group queue) having a +/- 50 Elo difference between each other. Only if he doesn't manage to do that, will he open some release vents and increase that range. 100 after two minutes and 1200 after five.

It's not: We drop twelve random single players in a group, average the Elo and then match it against another group of twelve random solos where we did the same.

That's how Karl Berg, the one who wrote the matchmaker, described how it works.

So true, after about 50 trillion games of MWO you may approach a correct Elo rating for MWO.

Once again, Elo is not magic, it can only account for so many variables before it becomes useless, the number of variables isn't set, but it certainly can not handle the Humongous amount of variables present in a MWO match.

Your post demonstrates you have no understanding of the difference between MM and Elo ratings.

It really is moot How the MM works, what criteria the MM uses to expand its search, how the MM averages Elo etc....

It is about trying to use an Elo system to rate players in MWO.

Elo was designed for chess, All successful adaptations of Elo outside of the field of chess also have few variables present.


In chess there is only one variable, the player.

In MWO there is only one constant, the player.

Now why would anyone think it would still work... magic maybe?

[Abivard]

Egomane, on 10 November 2014 - 08:06 AM, said:

Why should it? Elo is only applied during match making and for the end result adjustment. Are you attempting to move the goal post?

Both teams were near equal at the beginning. Both had near equal chances to win. There should be no further responsibility for the Elo formula.

Edit:
Your argument brought back to chess...
As you lose towers and knights, how does the chess Elo factor that in? It's totally throwing the result of, if one player loses more pawns then the other.

I take it you do not play chess?

Each player starts with the exact same pieces, in the same relative positions as the other player.

There is no," well one player has three rooks and the other has 2 queens so they are NEAR equal at the start"

Also, the chess pieces neither think nor move by themselves, they are controlled by a single player, there are also no dice rolls involved.

[Sandpit]

Player A has 100 Elo and a medium mech

Ok, that's where the MM system starts. Then we add in
Game mode selection (which can be narrowed to 1)
and
3/3/3/3 (horrible idea in my opinion but I digress)

Now we have 4 basic dynamics that the MM starts with. It must find other players in that Elo range, then players with the same game mode selected, then fill out 3/3/3/3.

When you hit launch the MM starts working its maths for this. If it is unable to find a match in xx seconds, it expands the search. Now it can't expand for game modes many times because of player selections. It can't break 3/3/3/3, so the only way it can expand and find opponents and teammates is to expand the Elo ranges. As it hits the timed "release valves" it continues expanding on eligible Elo ranges.

Groups work differently. Groups have their Elo averaged and then artificially boosted (which again, makes no sense being all groups only face off against other groups) and the process begins just like solo drops.

One tip that will help prevent you (and others) from getting "too far" out of their Elo range is if the MM takes more than about 45 seconds (knowing the actual time for the release valves would help), hit cancel and launch again. This will prevent the MM from expanding too far out of your individual Elo range.

[Sandpit]

Egomane, on 10 November 2014 - 08:06 AM, said:

Why should it? Elo is only applied during match making and for the end result adjustment. Are you attempting to move the goal post?

Both teams were near equal at the beginning. Both had near equal chances to win. There should be no further responsibility for the Elo formula.

Edit:
Your argument brought back to chess...
As you lose towers and knights, how does the chess Elo factor that in? It's totally throwing the result of, if one player loses more pawns then the other.

I think what he's getting at is that in chess you ahve the exact same pieces on both sides with no variation. In a game like MWO with all the customization options, my Stalker isn't equal to your Stalker. Also with weight class used that means according to Elo, that your Atlas is equal to my Awesome.

This also doesn't take into account things like quirks, modules, etc.

Elo doesn't work well here because it's the sole unit used to determine "balanced" pieces. Individual mechs need to have their own metric based on loadouts, modules, upgrades, etc. or at the very least these need to be accounted for in the Elo rating. That's where you tend to get major discrepancies.

Elo isn't bad, it's just not "good" when it's the sole determining factor in balancing your pieces.

[Egomane]

Sandpit, on 10 November 2014 - 03:24 PM, said:

I think what he's getting at is that in chess you ahve the exact same pieces on both sides with no variation. In a game like MWO with all the customization options, my Stalker isn't equal to your Stalker. Also with weight class used that means according to Elo, that your Atlas is equal to my Awesome.

This also doesn't take into account things like quirks, modules, etc.

It doesn't have to account for those things. If you have the same Elo as I ,in a differently equiped Assault, you are using your Assault build equally successfull as I am using mine.

