While it is true for many people that reach their true skill very fast, this isnt true for the bottom and top. The dependency will mainly help those. it isnt fun for people to first loose 20 games in 1v1 and after these 20 games again 20 for TGs as well. You can make the same argument for the top tier. They can just troll their opponents in their first games. For other players it should not really impact their rate in reaching their true skill in my opinion.
The main issue with the current dependency is that all ladders follow a completely different distribution, while this dependency only works if all ladders have the same distribution. That is the main reason why we currently get some strange result when someone with high rating on just one ladder switches to another ladder.
I think after fixing the inflation, all ladder will follow about the same distribution and thus this dependency would work fine and dont really need to change. It will just help the top and bottom tier to help reaching their true skill a little bit faster. It isnt really worth to change that at all.
I agree with the part of your post that i didnt quote.
You need to ask @GMEvangelos. Maybe there is no reset at all…
I did the team game analysis part. And here are the results Ladder |Games analysed | <40 rating difference | <100 rating difference 1v1 | 68737 | 85.9% | 97.8% 2v2 | 11692 | 70.2% | 95.3% 3v3 | 05382 | 72.4% | 95.8% 4v4 | 03672 | 70.5% | 94.0%
Here average ratings were used for teams. This data was taken from 1 week at the start of December. Only matches where all the players had a valid rating and rating_delta were considered. This excludes player playing their first 10 games. My observations in a previous post were from matches played over a 6 hour window. So this data should be more reliable.
From this it is clear that the matchmaking system tries to keep the average of a team as close as possible. It is doing a good job at it as well.
Some other statistics (1v1, 2v2, 3v3, 4v4):
Average standard deviation among ratings of all players in a match: 20. 51, 71, 90
Average difference between maximum and minimum: 20, 116, 193, 274
I blame the later mainly on the nature of premade parties.
Looking at the rating chart: https://ratings.aoe2.se/ I think it takes fewer than 10 losses to be at the true level for >99% of people. If look at pro players’ profile on aoe2.net you will find that most of them had 1600 ELO when they finished the first 10 games. Bottom players had around 500. And 500 ELO is already the bottom 1%.
BlockquoteCan someone tell us the calculation of ELO in teamgames? First thought would be it will be the same as HD and Voobly (see youtube against for that calculation), but this is not true.
the most common way team game elo is calculated is by taking the total of both players
Player 1 3000 ELO
Player 2 1500 ELO
Team ELO: 4500/2=2250
then you hit up search and the game will try and find a match with another team around the 2250 ELO mark.
This is how you get these uneven matchups. At least that’s generally how I’ve personally seen team game matchmaking to work, not sure if it’s working like that in this game.
Now if we look at team 2 for example
Player 3 - 2300 ELO
Player 4 - 2200 ELO
Team 2 ELO: 4500/2=2250
With this we have based on team rating alone a ‘balanced team matchup’ but based on the individual elo, that Player 1 would most likely accel at the game, while their team partner Player 2 will struggle, the Team 2 with player 3 and player 4 will perform both about equally, but still worse than player 1, now the question is if they notice that and double either player 1 which has better skills or completely demolish player 2 which has lower skill than both. If they would play it out as a normal 1vs1 one has a much easier time versus the opponent than the other.
Seems like i never reacted to this stats. I do think this is kind of what i would have expected. No real surprises to me. For me just shows that premades arent really the issue. The main reason why TG sucks at this moment is still the bug in the calculation. The devs dont need to compare your elo to the maximum of the enemy, but to the average elo of the enemies.
I am still waiting on a respons from the devs. This thread is already opened 10 months ago. It wasnt the first thread about this subject. It wasnt also the latest, because many other threads followed. It is really a shame that the devs completely ignore this issue. I really hope that the devs will look in this issue quickly. @GMEvangelos@BBQTurkman
Yes.
The problem is not that the matchmaking doesn’t match us with opponents that have a similar ELO but that the ELO is inflated for people who play more.
In my group of friends the one that plays the most teamgames has a 200+ higher ELO then another who is actually a little better and has a higher winrate (and would beat him without a sweat 1v1).
We all played several hundred Teamgames. By now the ELO system should have found an accurate position for all of us but it just keeps on changing (increasing…).
also the matchmaking should stop spawning games that have hundreds of elo difference between teams. i’ve played hundreds of these by now and they’re pretty much all a waste of time
Of course I can’t look inside of the developers heads, but I will try anyway to hypothesize how we got to this situation.
