How can ELO actually work if it allows for 4 man groups to go against pugs?
As I understand it, ELO simply averages teams ratings...
Why the hell doesn't a 4 man group get a higher rating by default, so they cannot go against full pugs? Like 2% higher rating per group member.
This would put coordinated 4 mans against the best puggers, or another 4 man group.
I don't get it.. It seems like such a simple thing to figure out..
0
Elo And Groups Question
Started by Mxxpower, Feb 27 2013 05:56 AM
7 replies to this topic
#3
-
-
- 143 posts
Member
- LocationMINNESOTA
Posted 27 February 2013 - 06:18 AM
Chavette, on 27 February 2013 - 06:11 AM, said:
Or just, have something like group+pugs vs group+pugs teams matches, to make it somewhat even, like in my sig.
PGI should hire me.
PGI should hire me.
Well I think the group added rating would accomplish this better, because a 2 man group really shouldn't HAVE to be matched against a 2 man, just a few good puggers on the other side.
1 or 2% increase in ELO rating per group member would solve this for ANY size group actually, 5, 6, 7 man, whatever size. By the time you have a full 8 man, your base rating would be so hi that you couldn't possibly go against a non 7 or 8 man team.
#4
Posted 27 February 2013 - 06:28 AM
Thats a good idea too.
It doesn't say it has to be 2 man vs 2man, there would be a pretty small chance of that happening, as the system already has to take in weight balance+elo balancing.
My method (seemingly yours too) help the issue of enabling 5-7man groups to play without an exponentially high chance of a stomp-fest happening.
It doesn't say it has to be 2 man vs 2man, there would be a pretty small chance of that happening, as the system already has to take in weight balance+elo balancing.
My method (seemingly yours too) help the issue of enabling 5-7man groups to play without an exponentially high chance of a stomp-fest happening.
#5
-
-
- Philanthropist
- 606 posts
Member
Posted 27 February 2013 - 06:29 AM
(1) I thought PGI was considering a modifier to group rating.
E.g., for example, if the average Elo of a group of 2 people was 1300, maybe they'd MM (Match-Make) at a 1400 or something? Or a separate Elo for you when you group . . . of course every new Elo rating category they add makes the data less reliable. If I need a 1, 2, 3, 4 and 8 person group ELO for every chassis I buy [hint, that stock chassis with SHS and standard internal using only the weapons I have already purchased that I bought to grind skill . . . is not as good as my fully customized machine], for example, the Elos will be a lot better in the long run, but only if I play another thousand games. If it takes 50 games to get a reasonable Elo picture, that's potentially 1,000 games while you're getting full elite on just one chassis.
(2) I think people need to be careful to remember that groups are 1-4 people. Any solution that only works with groups of 4 will eliminate any 2 or 3 person groups. Which means people who invite their friend to play pretty much aren't going to be able to help their friend.
(3) People need to remember that you can't just add qualifiers to the queue system willy nilly. Every additional requirement is going to add to queue length. If you want strict weight matching, class matching, elo matching, and premade size matching (so each team gets one 2 person team and one 3 person team of about the same ELO ratings) . . . you may have to be looking at 15-30 minute waits for matches. At some point, you'd rather just play 3 matches and risk two unfair matches rather than wait that long. I'm not saying the system shouldn't be tuned. I'm just saying understand that it is not as simple as just adding another restriction to the system.
(4) If it makes you feel any better, I had a 16 game losing streak in a 4 man premade last night. That's even worse than my 25% win ratio solo. So, you know. It's not all candy and rainbows. Granted, it was a pick up group, not a regular group. But they have to balance that carefully too. 4 new players who just started playing in the last 2 weeks and don't know each other =/= 4 experienced players who have played together as a group for months.
E.g., for example, if the average Elo of a group of 2 people was 1300, maybe they'd MM (Match-Make) at a 1400 or something? Or a separate Elo for you when you group . . . of course every new Elo rating category they add makes the data less reliable. If I need a 1, 2, 3, 4 and 8 person group ELO for every chassis I buy [hint, that stock chassis with SHS and standard internal using only the weapons I have already purchased that I bought to grind skill . . . is not as good as my fully customized machine], for example, the Elos will be a lot better in the long run, but only if I play another thousand games. If it takes 50 games to get a reasonable Elo picture, that's potentially 1,000 games while you're getting full elite on just one chassis.
(2) I think people need to be careful to remember that groups are 1-4 people. Any solution that only works with groups of 4 will eliminate any 2 or 3 person groups. Which means people who invite their friend to play pretty much aren't going to be able to help their friend.
(3) People need to remember that you can't just add qualifiers to the queue system willy nilly. Every additional requirement is going to add to queue length. If you want strict weight matching, class matching, elo matching, and premade size matching (so each team gets one 2 person team and one 3 person team of about the same ELO ratings) . . . you may have to be looking at 15-30 minute waits for matches. At some point, you'd rather just play 3 matches and risk two unfair matches rather than wait that long. I'm not saying the system shouldn't be tuned. I'm just saying understand that it is not as simple as just adding another restriction to the system.
