Login

russian armor

Addressing the issue of ELO-hell.

PAGES (13)down
18 Feb 2022, 15:23 PM
#202
avatar of GachiGasm

Posts: 1116 | Subs: 1



The lower you are on the ladder, the more probable it is to be matched with a higher skilled opponent. At the same time, this loss will result in a lower ELO adjustment.


It actually depends. Without taking into a considiration super low ranks, if you have multiple opponents simular to your rank, but among them there is a player with a significantly highter rank then yours, you still will lose significant amount of ELO. The same way you will lose significant amount of ELO if you are paired against lower ranks, having simular low ranks teammates.

But it actually would be really insteresting, if some one made a research of how actually ELO is calculated in a match with multiple people being much higter\lower rank then the others.

I even think that, there might be even some sort of way system tries to balance the match via matching differently ranked players. Because its a common situation where, for instance, top 100 player could be matched with multiple rank 700-1000 players, against full house of ranks 300-400 players.
18 Feb 2022, 15:39 PM
#203
avatar of Hannibal
Senior Moderator Badge

Posts: 3114 | Subs: 2

It actually depends. Without taking into a considiration super low ranks, if you have multiple opponents simular to your rank, but among them there is a player with a significantly highter rank then yours, you still will lose significant amount of ELO. The same way you will lose significant amount of ELO if you are paired against lower ranks, having simular low ranks teammates.

But it actually would be really insteresting, if some one made a research of how actually ELO is calculated in a match with multiple people being much higter\lower rank then the others.

I even think that, there might be even some sort of way system tries to balance the match via matching differently ranked players. Because its a common situation where, for instance, top 100 player could be matched with multiple rank 700-1000 players, against full house of ranks 300-400 players.

I agree here. I only know ELO systems very superficially and we also don't know how exactly Relic implemented their system, but shouldn't the ELO adjustment be made based on the team's winning chance?
Say you're a top player with only low skill players in your team vs a team of "normal" opponents, giving you a calculated win chance of 50%, shouldn't the ELO rating after the game adjust as much as if this was a match of "worthy" teammates and opponents? It shouldn't take individual ELO into account and punish the top player overly for losing the game.

Asking out of curiosity here, because this is how I always understood the system.
18 Feb 2022, 15:54 PM
#204
avatar of GachiGasm

Posts: 1116 | Subs: 1


I agree here. I only know ELO systems very superficially and we also don't know how exactly Relic implemented their system, but shouldn't the ELO adjustment be made based on the team's winning chance?
Say you're a top player with only low skill players in your team vs a team of "normal" opponents, giving you a calculated win chance of 50%, shouldn't the ELO rating after the game adjust as much as if this was a match of "worthy" teammates and opponents? It shouldn't take individual ELO into account and punish the top player overly for losing the game.

Asking out of curiosity here, because this is how I always understood the system.


I belive thats the one of the core problems with ELO here. Because the end result of ELO win\loss is based on an individual player rank. It isnt taking into a considiration whether you was paired with low\high rank players. If you are top player winning game against low ranks with low ranks teammates, you gain almost no ELO, yet you are loosing it as if you lost to a low rank.

I do belive that ultimately, system treat all games (except the ones which were obliviosly sabotaged by unability to find proper opponents\teammates) as if they have 50\50 chance of winning. Either that, or it was just a huge oversight or copy paste code done by relic in terms of how teamgames ELO works.

Since there are some signs of MM trying to balance the game rank wise, even if its unable to find players of your rank, but at the same time when it comes to ELO there is litteraly nothing adressing such match ups.

At least this is the idea I got from my observations.
18 Feb 2022, 15:58 PM
#205
avatar of T.R. Stormjäger

Posts: 3588 | Subs: 3

Everyone has yet to mention how queueing for a long time in 4v4 puts you vs anyone queueing without considering ELO anymore, leading to super unbalanced matchups in terms of rank.





"There is literally no problem with MM, just get good!"
- Karl "Cringelarp" Marx when called in to solve lelic's finances.