It doesn't matter what loadout we have. In chess there are some players who can mount a fantastic offense with their pawns, while others rely mostly on their knights and again others prefer their queen to do all the work. If they all are equally successfull, they have an equal Elo value and will be matched against each other. Their preference has no place in Elo. Neither does it have one in MWO.

If you change your mech or loadout, it is equivalent to changing your playstyle and your Elo will adjust accordingly, depending if you are more or less successfull with it, then with your previous way of playing.

Sandpit, on 10 November 2014 - 03:24 PM, said:

Elo doesn't work well here because it's the sole unit used to determine "balanced" pieces. Individual mechs need to have their own metric based on loadouts, modules, upgrades, etc. or at the very least these need to be accounted for in the Elo rating. That's where you tend to get major discrepancies.

Elo isn't bad, it's just not "good" when it's the sole determining factor in balancing your pieces.

It's not implemented to balance the pieces, but the players using them. It's not the mech that is rated. It's the player who is using it. Just like it is the player who is using his queen, rook, knight or pawn.

You are looking at it from the wrong end!

Abivard, on 10 November 2014 - 03:01 PM, said:

In chess there is only one variable, the player.

In MWO there is only one constant, the player.

See my answer towards Sandpit, for why you are wrong on this and in the rest of your post.

Abivard, on 10 November 2014 - 03:08 PM, said:

I take it you do not play chess?

I do play it! I may not be the best of the best and I would never dare claim to be, but I'm not someone who is beaten easily either.

Edited by Egomane, 10 November 2014 - 03:51 PM.

[Troutmonkey]

I played for 2 hours last night and only scored 1 or 2 wins. In nearly every single game my team was completely wiped by the opponent in a matter of minutes. The day before the same thing happened. Loss after loss after loss. Trying to coordinate the PUG was fruitless, and my team scattered to the hills over and over. I tried different mechs, different game modes, different play styles - all failed. Two weeks ago I was doing fine, but now my team seems cursed to lose and it's incredibly frustrating, and then it's just depressing. The terrible payouts under the new system just make matters worse.

It appears that I am currently in ELO hell. How the heck do I get out of this mess?

[Egomane]

Troutmonkey, on 10 November 2014 - 03:48 PM, said:

I played for 2 hours last night and only scored 1 or 2 wins. In nearly every single game my team was completely wiped by the opponent in a matter of minutes. The day before the same thing happened. Loss after loss after loss. Trying to coordinate the PUG was fruitless, and my team scattered to the hills over and over. I tried different mechs, different game modes, different play styles - all failed. Two weeks ago I was doing fine, but now my team seems cursed to lose and it's incredibly frustrating, and then it's just depressing. The terrible payouts under the new system just make matters worse.

It appears that I am currently in ELO hell. How the heck do I get out of this mess?

As I was in three of those lost matches, I can tell you that you died early in all of them. Among the first in your or even both teams each time. So maybe you were a part of the reason your team lost?

Your attempts to coordinate where not among the most constructive ones either. If you fill them with to much colorful language and complaints you will be ignored more often then followed.

It's easy to blame others or some obscure matchmaker, but I'd always look at my own current performance first, before I start doing so. Maybe today your reflexes were not up to the task. Maybe your mind was busy with something else. Or maybe you were simply outplayed by the other side.

From what I saw: You created your hell yourself!

Edited by Egomane, 10 November 2014 - 03:59 PM.

[Sandpit]

Egomane, on 10 November 2014 - 03:33 PM, said:

It doesn't have to account for those things. If you have the same Elo as I ,in a differently equiped Assault, you are using your Assault build equally successfull as I am using mine.

That's where the discrepancies come in though. All mechs are not created equal. If you take a stock loadout you're nowhere near as effective as I am with a totally customized loadout. If they implemented a metric that helped bring that into the equation it would help Elo become a better balance point. No single metric is ever going to be enough to balance matches in a game like this with all of the customization and variety offered. Elo simply cannot and was never intended to account for those types of metrics and differences in opponents. It simply puts a value on the player's individual skill level. It doesn't account for things like poor design and build because it can't. That's where adding in something to account for that would help balance matches further than Elo ever could on its own.

Egomane, on 10 November 2014 - 03:33 PM, said:

It's not implemented to balance the pieces, but the players using them. It's not the mech that is rated. It's the player who is using it. Just like it is the player who is using his queen, rook, knight or pawn.

