137 replies to this topic

[Mudhutwarrior]

Triordinant, on 14 January 2014 - 10:14 AM, said:

Yet another reason to have a premade-free, PUG-only queue. A much better experience, especially for new players.

I know two Mariks who will probably try to have you drummed out for saying that.

[Doctor Proctor]

Don't know if it was mentioned previously, but PGI talked awhile back about how they were changing the Elo rating system up a bit. Previously, new players started at the low end of the curve and had to work their way up the player distribution looked more like a slow leading up from the low end to a larger player population towards the high end. This meant that new players and veterans didn't really play together, so PGI decided to change this.

Now, when a new player rolls an account he's assigned a starting "cadet" (I can't remember the proper name PGI gave it) Elo score that places him near the middle of the distribution curve, in the midst of the largest population of players. This includes many veterans that weren't quite top players that were shifted down towards the center of the bell curve. This gets new and old players to mix together when they're first starting out, in the hopes that vets will help to teach the newbies.

Once they're out of their cadet bonus period the new player will shift down their actual Elo rating, which will probably be a bit lower, but still within the middle 50% of the bell curve. Sounds great, right?

Unfortunately, what this means is that you have players with a couple thousand matches under their belt and a stable of dozens of tweaked out mechs up against first time players in trial mechs that don't even understand the controls of the game. Since the quality of the players at this middle level is so highly variable who you get matched with is largely the luck of the draw, thus leading to some of the ridiculously swingy matches we get. This means it's hard for an individual to push his Elo score up very much since in the next match he could get a squad of newbies and tank the match. So, most players now just sit in the central 50% and don't seem to move around much, forever doomed to hear those magic words: "How do I target again?"

Edit: http://mwomercs.com/...ted-april-19th/
Found the post about the change.

Edited by Doctor Proctor, 14 January 2014 - 10:25 AM.

[Joseph Mallan]

Bhael Fire, on 14 January 2014 - 07:29 AM, said:

When I said premade, I was referring to any premade group....not just the 12-man groups.

Whenever you drop in a group, it makes it harder for Matchmaker to place your group...so you're bound to see other familiar players that are also in groups, mixed in with "filler" solo players of various Elo scores. This is why you see weird mixed up and unbalanced matches from time to time.

The best matches are those filled entirely with solo players. It allows MM to function more accurately because with solo players there's more granularity, and therefore less disparity between each player's Elo score.

You have an entirely different perspective on what makes for a good match than I do. A great match to me is when everyone is on he same page and we move in a coordinated fashion massing fire on each enemy in succession. That is an awesome match in my eyes. I miss full Lawman drops!

Mudhutwarrior, on 14 January 2014 - 10:20 AM, said:

I know two Mariks who will probably try to have you drummed out for saying that.

Not just Mariks kind sir!

[Zordicron]

Setting your mode to "any" can help with the tonnage issue, a little. Premade groups still F it up most of the time. But if you luck out and get 12 puggies vs another 12, you can successfully run lights and mediums in a somewhat fair tonnage match up, although obviously you have no choice of game mode.

Alternatively, I have had better matches by running lights in conquest only, mediums in assault, and anything bigger I leave on assault or any. I have been avoiding my big stuff lately because it is really difficult to find an even match lately with those. Conquest has been historically a total abomination for me, but, lately with the skirmish mode, there have been a lot lower tonnage totals in the mode and the "chaos" in this mode is more tolerable( like you can recover from n00bs running all over the dang map without a clue more easily when there are a lot more speedy mechs on the team)

I think there are less players on right now, and adding skirmish has also created a situation where more players are selecting a particular mode instead of any. bvut that is also speculation based on observation on my part.

[Ghogiel]

Hotthedd, on 14 January 2014 - 10:08 AM, said:

Source please.

I have seen absolutely ZERO evidence, either empirical or otherwise that suggests that any matches are using tonnage as a factor in the matchmaker.

