Question about squads and matchmaking

Here you can discuss the game in general and provide us your opinion or feedback on the project.

12.11.2018, 23:54

When you squad with someone, does the matchmaking system take both player ELOs into account to determine who you play with/against, or does it only take the squad leader?

I may not have explained well, but my example is my last match where I squaded with a clan mate with a much lower ELO than my own, and it put us in a match full of only players with ELOs of around 1000 and below.

Would someone be able to confirm please?
User avatar
Scaramanga
 
Posts: 124
Joined: 06.10.2015, 21:11

13.11.2018, 01:18

I would really like to see the "3 category" thing working, or at least someone who actually knows how it works explaining it, what are the ELO limits etc. Right now it seems to be working just like before. Of course the 1 tier system makes it so that there really isn't need for different categories, especially as those are thrown away when there simply aren't enough same category players in queue.

But yeah, this https://survarium.pro/matches/12321008 sure looks like what you described. I'm hoping it's only because there weren't any higher ELO players in queue.

Tsjuder
 
Posts: 1711
Joined: 13.04.2013, 15:50

13.11.2018, 10:19

Tsjuder:But yeah, this https://survarium.pro/matches/12321008 sure looks like what you described. I'm hoping it's only because there weren't any higher ELO players in queue.


Freaking nice. And here I am with my "rightfully earned" 2k> ELO getting thrown exclusively into 3rd bracket with freaking pros and getting my face caved in most of the time, while others get a free pass to play with average players...
User avatar
UnknownPlayer666
 
Posts: 769
Joined: 30.03.2017, 16:22

13.11.2018, 14:54

UnknownPlayer666:

Freaking nice. And here I am with my "rightfully earned" 2k> ELO getting thrown exclusively into 3rd bracket with freaking pros and getting my face caved in most of the time, while others get a free pass to play with average players...


When I play alone, I get put into matches with equal(ish) ELO players. This only happens when I squad, and every time I've squaded recently, my clan mates have been the squad leaders, which is what makes me think matchmaking bases the ELO "bracket" on the squad leader only.

Fun as it is to be the best player in a match, it's somewhat frustrating when your entire team is low skilled and not really able to properly play match objective
User avatar
Scaramanga
 
Posts: 124
Joined: 06.10.2015, 21:11

13.11.2018, 17:24

New elo formula seems to work. I played a lot of games since 0.55 and i had maybe 1 match that was freaking not balanced -> https://survarium.pro/matches/12314119
User avatar
Ventori
 
Posts: 414
Joined: 09.12.2015, 11:48

13.11.2018, 18:13

Ventori:New elo formula seems to work. I played a lot of games since 0.55 and i had maybe 1 match that was freaking not balanced -> https://survarium.pro/matches/12314119

Squad ELO multiplier (I believe you were in squad with Vimako?) + some people not played (enough) in .55 to raise their ELO to the right level cause this.

With the ELO brackets I wonder if it would be worth it to test different multipliers as you see:
3000*1,25= 3750 -> 750 increase
2000*1,25= 2500 -> 500 increase
1000*1,25= 1250 -> 250 increase
Not saying these are the brackets, just example numbers.
Or maybe just put a plain 500 increase in all?

More examples of balancing (good or bad) with squads would be nice, from anyone!

Tsjuder
 
Posts: 1711
Joined: 13.04.2013, 15:50

13.11.2018, 19:10

@Tsjuder, maybe this is a helpful sample match for you. Enemy team was a three man squad. We were two single players and one who couldn't stand the pressure :D All six players have quite similar ELO.
-> https://survarium.pro/matches/12323751

... they crushed us, but yet it was big fun :-)
User avatar
jule
 
Posts: 399
Joined: 27.05.2013, 15:22
Location: Hävern

14.11.2018, 01:44

UnknownPlayer666:Freaking nice. And here I am with my "rightfully earned" 2k> ELO getting thrown exclusively into 3rd bracket with freaking pros and getting my face caved in most of the time, while others get a free pass to play with average players...