That's the point though. It doesn't rate the equipment which helps lead to some of the discrepancies we see in the game. Now with that said, it's less of a factor at higher Elo ratings. The more skilled the individual player, the easier it is to overcome any imbalances created through the pieces themselves. At the lower levels of the Elo bracket it leads to much harsher results. A new(ish) player isn't going to be able to cope with and adjust to those discrepancies as easily as someone who's individually good at the game based on their skill.

Egomane, on 10 November 2014 - 03:33 PM, said:

See my answer towards Sandpit, for why you are wrong on this and in the rest of your post.

See MY answer as to why you're missing the point and your premise isn't wrong, it's just dismissing that there are real discrepancies in mechs, weights, builds, etc.

Elo is a good place to START, but using it as the sole method of balance isn't enough.

[Abivard]

Egomane, on 10 November 2014 - 03:33 PM, said:

It doesn't have to account for those things. If you have the same Elo as I ,in a differently equiped Assault, you are using your Assault build equally successfull as I am using mine.

It doesn't matter what loadout we have. In chess there are some players who can mount a fantastic offense with their pawns, while others rely mostly on their knights and again others prefer their queen to do all the work. If they all are equally successfull, they have an equal Elo value and will be matched against each other. Their preference has no place in Elo. Neither does it have one in MWO.

If you change your mech or loadout, it is equivalent to changing your playstyle and your Elo will adjust accordingly, depending if you are more or less successfull with it, then with your previous way of playing.


It's not implemented to balance the pieces, but the players using them. It's not the mech that is rated. It's the player who is using it. Just like it is the player who is using his queen, rook, knight or pawn.

You are looking at it from the wrong end!


See my answer towards Sandpit, for why you are wrong on this and in the rest of your post.


I do play it! I may not be the best of the best and I would never dare claim to be, but I'm not someone who is beaten easily either.

You are making so many false assumptions it isn't funny.

The two BIGGEST false assumptions are that MWO Elo ratings are a true reflection of players skill and that Match Maker Always functions the way it was intended.

Garbage in garbage out.

[terrycloth]

Ideally it would track your ELO separately based on your mech and loadout, but that would require dozens of games with each mech before the result would be sensible. I guess it could start with your average for that weight class? Then each game would change the specific mech a lot and the weight class rating a little.

Or check what weapons you were using? "He sucks using LRMs, so effective -50 ELO in this mech."

Or easy one, ELO penalty for mechs with no XP or that aren't elited.

It could also weight the ELO change based on your personal performance. "His team lost but he got 6 kills, probably shouldn't ding him that much."

Not sure all this complication would really be worth it though. KISS is always a consideration.

[Sandpit]

terrycloth, on 10 November 2014 - 05:28 PM, said:

Ideally it would track your ELO separately based on your mech and loadout, but that would require dozens of games with each mech before the result would be sensible. I guess it could start with your average for that weight class? Then each game would change the specific mech a lot and the weight class rating a little.

Or check what weapons you were using? "He sucks using LRMs, so effective -50 ELO in this mech."

Or easy one, ELO penalty for mechs with no XP or that aren't elited.

It could also weight the ELO change based on your personal performance. "His team lost but he got 6 kills, probably shouldn't ding him that much."

Not sure all this complication would really be worth it though. KISS is always a consideration.

Not if a rating was assigned to each component. Think of it as each component having an "Elo modifier" that adjusts accordingly and adds into your personal Elo. Then it doesn't have to adjust for each build in a match, it adjusts based on the component itself and just acts as a modifier to your base Elo.

[F4T 4L]

Egomane, on 10 November 2014 - 03:58 PM, said:

As I was in three of those lost matches, I can tell you that you died early in all of them. Among the first in your or even both teams each time. So maybe you were a part of the reason your team lost?

Your attempts to coordinate where not among the most constructive ones either. If you fill them with to much colorful language and complaints you will be ignored more often then followed.

It's easy to blame others or some obscure matchmaker, but I'd always look at my own current performance first, before I start doing so. Maybe today your reflexes were not up to the task. Maybe your mind was busy with something else. Or maybe you were simply outplayed by the other side.

From what I saw: You created your hell yourself!

Very amusing, but perilously close to name and shame, no?

[Targetloc]

AjerWerklWerkl, on 10 November 2014 - 07:19 AM, said:

This is the big question, yes! When a game ends, what factors affect the change in ELO per person?