It apparently still uses weight. I can cite some patch notes that aren't too old that mention it still being a factor eg> http://mwomercs.com/...42-17-sep-2013/

"
Gameplay

- 12 players pre-made group launch is 1st Person View only.
- Match making ELO and Weight Class thresholds narrowed down. NB: These changes will not affect 12 player groups.
- The longer a match took to kick off, the broader the range the MM uses in terms of matching ELO and weight class.
- This reduction means that the MM will no longer search as broad a spectrum over time (still increases but in a narrower range). It will be much more refined in terms of how much variance will be given in matching players and 'Mechs.
- Match making timeout increased to 3 minutes.
- This allows the MM more time to find players before terminating its search."


that was the last thing I recall. Odd that it actually mentions weight class matching, which I thought they ditched at some point in favor of actual weight matching due to the case of awesome vs atlas

And I do recall somewhere reading it uses Elo as the primary over weight. going to have to troll the ask the dev posts :\

[AdamBaines]

Mister Blastman, on 14 January 2014 - 09:52 AM, said:

Doing a few mid-day drops, I have observed way more nubs than I think I have ever seen before.

Conclusion: All the long-timers are quitting the game.

PGI better do something, fast.

Now that's just solid science...........?

[Mister Blastman]

Hotthedd, on 14 January 2014 - 10:10 AM, said:

Source please.

I have seen absolutely ZERO evidence, either empirical or otherwise that suggests that any matches are using tonnage as a factor in the matchmaker.


Once upon a time, many, many moons ago, PGI actually communicated with the players.

Hmm. I have a very witty reply but I'm not sure if it would violate the COC. Hmm... choices...

[Bhael Fire]

Joseph Mallan, on 14 January 2014 - 10:23 AM, said:

You have an entirely different perspective on what makes for a good match than I do. A great match to me is when everyone is on he same page and we move in a coordinated fashion massing fire on each enemy in succession. That is an awesome match in my eyes. I miss full Lawman drops!

Actually, I think we have similar ideas of what makes a great game.

When MM is working properly, I find that I get matched with other like-minded players that are keen on organizing into structured formations and coordinated attacks.

So yeah, the best matches are when your teammates instinctively know what needs to be done and everyone is on the same page, moving in unison efficiently and with lethal veracity.

[Joseph Mallan]

Bhael Fire, on 14 January 2014 - 11:29 AM, said:

Actually, I think we have similar ideas of what makes a great game.

When MM is working properly, I find that I get matched with other like-minded players that are keen on organizing into structured formations and coordinated attacks.

So yeah, the best matches are when your teammates instinctively know what needs to be done and everyone is on the same page, moving in unison efficiently and with lethal veracity.

So a well practiced team that you have played with for several months honing trust an friendship!

[Doctor Proctor]

Joseph Mallan, on 14 January 2014 - 11:31 AM, said:

So a well practiced team that you have played with for several months honing trust an friendship!

No, what he's saying is that if MM were working properly say, Bhael Fire and myself would get matched up on the same team. Why? Because we're both vets that understand the value of tactics and try to communicate in game, and should therefore be a higher Elo (and more likely to be matched together) than someone who just downloaded the client for the first time today.

I always solo drop, and there are people who will listen and chat, as well as understand tactics of how to deal with the enemy. You can call out targets to focus on and they'll do it, or say "Let's defend this point" and they'll listen, or even coordinate a complex capping strategy to win a Conquest match against a team that massively outweighs yours. The problem is that those of us who are like that are not getting matched with like minded people, which creates for a frustrating environment.

[Joseph Mallan]

Doctor Proctor, on 14 January 2014 - 11:55 AM, said:

No, what he's saying is that if MM were working properly say, Bhael Fire and myself would get matched up on the same team. Why? Because we're both vets that understand the value of tactics and try to communicate in game, and should therefore be a higher Elo (and more likely to be matched together) than someone who just downloaded the client for the first time today.

I always solo drop, and there are people who will listen and chat, as well as understand tactics of how to deal with the enemy. You can call out targets to focus on and they'll do it, or say "Let's defend this point" and they'll listen, or even coordinate a complex capping strategy to win a Conquest match against a team that massively outweighs yours. The problem is that those of us who are like that are not getting matched with like minded people, which creates for a frustrating environment.

However that guy who just joined today needs to see the game played right so he can possibly avoid having to unlearn bad habits learned playing with other pure Boots. Boots need a few weeks of training missions before being thrown into the mix I agree, but there is no better way for some to learn than being thrown to the wolves! Heck last time I played a computer game was back in the 90s. I was a console veteran but a PvP boot, I got my legs in 2 weeks and bearings in 4. Been a good ride ever since. that includes the last 6-8 months of almost PUG only!