Do you have anything constructive to add or is acting like a clown all you know?
18 Feb 2022, 16:08 PM
#206
avatar of BasedSecretary

Posts: 1197


I agree here. I only know ELO systems very superficially and we also don't know how exactly Relic implemented their system, but shouldn't the ELO adjustment be made based on the team's winning chance?
Say you're a top player with only low skill players in your team vs a team of "normal" opponents, giving you a calculated win chance of 50%, shouldn't the ELO rating after the game adjust as much as if this was a match of "worthy" teammates and opponents? It shouldn't take individual ELO into account and punish the top player overly for losing the game.

Asking out of curiosity here, because this is how I always understood the system.


Excellent point.

The ELO system, in and of itself, is just a way of removing x amount of points if you lose and you are the top dog and y amount of points (x>y) if you lose and you were the underdog. The thing is however, that the ELO system makes mathematical sense only if you can prove that the performance of every COH2 player can be modelled by the Normal "Bell Curve" Distribution. A further assumption is necessary because COH2 performance in the above sense is still not measurable. One cannot look at a sequence of clicks, orders or doctrine choice and derive a number to represent that player's skill. Performance can only be inferred from wins and losses. Therefore, if a player wins a game, they are assumed to have performed at a higher level than their opponent for that game. Conversely, if the player loses, they are assumed to have performed at a lower level. To simplify computation even further, Elo proposed a straightforward method of estimating the variables in his model (i.e., the true skill of each player). One could calculate relatively easily from tables how many games players would be expected to win based on comparisons of their ratings to those of their opponents. The ratings of a player who won more games than expected would be adjusted upward, while those of a player who won fewer than expected would be adjusted downward. Moreover, that adjustment was to be in linear proportion to the number of wins by which the player had exceeded or fallen short of their expected number.

Now here is where the retardedness begins in COH2.org: Normal distribution cannot accurately model the performance of lower ranked players. Every statistical organization worth its salt nowadays uses logistic distribution to model the performance of games. Even FIDE does (the chess cup).

Nobody uses the original ELO system anymore, because contrary to what most people here believe, "get good" cannot overcome basic statistical inadequacies. Most organizations use modified version to weigh heavily on ranking using something called the K-factor.

All that said, I only have 2 questions to relic:
  • What kind of modification of ELO is COH2 running?
  • Can we make any modifications so as to allow COH3's matchmaking to be more fair?



18 Feb 2022, 16:50 PM
#207
avatar of Rosbone

Posts: 2144 | Subs: 2

Since we are some what in agreement that ELO Hell is most prominent in 4v4:

Currently there are about 7800 players playing Coh2.


PLAYER COUNTS 4v4
Using COH2 SPECTATE I can see 7 pages of 4v4 matches. Each page holds approximately 19 matches.

Current 4v4 players ≈ 1064
Current players: 7818
Ratio: 1064/7818 ≈ 13.6%

So worst case players in queue could be (can check later):
Lowest player count time players ≈ 3000
3000 * .136 ≈ 408 players
408 / 8ppm = 51 matches

Just a data point. Nothing significant in this data yet.

PLAYER COUNTS 1v1
Current 1v1 players ≈ 190
Current players: 7818
Ratio: 190/7818 ≈ 2.4%

AVERAGE GAME TIMES
From PagePs site, we see most game modes have similar average game times at around 25 minutes.
18 Feb 2022, 23:30 PM
#208
avatar of Klement Pikhtura

Posts: 772

Guys, lets just agree that 2v2 is technically the most superior mode in the game and go home :clap:
19 Feb 2022, 03:39 AM
#209
avatar of Grumpy

Posts: 1954



The lower you are on the ladder, the more probable it is to be matched with a higher skilled opponent. At the same time, this loss will result in a lower ELO adjustment. I am not a computer scientist, but shouldn't the ELO system fix this scewed distribution by estimating the win chances before the game? I always though that this is how it works.

I think the way that the match maker works is that it tries to find games for you with people within a certain range. My guess is that it's something like 0-30, 31-100, 101-300, 300-1000, 1001-3000, You'll get matched quickly if there are enough people in your range. If you don't find a match within a certain length of time, it starts opening up the range. It seems like it is more focused on making sure that people can get games instead of waiting forever. It's a somewhat valid point, as waiting 30 minutes to find a game would bother the legions of casual players more than the mismatches.

I have had active ranks from 17 to 3000+ at the same time. On the 4v4 team that was ranked in the top 20, if we got a match right away then it was another well-ranked team. If the search went a long time, we knew we were going to get worse ranked players.