You could have it worse, I have some 1,7k ELO (granted I haven't played much in .55, or for a while before that...) and I get matched with the same people, In the last few games I played I was top 3 lowest ELO player in all of them. While highest ELO was most of the time 2,3k-2,6k...
But one could have it even worse, https://survarium.pro/matches/12324441 ,1k ELO player with 2k... Of course his team won, because the lvl 100 players on his team don't have correct ELOs (at least for Research)...

@jule, don't know if that helps much. MM couldn't do anything else. Even if it feels stupid, MM just tried to make a match in reasonable time...

Tsjuder
 
Posts: 1711
Joined: 13.04.2013, 15:50

14.11.2018, 11:12

Scaramanga:When you squad with someone, does the matchmaking system take both player ELOs into account to determine who you play with/against, or does it only take the squad leader?

I may not have explained well, but my example is my last match where I squaded with a clan mate with a much lower ELO than my own, and it put us in a match full of only players with ELOs of around 1000 and below.

Would someone be able to confirm please?


That is normal I guess. If you think that in Survarium you have MM who is working and can see the difference of ELO of each players it is a fake news.
For me the MM good only around 0.24, 0.27 I do not remember and from that time ....

The MM is so good that it is even not able to calcul that you have 3 squad in a team and no one in the opposite one.
The most funny is a match with 3 top squad like ILS and in other no squad of normal player = farm camping = most players leaving the match.

In VG they have a lot of idea, problems is they are never able to make perfect.
User avatar
William-g
 
Posts: 1257
Joined: 07.04.2014, 10:57
Location: warsaw

14.11.2018, 11:18

William-g:The MM is so good that it is even not able to calcul that you have 3 squad in a team and no one in the opposite one.
The most funny is a match with 3 top squad like ILS and in other no squad of normal player = farm camping = most players leaving the match.

You are free to show examples of such matches. I like to think I have been pretty good at explaining MM before, .55 is little bit different, not only Scaramanga's case of different level players in squad (which as you can see we're trying to figure out here) but with the "bot farming compensation" with >2k ELO dropped to 2k. Other than that MM does pretty good job.
The only fake news here is your ".24 MM was better".

Tsjuder
 
Posts: 1711
Joined: 13.04.2013, 15:50

14.11.2018, 12:47

Tsjuder:
You are free to show examples of such matches. I like to think I have been pretty good at explaining MM before, .55 is little bit different, not only Scaramanga's case of different level players in squad (which as you can see we're trying to figure out here) but with the "bot farming compensation" with >2k ELO dropped to 2k. Other than that MM does pretty good job.
The only fake news here is your ".24 MM was better".


You have a case here: https://survarium.pro/matches/12299065
After 2 or 3 minutes we where only 3 players who still play the match, even dables ask us to leave the match because it was really stupid situation and only [в-у] DeD_Validol play the match to the end. We couldn't even respraw as soon we respraw who where dead.

An other game here where Hand of Middas where in our base after two minutes https://survarium.pro/matches/12324933 impossible to respraw.

And sorry but from what I remember around version 0.24 it was a better balanced MM and one of the reason is that it was less players than now.
Last edited by William-g on 14.11.2018, 12:51, edited 2 times in total.
User avatar
William-g
 
Posts: 1257
Joined: 07.04.2014, 10:57
Location: warsaw

14.11.2018, 17:04

William-g:

Oh you meant only that, I thought you meant "3 squads in one team", but you meant "a 3 man squad in one team and no squads in other".
Well what do you want MM to do? Go all "you can't do that", on players with >2,5k ELO when they try to make a squad? On Hand of Midas if he even tries to play? Sure, it feels unfair and broken when you get matched against these, but there is no real solution here. You can't forbid people from playing just because they are good.

