Login

russian armor

Explain this ladder sink

17 Aug 2022, 01:58 AM
#1
avatar of Reverb

Posts: 319

Match 1:
I play as random, matched against arranged team. Loss: lose 32 ladder positions.


Match 2:
Next match same thing after about 1 minute queuetime, I play as random, I match with the exact same partner, against the exact same arranged team opponents. This time is a win, gain 4 ladder positions.

So me and my teammate had lower ladder in the second, the other team had slightly higher ladder, and I am awarded less ladder positions.


What is the point? The ladder is so bugged, it makes absolutely no sense. Shouldn't you gain more ladder from coming back from a 'deficit' position on the ladder?
And the loss of ladder to gain difference is so extreme there is no way it is functioning correctly.
17 Aug 2022, 07:56 AM
#2
avatar of Hannibal
Senior Moderator Badge

Posts: 3114 | Subs: 2

Your ELO was higher than the enemy team's. The ELO estimate expected you to win, but you lost. Therefore, you get a "high" penalty and downranked.

Second game, your ELO and the enemy team's ELO are more closely matched. The game is closer to 50:50, therefore you don't get as much of a bonus if you win.

Basically, everything working as intended.

The same thing occurs if your ELO was always higher than the ELO of the enemy team (which, if I get you right, probably was not the case). Then you get penalized for losing, but don't gain much from winning since you were expected to win in the first place. Also here, everything working as intended.
17 Aug 2022, 11:37 AM
#3
avatar of aerafield

Posts: 3032 | Subs: 3

Your ELO was higher than the enemy team's. The ELO estimate expected you to win, but you lost. Therefore, you get a "high" penalty and downranked.

Second game, your ELO and the enemy team's ELO are more closely matched. The game is closer to 50:50, therefore you don't get as much of a bonus if you win.

Basically, everything working as intended.

The same thing occurs if your ELO was always higher than the ELO of the enemy team (which, if I get you right, probably was not the case). Then you get penalized for losing, but don't gain much from winning since you were expected to win in the first place. Also here, everything working as intended.


This.

The only major flaw here with the CoH2 system is that it doesn't compare the ELO on a team-basis, instead it just compares your personal ELO to the enemy team. Which becomes a big problem if the matchmaker screwed up completely or couldn't find a good match for you.

If I as top 20 player get dumped into a complete bullshit 4v4 match with three rank 10.000 teammates and lose against four rank 5.000 enemies, my ELO gets nuked because the system simply saw it as "Wow how can a top 20 player lose against rank 5.000", even if there was 0% chance of winning.

I hope this changes in CoH3
17 Aug 2022, 11:48 AM
#4
avatar of Protos Angelus

Posts: 1515



This.

The only major flaw here with the CoH2 system is that it doesn't compare the ELO on a team-basis, instead it just compares your personal ELO to the enemy team. Which becomes a big problem if the matchmaker screwed up completely or couldn't find a good match for you.

If I as top 20 player get dumped into a complete bullshit 4v4 match with three rank 10.000 teammates and lose against four rank 5.000 enemies, my ELO gets nuked because the system simply saw it as "Wow how can a top 20 player lose against rank 5.000", even if there was 0% chance of winning.

I hope this changes in CoH3


I will chime in with my experience. I do think it compares on per-team basis.

I know that a lot of times, when I was paired with rank 5000 (so ~20 + 5000 + 5000) against rank ~top 400 I lost no ranks. Heck I remember times when I had 2-3 games in a row with such "perfect" MM and I lost 0 ranks. I was I think rank 8. Lost 3 games in a row. Horrible games, was still at 8 the next two days.

However, I do lose rank when I get a match such as:
Me (rank top 20 99% of time) + rank 2000 + rank 1500
VS
Rank top 100 + rank 1000 + rank 1000

I know in such games I usually lost around 5-7 ranks.

It leads me to conclude that there is some sort of per-team basis of ELO. Not just your own against the enemy, for each player.
I seldom lost ranks where the MM was an extremely fishy loss, and vice-versa. I seldom gained ranks by winning such games.
17 Aug 2022, 13:27 PM
#5
avatar of NorthWeapon
Donator 11

Posts: 615

Your ELO was higher than the enemy team's. The ELO estimate expected you to win, but you lost. Therefore, you get a "high" penalty and downranked.

Second game, your ELO and the enemy team's ELO are more closely matched. The game is closer to 50:50, therefore you don't get as much of a bonus if you win.

Basically, everything working as intended.

The same thing occurs if your ELO was always higher than the ELO of the enemy team (which, if I get you right, probably was not the case). Then you get penalized for losing, but don't gain much from winning since you were expected to win in the first place. Also here, everything working as intended.


Very good explanation. In other words, this ELO algorithm is not exclusive to Relic, and it rewards/penalizes based on probability of winning/losing
17 Aug 2022, 14:21 PM
#6
avatar of BasedSecretary

Posts: 1197

^^^Dumb replies by people with 0 experience in simple one-variable statistical models.

The ELO model is very outdated and few, if any, competitive games still use it. Valve is famous for being the first to ditch the favored ELO model back in 2014, followed by -I think- the makers of LOL.

ELO is fundamendally flawed because it "predicts" outcomes on a normal distribution basis rather than a on a more flexible system ie. logistical distribution. I have talked about this before.

I do not want to bore you down with details because it's pointless. The gist is that after 50 years of ELO system the research has shown that it cannot accurately predict games after a certain threshold (I think it was ELO1400 and upwards) and it either nukes you up or down without necessary deserving it.

Most games use a heavily modified ELO system called Glicko-II which artificially trends the game to logistical models or a system created by Microsoft which I have no idea how it works.

Couple that with the downright retarded implementation of vanilla ELO on COH2 (the weights are on the player and not on the team as a whole, meaning the game expects you to carry your teammates or get carried by them) you understand the result will not be nice.

PS: Fun fact - PUBG is the only game to utilize the very first ELO implementation for its competitive ranks.
17 Aug 2022, 22:20 PM
#7
avatar of Reverb

Posts: 319

Ok explain this ladder sink I just experienced:


I queue match, I have a rank 400 teammate, the other team is made of top 100 players (one top 50).

After about 20 seconds one opponent drops, the other drops about 5 seconds later.

My rank goes from 332 okw to 362. I just lost 30 ladder from a win.

I've dropped about 150 ladder positions since I reinstalled this game yesterday in 5 matches.
18 Aug 2022, 00:48 AM
#8
avatar of Rosbone

Posts: 2147 | Subs: 2

jump backJump back to quoted post17 Aug 2022, 22:20 PMReverb
My rank goes from 332 okw to 362. I just lost 30 ladder from a win.

This happens when you lost a game before that had not finished yet.
1 user is browsing this thread: 1 guest

Ladders Top 10

  • #
    Steam Alias
    W
    L
    %
    Streak
Data provided by Relic Relic Entertainment

Replay highlight

VS
  • U.S. Forces flag cblanco ★
  • The British Forces flag 보드카 중대
  • Oberkommando West flag VonManteuffel
  • Ostheer flag Heartless Jäger
uploaded by XXxxHeartlessxxXX

Board Info

1027 users are online: 1027 guests
0 post in the last 24h
11 posts in the last week
26 posts in the last month
Registered members: 49992
Welcome our newest member, xewiy33830
Most online: 2043 users on 29 Oct 2023, 01:04 AM