In this case, please describe the bug to me, so that I will know about it.
In almost every match (excluding very rare cases where all players have almost same elo), winners gain more elo points than losers lose. This leads to massive inflation and to players climbing higher when playing more games regardless of their skill (as is illustrated by the player having high elo despite losing most of their matches).
I recommend going to aoe2.net and sorting both 1v1 and TG ladders by number of games played (to exclude players who have not played enough games yet). You can see that in 1v1 ladder, everyone except the best players who often canât find equal opponent have 48-52% winrate, while in TG ladder players have winrates all over the place.
Itâs not elo. Elo in rating context refers to certain equations.
I would call it just âratingsâ or ârating systemâ.
I my opinion it should never be the case that the rating points of all participants of a match are not conserved.
This is too simple to fix in order to get rid of possible problems.
@RookyJupiter539 can you link a game, were this is the case?
Well in TGs the winrate is all over the place because of the factors i mentioned (premade teams, alf-f4 mapdodging, maps specialists, ratings are not adjusted yet/numbers of games played)
1 Like
TG elo are inflated. As result you gain elo for just playing the game. Not for being a better player. As result elo doesnt mean much for TGs.
They messed up the equations. Based on your response i assume you know the equations of for the elo rating. What the devs for TG elo is that they compare your elo to the highest elo on the other side in the equations. As results:
If you win:
Your elo is probably lower then the highest elo of a enemy. So you arent expected to win based on the theory behind elo. Thus you are awarded by extra points.
If you lost:
Your elo is probably lower then the highest elo of a enemy. So you was already expected to loose based on the theory behind elo. As result you loose little elo.
The most easy fix is to just compare your elo to the average elo of the enemy and it already will be much better.
HD and Voobly even take a different approach: They compare the average elo of both teams to determine the total gain / lost for each team and then just divides that number equally between all players within a team.
Both solutions will work. In that other thread even some more complication solutions are posted, that also deal with premade and things like that.
1 Like
11111111 this is so crooked
It was obvious to me that the rating system is not an Elo one.
It doesnât have to an Elo system, but the current system is so bad.
Everything is said in the thread you referred (Analyses of the ratings - Spotting the issues)
Itâs a known issue for more than a year.
No misconception mate.
The TG ELO system is broken and it rewards playing as much as possible to climb. Any proper system balances so that a player reaches 50% winrate after a certain amount of games. To have 39% and still be climbing, means the system is crap. The fellow I posted has 2k games⊠No way he has played 600 of them against high level teams, right?
Only 5% of the playerbase has under 1000 TG ELO, which means points are being created from thin air (lots and lots of points).
1 Like
Itâs not an ELO system, so you canât blame it for not working like an ELO system 11
Not having the rating points of all players conserved doesnât mean that your leaderboard position canât be determined.
You cannot climb arbitrary high with your leaderboard position, even if you would play a lot.
It is true that you need a lot of games to get to your actual rating, which should be changed.
But I would not call it broken because of the fact that there is inflation.
Itâs just that the number of games you need to play to reach and keep your leaderboard position (the inflation) is probably way too high.
I guess that is what you are referring to.
And even in an ELO system for more simple matchmakings like in chess you can have a below 50% winrate and still maintain your ELO rating (in case of aoe2:de this would be climbing, because of rating inflation). I donât see why you are so stubborn with your alt-f4 because of your teammates winrate 11
The system used is a slightly adjusted Elo system. So naming the system an Elo system is perfectly fine to me.
For 1v1 it is just the plain Elo system. The K-value depends on your rating and the number of games which you have played. In you first +/- 10 games the K-value will be higher, so you will gain or lose more points. After those initial games your K-value becomes around 30-32. After a certain rating (i dont know the exact value) the K-value drops even lower. But this mainly has an impact on the pro scene. All of this is pretty much in line with chess as well, where the K-value also depends on some rules:
Elo isnt meant to be a team game rating system. So they need to adjust the rating system to replicate some 1v1. On one side of the equation you have your own rating. On the other side of the equation they picked the max rating on the other side. Then they uses these numbers like it was a 1v1 and apply the Elo calculation.