One interesting problem with ELO (and all those other point systems discussed in some topic months ago) and Hand of Midas and other godlike players: It's calculated compared to the points of others; When you kill someone with low ELO your ELO rises only little, when you kill someone with high ELO, your ELO rises a lot. Opposite happens for getting killed. As ELOs were dropped to max 2k, the much better players first got quickly to slightly higher ELO, but after while it slowed down to almost nothing, as the enemies they keep killing don't have correct ELO. Let's say someone should have >3k ELO, but it was dropped to 2k (actually really good example https://survarium.pro/players/Kube rating TDM 3,5k PvP 2,4k) and someone with >3k ELO kills him, the ELO should increase "significantly" (noticeably?), as he killed one on the same skill level, but it only increases tiny amount because the game thinks he had only 2k ELO.
Same is of course true for all slightly above 2k, but for those it balances out much better.
Hopefully that made some sense to someone? The point being: ELO reduction was possibly a mistake whose consequences we will have to bear with for a long while.
I'll quote myself from this topic (those other ranking systems were also discussed there) viewtopic.php?f=3&t=16016
Tsjuder:I think they shouldn't reset ELOs after the change. That's because even if the values are false they serve as better start values than to reset everyone to 1k. Especially as the MM will be able to match similar ELOs in same match with a bigger pool to choose players from. Meaning people who got high ELOs with bot farming will play against others with high ELO, if high rank is only caused by bot farming their rank will drop as they are going to lose against those who really "deserve" high rank. And opposite happens with those whose ELO is low for not doing bot farming. Even those who "deserve" high ELO would get their ELO dropped, as they will still get killed by lower rank players, which should cause huge drops for them.
If they did reset ELO, hmm I hope this really doesn't need explaining... Everyone would start from the same line, MM would put all different skill level players in the same match, causing high rank players to smash lower rank players until ELO catches up. So if they reset it or not it will with time correct itself, it boils down to do you want to punish low rank players or bot farmers.

I think VG tried to go the Middle Way by not fully resetting, but only reducing >2k to 2k, but I fear it still didn't work.

Tsjuder
 
Posts: 1711
Joined: 13.04.2013, 15:50

14.11.2018, 18:19

Tsjuder:When you kill someone with low ELO your ELO rises only little, when you kill someone with high ELO, your ELO rises a lot.

Tsjuder:Meaning people who got high ELOs with bot farming [...]


Mhh.. well I am not really up-to-date on (at?, in?) this topic and maybe this is a bit naive, but I wonder why, if ELO calculation distinguishes between killing higher ELO players and lower ELO players, then why does it not also distinguish between killing a BOT and killing a human player? Because this should have prevented the ELO "poisoning" through bot-farming in the first place, shouldn't it? :O

Sorry this is a bit off the actual topic.
Last edited by jule on 14.11.2018, 18:23, edited 2 times in total.
User avatar
jule
 
Posts: 399
Joined: 27.05.2013, 15:22
Location: Hävern

14.11.2018, 18:35

jule:


Mhh.. well I am not really up-to-date on (at?, in?) this topic and maybe this is a bit naive, but I wonder why, if ELO calculation distinguishes between killing higher ELO players and lower ELO players, then why does it not also distinguish between killing a BOT and killing a human player? Because this should have prevented the ELO "poisoning" through bot-farming in the first place, shouldn't it? :O

Sorry this is a bit off the actual topic.