I don't know how ELO gets adjusted after a match. I don't know what my actual ELO number is, I just see the change in ranking from one game to the next so I can't tell how it makes the changes. I'm sure there is an algorithm but there are too many variables to guess at how it works.

I do think it would really help if Relic introduced Ranked and Unranked like many other online games have. I'd also like to see it so that in order to play ranked, you would have to beat at least a hard computer in a 1v1, or something to that effect.

The other improvement that I'd like would be to give new teams a elo rating based on either the average elo of each individual or their best ELO.
19 Feb 2022, 06:30 AM
#210
avatar of Esxile

Posts: 3602 | Subs: 1

There is flaw in your reasoning, both players should have wait enough time to be matched together. But that's not the case, I've been matched really quickly against players that were not close at all to my ELO ranking. I recall having been matched vs dutchman once in like 10 seconds while I was myself around rank 200.
So maybe he waited 2 or 3 minutes before getting an oponent with the searching range increasing but what about me, the other player?
And that's just an example I have in mind, but it happens regularly.
19 Feb 2022, 13:55 PM
#211
avatar of Grumpy

Posts: 1954

jump backJump back to quoted post19 Feb 2022, 06:30 AMEsxile
There is flaw in your reasoning, both players should have wait enough time to be matched together. But that's not the case, I've been matched really quickly against players that were not close at all to my ELO ranking. I recall having been matched vs dutchman once in like 10 seconds while I was myself around rank 200.
So maybe he waited 2 or 3 minutes before getting an oponent with the searching range increasing but what about me, the other player?
And that's just an example I have in mind, but it happens regularly.


I don't think the wait has to be two-sided. In your case, just Dutchman might have been waiting several minutes. When I played 1v1's, I sometimes had the same experience. When my ranking would go below something like 300 is when I'd get those matchups. Those could be a little deflating when it became obvious that not only was their micro better, they knew the maps better and I found myself being constantly outpositioned.

Also, the 1v1 matchmaking has a greater challenge than other modes, particularly 4v4, because of the low number of players. I play in the Pacific time zone (western United States) and the numbers in 1v1 matchmaking would sometimes bounce around from 38/62 to 43/57, etc. When I would see that, I'm pretty sure that there is less than 10 people searching 1v1 and the ratios are going from 3/8 to 3/7 (which is a lot more likely than having a lot of players and the matchups just happen to create those ratios that simplify to single digit numerators and denominators.
21 Feb 2022, 06:47 AM
#212
avatar of Rosbone

Posts: 2144 | Subs: 2

Is this supposed to NOT make people delete the game?



3 separate days of playing on the same maps :thumb:

The server is in fine shape. Nothing to see here.



If only I was as inept as Relic. Then I could just live on oblivious to how bad they really are. :guyokay:
21 Feb 2022, 09:49 AM
#213
avatar of MassaDerek

Posts: 197

jump backJump back to quoted post21 Feb 2022, 06:47 AMRosbone
Is this supposed to NOT make people delete the game?



3 separate days of playing on the same maps :thumb:

The server is in fine shape. Nothing to see here.



If only I was as inept as Relic. Then I could just live on oblivious to how bad they really are. :guyokay:


Stop playing 4v4 if you want to stop raging, there's just way too many players for that mode to ever be properly balanced or properly matched with even ranks.
21 Feb 2022, 13:37 PM
#214
avatar of BasedSecretary

Posts: 1197

To be fair, it's at least laughable trying to find a decent matchup in 4v4.

The game mode, maps and overall mentality just do not allow for it. No tactics, no strategy, just pure old fun. Most of the times I go for 4v4 is to either troll or laugh my ass off.

The real problem for me is the fact that we have no "rank reset" mechanic, kind of like seasons for LOL and DOTA which allow for faster upward mobility of players. I would gather that this is one of the reasons MM is in such a bad taste, keeping stats that were valid maybe 5 years ago.

In COH3 let's see at least a season system. That would be a good start. Also, let's also consider making the MP at some point F2P so that more (many more) people could join the low ranks and allow for better and larger pool.