Edited by Joseph Mallan, 14 January 2014 - 12:01 PM.

[BarHaid]

So what we need is a complicated AI that determines our play style by our actions and chat logs, and then compiles a physiological profile to add to our W/L and K/D ratios.

...

Something off the shelf, and cheap.

...

Instead of Community Warfare.

.
..
...

Naw, I'll pass on that.

[Joseph Mallan]

BarHaid, on 14 January 2014 - 12:01 PM, said:

So what we need is a complicated AI that determines our play style by our actions and chat logs, and then compiles a physiological profile to add to our W/L and K/D ratios.

...

Something off the shelf, and cheap.

...

Instead of Community Warfare.

.
..
...

Naw, I'll pass on that.

No. We need to jump in with both feet and figure it out as we go along. It's how I been doing things for decades.

[Akulakhan]

BarHaid, on 14 January 2014 - 12:01 PM, said:

So what we need is a complicated AI that determines our play style by our actions and chat logs, and then compiles a physiological profile to add to our W/L and K/D ratios.

...

Something off the shelf, and cheap.

...

Instead of Community Warfare.

.
..
...

Naw, I'll pass on that.

Aww, you think you'll actually get community warfare someday! That's so cute.

[Doctor Proctor]

Joseph Mallan, on 14 January 2014 - 12:00 PM, said:

However that guy who just joined today needs to see the game played right so he can possibly avoid having to unlearn bad habits learned playing with other pure Boots. Boots need a few weeks of training missions before being thrown into the mix I agree, but there is no better way for some to learn than being thrown to the wolves! Heck last time I played a computer game was back in the 90s. I was a console veteran but a PvP boot, I got my legs in 2 weeks and bearings in 4. Been a good ride ever since. that includes the last 6-8 months of almost PUG only!

If you would've looked at the post that I linked earlier you would see the old Elo graph that PGI thought was garbage. In fact, it was actually right where it should be. New players were starting probably at the bottom left, but the upper 50% of them were crossing over with a big bump of veteran players. This means that fresh recruits fight each for awhile to learn controls and such, but as they start to get better they'll get more vets in their game that can teach them more advanced tactics. Around 1550 Elo the newbies tend to disappear and the rest of the Elo ranks are solely veterans that already have a handle on the game.

What we have now instead is that from about 1145 - 1395 Elo is where most of the cadets are, which also happens to overlap with the center of the veteran bell curve. When newbies are getting out of that starter Elo they're obviously staying around the same central zone in the bell curve and remaining stagnant. So the middle 50% or so (probably more, but we can't see the vertical scale in these graphs) just sits in the middle with a large portion of their drop being made up of new recruits every time that prevent them from being able to coordinate well with other veteran players.

BarHaid, on 14 January 2014 - 12:01 PM, said:

So what we need is a complicated AI that determines our play style by our actions and chat logs, and then compiles a physiological profile to add to our W/L and K/D ratios.[snipped]

No, because players that work together as a team will tend to naturally rise higher in the game since it's so team focused. However, they can't do that very well right now because half the time there might be people in their drop who don't know the difference between all chat and team chat when talking about what the team is planning to do.

This is why premades in PUG matches tend to do so well, is that they already have the coordination down and just have to execute on it. Solo droppers have to get everyone to coordinate, and if they can't, must perform at absolutely stellar levels to turn a match around at times.

Edited by Doctor Proctor, 14 January 2014 - 12:15 PM.

[Ghogiel]

Doctor Proctor, on 14 January 2014 - 12:13 PM, said:

If you would've looked at the post that I linked earlier you would see the old Elo graph that PGI thought was garbage. In fact, it was actually right where it should be. New players were starting probably at the bottom left, but the upper 50% of them were crossing over with a big bump of veteran players. This means that fresh recruits fight each for awhile to learn controls and such, but as they start to get better they'll get more vets in their game that can teach them more advanced tactics. Around 1550 Elo the newbies tend to disappear and the rest of the Elo ranks are solely veterans that already have a handle on the game.

What we have now instead is that from about 1145 - 1395 Elo is where most of the cadets are, which also happens to overlap with the center of the veteran bell curve. When newbies are getting out of that starter Elo they're obviously staying around the same central zone in the bell curve and remaining stagnant. So the middle 50% or so (probably more, but we can't see the vertical scale in these graphs) just sits in the middle with a large portion of their drop being made up of new recruits every time that prevent them from being able to coordinate well with other veteran players.

