13 replies to this topic

[CGB Behemoth]

Some facts:
1) Matchmaking system based on player ELO score & mech weight.
2) ELO score is at first place when MM works.
3) Very often there are 1-3 really good players (with high ELO) and 8-11 worse (sometimes REALLY BAD players) players in team.
4) When such team meets 12 middle ELO players (that happens really often) - it is doomed!

So what programmers can do with easy to make MM vore balanced?

Answer is - MAKE ELO GROUPS.

Something like:
1) First group players from 0-500 ELO score
2) Second group 500-1000
3) Third group 1000-1200
4) Fourth group 1200-1400
... and so on.

It is easy enough to make players from first group never play against third group or higher (and only 1-2 players from second group, to make some challenge and learning).

P.S. Sorry for my bad English. Hope text is understandable.

Edited by Behemothk, 17 October 2013 - 07:20 AM.

[Ximix]

Where do you find your ELO?

[Ironwithin]

You don't, it's a figure the matchmaking-system uses to match players of roughly the same experience and it is not publicly released. Only PGI knows your Elo-rating.

[QuantumButler]

Behemothk, on 17 October 2013 - 07:19 AM, said:

Some facts:
1) Matchmaking system based on player ELO score & mech weight.
2) ELO score is at first place when MM works.
3) Very often there are 1-3 really good players (with high ELO) and 8-11 worse (sometimes REALLY BAD players) players in team.
4) When such team meets 12 middle ELO players (that happens really often) - it is doomed!

So what programmers can do with easy to make MM vore balanced?

Answer is - MAKE ELO GROUPS.

Something like:
1) First group players from 0-500 ELO score
2) Second group 500-1000
3) Third group 1000-1200
4) Fourth group 1200-1400
... and so on.

It is easy enough to make players from first group never play against third group or higher (and only 1-2 players from second group, to make some challenge and learning).

P.S. Sorry for my bad English. Hope text is understandable.

That's all well and good, but MWO probably doesn't have enough players to sort into groups and still have people find matches.

Searching for a match takes long enough as it is, they should just bring back hard 1 to 1 weight based matchmaking since ELO just isn't working and nothing they can do will fix it.

[Svidro]

QuantumButler, on 17 October 2013 - 09:45 AM, said:

That's all well and good, but MWO probably doesn't have enough players to sort into groups and still have people find matches.

Searching for a match takes long enough as it is, they should just bring back hard 1 to 1 weight based matchmaking since ELO just isn't working and nothing they can do will fix it.

Pretty much this, though I would disagree that ELO "just isn't working." I just made a new account to play around with trial mechs and wheeeee were the people I was playing against bad. Though it was nice seeing a mix of mechs rather than Cata 3d, cata3d, victor, victor, HGN, jenner

[Svidro]

Actually, having thought about it another few seconds, this might work IF
1. The ranges were dynamic based on the CURRENTLY ACTIVE playerbase. Basically you have a lot of buckets, and the actual ELO min and maxes are adjusted to keep each bucket at least a certain amount full. The high end bucket might have a very wide ELO bracket, but hey, it would be the high end, so if you are there, you deal with it I guess. It would also allow for a bucket that is mostly for people who are in their Cadet bonus phase, and the "worst" players.
2. Should help solve longer queue times if players within a bucket are mostly treated as equal. I'm sure this would result in some unbalanced matches, but we are currently seeing that anyway, right? It would also allow for a re-emphasis on weight matching.
3. WTB weight limits for drops.

[SchwarzerPeter]

As far as i know the MM is working with brackets as you described right now. The problem is that sometimes the matchmaker does really weird thinks to fill up the matches, which results in such poor roflstob combinations.
And they have loosen the weight preference, because high elo players where experiencing very long waiting times.

Oh and sometimes there seem to be not enough players around for a good matchmaking

Last word from the dev's was that they look into it, maybe with longer waiting times, but with more feedback from the matchmaker (not just the spinning circle).

Edited by SchwarzerPeter, 18 October 2013 - 03:31 AM.

[Kageru Ikazuchi]

Behemothk, on 17 October 2013 - 07:19 AM, said:

...
So what programmers can do with easy to make MM vore balanced?

Answer is - MAKE ELO GROUPS.
...

The problem with Elo groups is that some people will try to game the system so that they are usually in the top of their group.

For example, an exploit gamer realizes that he just transitioned up into the "Gold" tier, he decides to tank a couple of dozen games (maybe farming XP or C-Bills running new mechs), until he's back in his "Silver" tier, then he goes on a few ROFL stomps, moves back up to "Gold" ... wash, rinse, repeat.

Also, what if you're grouped with IRL friends, who happen to be of a different skill level ... maybe you're a "Gold", two of your friends are "Bronze" and one "Diamond" ... how does the match maker deal with that in a tiered system? Average? Highest?