Most successful multiplayer games are F2P and most of them are not even P2W their business model is selling skins and weapons. I think overall it would be a good choice.
21 Feb 2022, 15:47 PM
#215
avatar of Protos Angelus

Posts: 1515

You seriously need to stop raging, it's bad for your health. You're playing 4v4:

Population max: 800
Map size: Slightly bigger than 2v2s, definitely not proportionally bigger compared to population --> high pop density
# of players: Fu**high
# of cretins: Fuc*ton
Micro intensity: Either extremely low or extremely high
# of tactics/strategies: low

Heck, last game I was playing against some top 50 players and was trading nicely, overall all 3 of us were doing well, having better territory control, etc.
About 24? minutes in, the calm before the storm. 3x pgren terminators with G43s, A-move raketen (donated from okw player that got ra*ed by my team), 5 vet2/3 grens with LMGs, 2x pgrens with shrecks, 2x flakpanzer, panter, 2x werfer strikes and a brummbar all came rushing down on top fuel/cutoff of Winekkendonk ( I had bottom left spawn ).

I lost almost everything. A massive 2v1 blob charge. Mines? Triggered by brummbar explosion or werfers. 2x G43 pgrens literally wiped a vet3 ranger in yellow crater in <2 seconds coming from the corner. Most people would rage and leave. I just watched how my team countered them and completely surrounded and 5 minutes later they surrendered.
That is common in 3v3s. Overextending and getting trapped.
The trick is to not rage and flame, because unless you're playing with newbies, if you're getting ra*ed by 2 guys, that means their 3rd guy is getting ra*ed by your teammates. You just need to keep playing..... In 4v4s, that extra one player on each side means that there is a greater probability that in such scenarios, they will just straight up leave if they lose a lot of units. In 3v3s, that's less common, people tend to play on.
Moral of the story: Don't expect 4v4s to have 3 good teammates in randoms. One or two are possible, but all 3? Nah.
I had 2 great random teammates in my last game and despite the losses, won. That sh** doesn't happen in 4v4s. Not only are mega wipes more common (lots more werfers and lefhs and stukas and katys), but having more players, the chances are higher than one of you will leave.
21 Feb 2022, 18:15 PM
#216
avatar of Rosbone

Posts: 2144 | Subs: 2

Uhhhh did anyone look at the pictures or read what I wrote? I dont think you did.
21 Feb 2022, 19:55 PM
#217
avatar of Yourcall

Posts: 40

jump backJump back to quoted post18 Feb 2022, 16:50 PMRosbone

AVERAGE GAME TIMES
From PagePs site, we see most game modes have similar average game times at around 25 minutes.


We have to keep into account that you single handedly drag down this average gametime by about 5 minutes.
21 Feb 2022, 20:05 PM
#218
avatar of Rosbone

Posts: 2144 | Subs: 2

We have to keep into account that you single handedly drag down this average gametime by about 5 minutes.

Why thank you. Real recognize real.
22 Feb 2022, 23:55 PM
#219
avatar of Grumpy

Posts: 1954

jump backJump back to quoted post21 Feb 2022, 18:15 PMRosbone
Uhhhh did anyone look at the pictures or read what I wrote? I dont think you did.


I did but don't know what to make of it. I sometimes get streaks like yours. What are you using for vetoes?
23 Feb 2022, 05:32 AM
#220
avatar of Rosbone

Posts: 2144 | Subs: 2

jump backJump back to quoted post22 Feb 2022, 23:55 PMGrumpy
What are you using for vetoes?

Essen Steelworks - Too tight in city area. RNG shitfest.
La Gleize - Too big and weird fences everywhere.
Lorch Assault (West Wall) - If I get bottom I /L.
Vielsalm - Too big and wonky city fighting.
City 17 - Worst 4v4 map ever made by far.



Total Games: 40177

Games on Nordwind: 2854
Odds to get Nordwind: 2854 / 40177 = 7.1%

Games on Lienne: 3905
Odds to get Lienne: 3905 / 40177 = 9.7%

So what are the odds you get Nordwind 3 games in a row?
So what are the odds you get Nordwind 5 out of 6 games in a row?
So what are the odds you get Lienne 3 out of 5 games in a row?
PAGES (13)down
0 user is browsing this thread:

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

412 users are online: 412 guests
1 post in the last 24h
7 posts in the last week
39 posts in the last month
Registered members: 49063
Welcome our newest member, jennifermary
Most online: 2043 users on 29 Oct 2023, 01:04 AM