If it's only Win/Loss, then we have a problem... I would think it could be a weighted formula of Match Score (which already includes everything else, like kills, assists, TAG, etc...), Win/Loss, and average opponent ELO.

The changes in Elo are affected by win/loss and average opponent Elo.

Kills, assists, etc aren't counted because they're noise. Elo isn't trying to rank your personal skill, it only tries to predict how likely you are to affect your side winning.

Some players are very good, but take big risks by trying to flank on their own or using splat builds or whatever. Sometimes they get really amazing scores, but they might not be consistently helping their team in a way that wins the match. (Looking at 2 tournaments ago, some players get high scores by waiting until the end and trying to clean up).

A lot of players in the middle ground are very inconsistent. They'll have a really great match and think, "man, why am I getting paired up with these losers?" and then they will have a really crap match and think, "man, I wouldn't have died so fast if my team wasn't a bunch of scrubs." They only focus on their good matches, but across multiple multiple games, their performance is pretty average, so their Elo settles in the middle... along with all the other players that have a 50/50 chance of doing either 600 damage, or 60 damage. Grab 6 or 8 pilots each with a 50/50 chance of doing either 600 or 60 damage and... well, you get the quality of matches people complain about.

The way to climb out of that is to be consistently better than average.

[Troutmonkey]

Egomane, on 10 November 2014 - 03:58 PM, said:

From what I saw: You created your hell yourself!

Yes, I am 100% solely responsible for the inept actions of my entire team, and they are in no way responsible or to blame for my loss. I should learn to carry harder, as only my actions affect the outcome of a game

[Vaderman]

Not to mention folks that group often with good players might end up with a highly inflated ELO, and not be as good as they think. That probably leads to a lot of the imbalance as well.

Maybe having a team ELO and a solo ELO per player would even out the playing field more? I dunno.

[Egomane]

Sandpit, on 10 November 2014 - 05:10 PM, said:

That's where the discrepancies come in though. All mechs are not created equal. If you take a stock loadout you're nowhere near as effective as I am with a totally customized loadout.

If we have the same Elo rating, then yes, we are considered equally effective. Otherwise we wouldn't have achieved the same Elo rating.

Again... changing your loadout or mech is changing your playstyle, just like it is in chess. And your success with the changed playstyle will be reflected in your Elo after a short while, by either raising or falling to your new level you should be having.

As I understand it, you want a system for equipment rating, but equipment is not a perfect matchmaking balancing tool either. It brings its own shipload of cans of worms with it.

And if you combine the two systems the matchmaker will get unneccessarily complex and I will guarantee you, the matches will still as often be loopsided as they are now.

Elo is a simple system and it has been demanded often enough for PGI to keep it simple. Making it more complex will not neccessarily make it more effective and it will burn a lot of working hours and money for diminishing returns.

Abivard, on 10 November 2014 - 05:17 PM, said:

You are making so many false assumptions it isn't funny.

The two BIGGEST false assumptions are that MWO Elo ratings are a true reflection of players skill and that Match Maker Always functions the way it was intended.

Educate me! Throwing big comments into a discussion without facts is not really a discussion and will not help you sway me into your direction.

I didn't mention the word "skill" a single time. I didn't say that Elo reflects skill. It doesn't do that in chess either. It reflects success rates. That's all it does and it's all it should do. Don't try to twist words here. If you believe it should reflect skill, then it's you who doesn't understand Elo.

F4T 4L, on 10 November 2014 - 05:35 PM, said:

Very amusing, but perilously close to name and shame, no?

Close, but he himself came here and provided his point of view how his team always was bad. I simply added mine and included him in the picture he was painting. Of course... if you beleive I did name and shame him, please report me. I don't want a special treatment. If I step over the line, I need to be made aware of it.

Troutmonkey, on 10 November 2014 - 07:46 PM, said:

I should learn to carry harder, as only my actions affect the outcome of a game

You did not carry in those matches, not for a single moment, as far as I witnessed it. You complained (even before anything happened) and died early and that was all there was to your performance, on the three matches where I had the honor to witness your greatness.

Vaderman, on 10 November 2014 - 10:13 PM, said:

Not to mention folks that group often with good players might end up with a highly inflated ELO, and not be as good as they think. That probably leads to a lot of the imbalance as well.

Maybe having a team ELO and a solo ELO per player would even out the playing field more? I dunno.

In my eyes that would not be the worst addition, as it could also be the other way round.

Edited by Egomane, 11 November 2014 - 06:03 AM.