^Legit question.
User avatar
UnknownPlayer666
 
Posts: 769
Joined: 30.03.2017, 16:22

14.11.2018, 18:39

jule:Mhh.. well I am not really up-to-date on (at?, in?) this topic and maybe this is a bit naive, but I wonder why, if ELO calculation distinguishes between killing higher ELO players and lower ELO players, then why does it not also distinguish between killing a BOT and killing a human player? Because this should have prevented the ELO "poisoning" through bot-farming in the first place, shouldn't it? :O

Damn you caught me again. ;)

Not really sure what the deal with this is/was. Either they changed the way ELO increases/decreases to win/lose instead of kill/death (meaning ELO would be affected by real players in enemy team) at some point, maybe even taking account personal score? Or maybe only score? I don't know, that seems like a bad way to calculate ELO, but you're right, giving bots some ELO (they need to have it for bot farming to increase ELO) is maybe even worse solution... Maybe it was a quick solution so the game doesn't freak out when you get a kill and the enemy doesn't have ELO at all?
Thanks for pointing this out, I somehow didn't think about this at all, and sorry for not having a good explanation for this...
Edit: Almost forgot, not related to bot farming, actually the opposite:
It would be super interesting to see some numbers on Rating ELOs and winrates (preferrably a graph)- before .55 and after once we have a new season (I expect we'll have similar ELO categories for League aswell?). It would show very well if the MM performs as it's supposed to - meaning 50% winrate for everyone (well in the long run anyways).
Sadly I think it's something VG won't share, and I don't know how hard would it be to get that info from survarium.pro ...
Last edited by Tsjuder on 14.11.2018, 18:45, edited 1 time in total.

Tsjuder
 
Posts: 1711
Joined: 13.04.2013, 15:50

14.11.2018, 18:46

Tsjuder:
Oh you meant only that, I thought you meant "3 squads in one team", but you meant "a 3 man squad in one team and no squads in other"....


No,no I really mean 3 squad in one team, like a squad of 3 and 2 of 2 or 3 squad of 2 in one team and other team, no squad at all.
User avatar
William-g
 
Posts: 1257
Joined: 07.04.2014, 10:57
Location: warsaw

14.11.2018, 18:52

William-g:No,no I really mean 3 squad in one team, like a squad of 3 and 2 of 2 or 3 squad of 2 in one team and other team, no squad at all.

Well could you point out the squads? Apart from top3 it's not clear who are in squads and who are not
https://survarium.pro/matches/12299065
And I don't know if it really mattered if others were in squads also, as the 2xILS+GG is what caused you to lose. IMO.

Tsjuder
 
Posts: 1711
Joined: 13.04.2013, 15:50

14.11.2018, 19:03

Tsjuder:
It would be super interesting to see some numbers on Rating ELOs and winrates (preferrably a graph)- before .55 and after once we have a new season (I expect we'll have similar ELO categories for League aswell?). It would show very well if the MM performs as it's supposed to - meaning 50% winrate for everyone (well in the long run anyways).
Sadly I think it's something VG won't share, and I don't know how hard would it be to get that info from survarium.pro ...

I would start from this:
It would be super interesting to see full ELO equation, to understand how is it calculated.
Then we could compare it with results, graphs, etc. of season leagues or matches in some period of time.
Without it the whole topic is like reading tea leaves (if you know what I mean).
User avatar
Tanatloc
 
Posts: 740
Joined: 08.03.2016, 17:01

15.11.2018, 00:19

I don't know why they didn't just reset everyone's PVP ELOs to match their rating ELO? To me that makes much more sense than just bringing everyone down to 2k. I assume at least one of the programmers must have thought of this at some point, no?
User avatar
Scaramanga
 
Posts: 124
Joined: 06.10.2015, 21:11

15.11.2018, 10:02

Scaramanga wrote:I don't know why they didn't just reset everyone's PVP ELOs to match their rating ELO? To me that makes much more sense than just bringing everyone down to 2k. I assume at least one of the programmers must have thought of this at some point, no?


Dima did not think a complete reset was the best idea. According to him and all his calculations, bringing the super-high ELO players down to 2000 was the best solution. Now, you may not agree with him. But at this stage, it's really a moot point. Overtime, ELO will adjust per player to better reflect their skill.
User avatar
joewillburn
Vostok Games
 
Posts: 6419
Joined: 03.04.2013, 11:25

In order to reply you must be signed in.

Return to Game Discussion



cron