The middle ranges of the sub 50 games player group is nearly the same range between to 2 graphs. The only thing wrong with the old graph was that once you played the {Scrap} out the game your Elo was inflated which is why there was a huge section player base all at the maxium Elo range possible. The longer that issue would have went on then more noobs, medium, vet, elite players would have made it 2800.

What's MORE bizarre in the old graph you can see players with sub 50 games played in the 2000+ elo ranges.

Edited by Ghogiel, 14 January 2014 - 12:52 PM.

[Doctor Proctor]

Ghogiel, on 14 January 2014 - 12:50 PM, said:

The middle ranges of the sub 50 games player group is nearly the same range between to 2 graphs. The only thing wrong with the old graph was that once you played the {Scrap} out the game your Elo was inflated which is why there was a huge section player base all at the maxium Elo range possible. The longer that issue would have went on then more noobs, medium, vet, elite players would have made it 2800.

What's MORE bizarre in the old graph you can see players with sub 50 games played in the 2000+ elo ranges.

I didn't say it was absolutely perfect, but the distribution of new players and the bulk of existing players made for a much better skill gradient than the "everyone in the same bucket!" mentality that exists now.

As for the group of those at the high end, well that can be solved in a couple of ways. Either 1) Reduce the Elo cap to lower level so that they'll rejoin some of those players in low pop section of the mid-to-high Elo bracket, 2) give diminishing returns on Elo gains as you move up the rankings or have win streaks in order to slow the progression up the brackets after you fall off the large hump where the bulk of the vets are, or 3) simply don't worry about and just increase the acceptable Elo range for MM for high Elo players.

[Sug]

Elo is, sadly, working. In 2 years I've maybe played 4 matches in assault mechs, finally got in my phoenix battlemaster last night for a few rounds, was shocked at the difference in players. It was very obvious I was in a lower Elo range.

Didn't recognize any names and I saw mechs I had forgotten were in the game. Trebs?? Dragons? Dear god every team had at least one Awesome on it. It was madness. And ZERO pop tarts. Lots of LRMs, who uses that **** anymore? Soooo many pulse lasers..

Omg, People were using NARC....

Seriously there was like, a Commando(?) using NARC...

[focuspark]

zhajin, on 14 January 2014 - 06:33 AM, said:

the problem is elo score is team based not player based. this means that it will mix high and low elo players to equal out a mid level player. this and the drop weight issues are pretty common during low population times.

also keep in mind that groups pretty much screw up elo completely.

overall there are too many factors and not enough players to match properly most of the time. some of it I suspect is poor decisions on pgi's part. but mostly its just to complex a system, with too small a player pool.

but overall it still seems better than the day of completely random matching.

Quoted for truth. People need to understand that ELO will not solve your personal match placement. It does matching and then predicts the winners. While it does attempt to drop balanced teams, it more than likely just drops A vs B with the assumption that A will steam-roll B and when that happens does a zero score modification.

ELO is a scoring mechanic, not a match making mechanic.

Bhael Fire, on 14 January 2014 - 07:29 AM, said:

When I said premade, I was referring to any premade group....not just the 12-man groups.

Whenever you drop in a group, it makes it harder for Matchmaker to place your group...so you're bound to see other familiar players that are also in groups, mixed in with "filler" solo players of various Elo scores. This is why you see weird mixed up and unbalanced matches from time to time.

The best matches are those filled entirely with solo players. It allows MM to function more accurately because with solo players there's more granularity, and therefore less disparity between each player's Elo score.

To make things worse, when a highly skilled premade (of 4) drops it's often with 8 newbs vs 12 moderate players. Then results is often a slaughter - depending on the premade it could be them or the other; but since steamrolled slaughters aren't fun it sucks for everyone.

[Ghogiel]

focuspark, on 14 January 2014 - 01:34 PM, said:

Quoted for truth. People need to understand that ELO will not solve your personal match placement. It does matching and then predicts the winners. While it does attempt to drop balanced teams, it more than likely just drops A vs B with the assumption that A will steam-roll B and when that happens does a zero score modification.

ELO is a scoring mechanic, not a match making mechanic.

regardless of it being a scoring system, Elo rating is actually used in the MM component of MWO to place players in matches.