Based on this info i am not sure why we cant call this system an elo system. It just follows the rules to calculate a rating based on the Elo rule set. So to me the system used to calculate the ratings is the Elo system. No doubt about this.
Btw, you write it is Elo, not ELO. Elo isnt a abbreviation, but it is name after a person. So only the first letter is a capital, not all letters.
Currently this seems pretty possible, because of the inflation. Because the system uses the max Elo of the enemies, it is pretty common to gain 20-30 elo for winning a game, while just losing about 5 after losing a match.
Match making makes sure teams are pretty equal, based on their average rating. So based on that, you expected to win or lose half of the games. But because the system uses the max Elo of the enemies, it is pretty common to gain 20-30 elo for winning a game, while just losing about 5 after losing a match. So you can even have a winrate about 20% against equally skilled teams (on paper) and still climbing the ladder.
This is all the result of picking the max rating of the enemy. The inflation is pretty much real and even some big noobs tend to be 2400 elo for TGs. That is all because of the inflation as result of the bugged Elo calculation.
Because TG Elo is meaningless as result of the bug, win rate is the next best thing. So people look at the win rates of their ally and determine if they want to play with these players based on their win rates. Low win rates with many games means in most cases a trash ally. And people dont want to play with such ally.
I can see why people look at the win rates and i understand the conclusions. Still i think Alt+F4 in that case isnt really good behaviour. I considered Alt+F4 as toxic behaviour.
1 Like
You are missing the entire point of the argument. There ISNT an âactual ratingâ on TGs because inflation breaks it.
A 1000 1v1rating player can play 3000 games and get to 2400 TG rating with 39% win rate.
A 1500 1v1rating player can play 300 games and get to 2400 TG rating with 55% win rate.
Which one is correct? They are both 2400 but are they evenly skilled?
The whole purpose of the rating system is to help establish balanced matches. Whilst individually there is no way to rectify the system, at least one can avoid wasting time on a match that would never happen if the system was working properly.
1 Like
You are getting there, itâs a lot about the numbers and rate of games to play in order to reach your actual rating.
Itâs just way too many.
Iâm too lazy to run a simulation on how fast the convergence is, also there is the information missing on how the match participants are set together.
And the TG rating is not an Elo system, because the points are not conserved and the ratings inserted in the Elo equation donât represent the team as a whole.
Itâs like applying Newtonâs equations for small actions and still call it Newton mechanics. It doesnât work.
The thing is that first i couldât understand you, because you are mixing overall winrate, Elo system, point inflation all together.
Lets approach this from another angle. I canât seem to figure why the devs donât want to fix this. I believe this has been raised many many times. Is it because people enjoy team games and higher ratings and so if people play more teams games, they will have higher rating and in turn play the game more? I think there are problems with that way of thinking. Because while I enjoy playing team games and higher rating, I also enjoy winning. With the way the current rating system is setup its really hard to win after a certain increase in rating because of skill level. I started playing this game playing team games with my friends. I lost a lot initially so my win rate was bad, but then as I played more my win rate has no chance of improving or even staying the same, as my rating keeps going up and that isnt exactly tied to increase in skill level. So at the end of the day, it leads to bad experience for everyone and turns people off the game. Oh and on top of that I decided to play 1 v 1 after playing TGs, and my rating was 1500-1600 after my first game⊠and then i had to lose around 10-15ish 1 v 1 games before I was even close to finding a competitive game⊠and now my 1 v 1 win rate is terrible as well⊠lets just fix this please devs, you will have wayy more people playing the game and enjoying it as well!
3 Likes
The problem about the current TG elos is that there is no guarantee that the opponent team, which has a similar elo compared to your team on average, will have a similar skill level you. Itâs totally random what type of opponents youâd have. Youâll either have those players with many team games and low win rates/low 1v1 elo (low skilled players) or with few team games and high win rates/high 1v1 elo (high skilled players). Based on your opponent you might have a very easy or hard matchup. Because of this randomness team games do not provide a competitive environment.
1 Like