(4) If it makes you feel any better, I had a 16 game losing streak in a 4 man premade last night. That's even worse than my 25% win ratio solo. So, you know. It's not all candy and rainbows. Granted, it was a pick up group, not a regular group. But they have to balance that carefully too. 4 new players who just started playing in the last 2 weeks and don't know each other =/= 4 experienced players who have played together as a group for months.
#6
-
-
- 143 posts
Member
- LocationMINNESOTA
Posted 27 February 2013 - 06:35 AM
DanNashe, on 27 February 2013 - 06:29 AM, said:
(1) I thought PGI was considering a modifier to group rating.
E.g., for example, if the average Elo of a group of 2 people was 1300, maybe they'd MM (Match-Make) at a 1400 or something? Or a separate Elo for you when you group . . . of course every new Elo rating category they add makes the data less reliable. If I need a 1, 2, 3, 4 and 8 person group ELO for every chassis I buy [hint, that stock chassis with SHS and standard internal using only the weapons I have already purchased that I bought to grind skill . . . is not as good as my fully customized machine], for example, the Elos will be a lot better in the long run, but only if I play another thousand games. If it takes 50 games to get a reasonable Elo picture, that's potentially 1,000 games while you're getting full elite on just one chassis.
(2) I think people need to be careful to remember that groups are 1-4 people. Any solution that only works with groups of 4 will eliminate any 2 or 3 person groups. Which means people who invite their friend to play pretty much aren't going to be able to help their friend.
(3) People need to remember that you can't just add qualifiers to the queue system willy nilly. Every additional requirement is going to add to queue length. If you want strict weight matching, class matching, elo matching, and premade size matching (so each team gets one 2 person team and one 3 person team of about the same ELO ratings) . . . you may have to be looking at 15-30 minute waits for matches. At some point, you'd rather just play 3 matches and risk two unfair matches rather than wait that long. I'm not saying the system shouldn't be tuned. I'm just saying understand that it is not as simple as just adding another restriction to the system.
(4) If it makes you feel any better, I had a 16 game losing streak in a 4 man premade last night. That's even worse than my 25% win ratio solo. So, you know. It's not all candy and rainbows. Granted, it was a pick up group, not a regular group. But they have to balance that carefully too. 4 new players who just started playing in the last 2 weeks and don't know each other =/= 4 experienced players who have played together as a group for months.
E.g., for example, if the average Elo of a group of 2 people was 1300, maybe they'd MM (Match-Make) at a 1400 or something? Or a separate Elo for you when you group . . . of course every new Elo rating category they add makes the data less reliable. If I need a 1, 2, 3, 4 and 8 person group ELO for every chassis I buy [hint, that stock chassis with SHS and standard internal using only the weapons I have already purchased that I bought to grind skill . . . is not as good as my fully customized machine], for example, the Elos will be a lot better in the long run, but only if I play another thousand games. If it takes 50 games to get a reasonable Elo picture, that's potentially 1,000 games while you're getting full elite on just one chassis.
(2) I think people need to be careful to remember that groups are 1-4 people. Any solution that only works with groups of 4 will eliminate any 2 or 3 person groups. Which means people who invite their friend to play pretty much aren't going to be able to help their friend.
(3) People need to remember that you can't just add qualifiers to the queue system willy nilly. Every additional requirement is going to add to queue length. If you want strict weight matching, class matching, elo matching, and premade size matching (so each team gets one 2 person team and one 3 person team of about the same ELO ratings) . . . you may have to be looking at 15-30 minute waits for matches. At some point, you'd rather just play 3 matches and risk two unfair matches rather than wait that long. I'm not saying the system shouldn't be tuned. I'm just saying understand that it is not as simple as just adding another restriction to the system.
(4) If it makes you feel any better, I had a 16 game losing streak in a 4 man premade last night. That's even worse than my 25% win ratio solo. So, you know. It's not all candy and rainbows. Granted, it was a pick up group, not a regular group. But they have to balance that carefully too. 4 new players who just started playing in the last 2 weeks and don't know each other =/= 4 experienced players who have played together as a group for months.
That is why I like the percent based system. It doesn't mean you will always have groups on both sides, but there is a higher and higher chance as you get up to 4 people because the group rating would be so high.
4 man group avg ELO X 8% - 2% per group member.
You are right in that allowing groups ONLY to fight groups would slow the queue down too much, this system would alleviate that.
#7
-
-
- Stone Cold
- 11,975 posts
Welcoming Committee
- LocationNewfoundland
Posted 27 February 2013 - 06:40 AM
Other games do account for group play as a force multiplier and adjust the Elo matching to account for this. Give PGI time to analyze the data and see if they need to follow suit.
#8
-
-
- Wrath
- 1,436 posts
Member
- LocationGermany
Posted 27 February 2013 - 11:14 AM
Redshift2k5, on 27 February 2013 - 06:40 AM, said:
Other games do account for group play as a force multiplier and adjust the Elo matching to account for this. Give PGI time to analyze the data and see if they need to follow suit.
Other games have a lobby and dont even run into this dead end road!
1 user(s) are reading this topic
0 members, 1 guests, 0 anonymous users