I don't think we're ever going to get completely away from having inexperienced pilots in matches with average skilled pilots, or average skilled pilots dropping with top tier pilots ... especially when any four players can group up together.

[drinniol]

Maybe you like waiting ten minutes for a match, I sure don't. Two teams equally matched with Elo and weightclass does not necessarily mean one team won't get stomped. You only need to have one or two stupids far the dominos to topple.

[Kaijin]

Bear in mind MWO's Elo doesn't go on individual player skill, but on match wins/losses, so luck plays a major part. Lucky if your team has less morons than the opposing team. Unlucky if the reverse. Elo go up. Elo go down. It's rather silly.

Edited by Kaijin, 17 October 2013 - 04:46 PM.

[Kageru Ikazuchi]

Kaijin, on 17 October 2013 - 04:46 PM, said:

Bear in mind MWO's Elo doesn't go on individual player skill, but on match wins/losses, so luck plays a major part. Lucky if your team has less morons than the opposing team. Unlucky if the reverse. Elo go up. Elo go down. It's rather silly.

Yes, over a small sample size, the random nature is significant, but over hundreds or thousands of games, your ability to affect the impact of a game with and against players of different skill levels can be measured.

[SchwarzerPeter]

You don't need hundreds of games. If you win 5-6 matches in a row you sure had some value to your team. Vice versa if you lose all of them.
In 10 matches a ELO system can determine your level pretty accurate. Not perfect, but accurate enough.

The system works not with win/loses btw, it just states whether you should win a match or not. If you win a match you should win, you just gain a very small amount of elo points (or even nothing at all, depends on the elo difference). If you win a match you should lose, you gain a huge amount of points. Same for matches you lose: you lose very little (or nothing) in matches you should lose. If you lose a match you should win, you lose a huge load of points.

Thats the reason why a win/lose rate don't represent your elo skill. In long term the upper 10% will win significant more matches than they lose. Opposite for the lower 10%. The rest should be pretty balanced in a win/lose rate.

Thats the elo system. The system works, whether you lose or win a match. But it won't be a nice experience for the player if the Matchmaker screws things up. If you get placed 10 times in matches you should lose, the elo system might be right, but you hate these matches nevertheless.
Its not the elo system who screws things up, its the matchmaker.

Edited by SchwarzerPeter, 18 October 2013 - 03:58 AM.

[1Sascha]

Based strictly on word-of-mouth here, but:

Is it true that new players are automatically classified as "average" by the ELO-system? That would explain quite a lot of weird player behavior/incompetence I've been seeing.

If this is true, I don't understand the reasoning behind it. New players should be ranked as "rookies".. not as "average". Anything else would be like hiring a high-school graduate and sticking him into a middle-management type of position straight away.


Some of the guys I've been watching since launch don't just look like they've never played a Mechwarrior/FPS-game. They look more like they've never played *any* sort of computer-game.

Quote

If you win a match you should lose, you gain a huge amount of points

Really? If that is the case then that's utter non-sense, IMO. Basing your rating on anything else than *your* performance is kinda silly. Meaning that with this system as described by you, you could get a huge increase in your ELO just by getting lucky. Or, and that would be even worse, by simply rushing for cap, thus winning a game by simply standing around.

If anything, it should be personal achievements that help you "rank up". Like total kills, K/D, components destroyed, spotting assists, etc. You know? Stuff that actually *helps* the team?

Simply winning a map doesn't say anything about your own skills. I've had games were I played like an utter id!ot, yet my team won the match. OTOH, I've had matches where I killed 3 or 4 Mechs, amputated a few more and still our team lost. Which of the two should count towards my ELO, I wonder?



S.

Edited by 1Sascha, 18 October 2013 - 04:08 AM.

[Electron Junkie]

PGI's Elo is busted plain and simple and there is now way to fix it with the amount of customization that MWO allows.
Everyone just need to accept that it will never, ever work. PERIOD.

Going to a strict combat rating off of match score would be preferable at this point and don't even try to stretch it out to different Elos for the same player across different weight classes. Yet matchmaker is only the tips of the issues.

Hit detection and HSR makes this game damn near futile to play when you have an awesome ping compared to those that are connecting to overseas (20~30ms vs 200~300+ms). I recieve "ghost" damage where I hear no impact, see no weapons fire, recieve no screen shake or any feedback at all other than a weapon being destroyed through a fully armored section. Nothing quite like running to battle in an Atlas when both 10-X ACs are destroyed and I have NO EXTERNAL DAMAGE.

I get HSR rubberbanded to death more time than I can count. I will see an assault while in just about any light mech. I cut, turn and run and will make it more than 100m around a bend and OUT OF LINE OF SIGHT or through a tunnel but am pulled back that same 100m to the death animation or a leg missing.

It honestly feels like I am playing on Xbox again where players with cruddy connections are give so much lag comp that they can more or less rule the games.

Edited by Electron Junkie, 18 October 2013 - 06:18 AM.