Probably the developers were thinking: ‘If we use the average rating in our calculations, that can be really unfair when e.g. a 2k player and a 1k player match up against two 1.5k players.’ So they decided to let rating gain/loss be dependent on how your rating compares to the highest rated player on the other team. Now in actuality, they may not realize it but their solution did literally nothing to solve the issue it was intended to solve, as the actual MATCHMAKING system still looks at average Elo. The only thing the solution did was introduce an additional problem: Rating inflation.
Developers have the tendency to want every change to be a strictly positive change, and to never go backwards. They may be thinking: 'Yes, @WoodsierCorn696 s solution would solve the problem of rating inflation, but it does not address the initial reason why we have the current system, and so his solution may worsen the problem of people of wildly different rating teaming up (even though they would be incorrect in thinking so).
I bet the developers only want to change the current system if they come up with a new system they are fully satisfied with, and they cannot afford the time of thinking about this problem for too long.
By the way, I don’t even know if the Elo inflation itself is the biggest cause of unbalanced team games. The current system has a load of other side effects that seriously imbalance team games, such as experienced by the OP of this post: Team elo is not working
I myself have been the unwilling cause of imbalanced teamgames… I have played lots of 4v4s with three friends. When we started playing together, our team Elos were quite different, but after a while they became almost the same (the high rated players in our team would get few points for winning, and the low rated players many, and vice versa every time we would lose). Last weekend, I played seven 2v2s with one friend, and we are the strongest members of the 4v4 group. Because our rating undervalued our skill, our matchups were too easy, and we won all seven of our games… Previously I wrote that this current system probably was designed to mitigate the problem of people of different ratings teaming up, and I said it did nothing to solve the issue, but that was an understatement: it is making it worse.
The dev can read this thread for ideas on how to solve those issues as well. People already shared ideas on how to deal with unbalanced teams (a 2k player vs 1k player vs 2 1.5k players).
This is also the result of their decision to compare your elo to the highest elo on the enemy side. This wont happen with HD or Voobly elo. On HD and Voobly there was a calculation for the whole team, comparing the average elo from 1 team with the other team. Then as second step, that amount of elo was distributed over the players, pretty much equally.
So most of the issues with the current TG rating will have its roots in the bugged calculation. At least, i see the calculation as bugged. I cant see the minds of the devs as well. I wish they shared their vision with the players. Currently we just hear nothing from the devs. They seems to not really respond on bug reports, they dont really respond to ideas, you hear nothing from them.
I will already be happy if they come with a statement about how they see the current TG ladder/ratings. But currentlly we dont know anything about their plans.
Yes, they should. I was hypothesizing why they don’t, though, and maybe it’s because they just don’t want to commit the time. But truthfully, I don’t know.
Oh, I agree with you. I am just saying that the problems it causes to teamgames are not only due to inflation.
Probably going to add insult to injury, but among all the broken things, this system makes it very easy to boost your friends by simply queuing together. For example if a 1500 Elo queues in duo with a 2500 Elo, the 1500 Elo player gains around +140 for a win and -1 for a loss.
It makes it difficult to not boost your lower Elo friends by queuing together and to not tank your own rating, too. I have done what you describe, but it is not like I want to cause inbalanced games…
From my experience the last part is wrong tho. You lose very little Elo by queueing with someone lower than you. For me it’s around +25 for a win and -30 for a loss.
One solution would be to avoid partnering up for ranked play when the gap between players is too wide, to avoid the “average ELO” of the team being a math equation that doesnt really represent any player. 1x 2000 and 3x 1500 should beat 4x 1625 any day.
The selected method of rewarding based on highest ELO is even worse than using average ELO. Above still happens AND the current system rewards victories OR playing a lot. You can find several 2200+ players with 40% win rates.
So the devs announced today that there will be a new civ. Did they also mention something about this issue? Or are they still silenced about this issue? @GMEvangelos please react to this thread about the point of view of the devs concerning this issue. It is already an issue from the release of DE. Until now the devs seems to just ignore this issue. I have seen a statement about this issue at all. The bugged team game ladder is killing all the fun for playing team games at the moment.
I do not know any of the details regarding this issue, but I can confirm that it is being discussed internally as an issue in need of a fix. I cannot say what or when something will change in the game (yet), but can say that your concerns have not fallen on deaf ears.