Matchmaking should be done by both ELO AND WIN RATE, inconsidering the ELO system has been FIXED

Hi All who will read this and hi to all of you who will feel that rage inside you once i remind you of the problem AOE2 DE faces today.

This is mostly regarding RM TG ELO as thats the one that is disproportionately out of sync with the reality.

I dont remember whats is the ELO that a new player starts with but i think its 1000 and 1000 right? Correct me if im wrong.
First issue is that a 1000 RM player will correspond to around 2000 RMTG not 1000.
I.E. in a 4v4 game this might happen: 2200,2200,2200,2200 vs 2600,2650,2700,1200 (low elo with 90% win rate).
The type of matchmaking that just wasters 30 minutes of your day and leaves you a little… angry at it.

The 90% win rate then leads to the SECOND ISSUE.
In most of these games that waste your time the “low elo” player usually has around 20-40 games already registered.
With a 80% + win rate, how poorly developed the formula has to be in order to still not have that player at the correct ranking?
That is just mediocre work on the developers side.

I come from chess.com as many of you might as well. (You might see where im going with this).

  • First suggestion: Give options.

When you create an account, give 3 ranks to start with, 800 for beginners, 1200 for interms and 1500 for players who have played hundreds of games in other accounts or platforms.

  • Second suggestion: Improve the damn formula.

Elo adjustment should have a variable in regard to streak and the amomunt of games a player has.

10 games with 9 wins should be enough to get your elo high enough for your rank.
Think about it, players with 29-1 raking 1600-1800 rmtg… how many days have they been playing to get there and its still easy as heck for them. Imagine playing for 10 straight days, 3-4 games daily and still not catching up.

  • Third suggestion: Win Rates

Dont just balance by elo. A team full of 40% win rate vs a team with 60% win rate is not balanced wouldnt you think?
Specially considering that 10% winrate +/- is a huge deal in this game, you dont see playters with 30% win rate or 70% win rate with a thousand games played too often do you?

Balancing should be done by both ELO AND WIN RATE, inconsidering the ELO system has been fixed.

Now for anyone interested i can updoad screenshots of teamgames with this issue which have been the majority in the past week for me, its incredible.
Many of these games i’ve streamed on Twitch so that can server as proof as well.

Thanks to the readers, please to the developers.
Oleksiy.

1 Like

You can’t pair people by winning rate, that would cause a huge mess, i would be facing lierey in 1x1, cause i don’t play that much but our winning rate is similar, see it doesn’t make sense and it would never find fair matches.

The devs recently changed the calculations, but they still arent perfect and they still cleaned up the mess of the previous calculation.

Elo was inflated in the previous calculation. As result the average moved slowly upwards. It started by 1000 elo (which is also the startin elo for that reason), but slowly increased to about 1500-1600 as a bug in the calculation.

Now the made a fix, so the system isnt inflated anymore, but there are still other exploits (mainly with smurf accounts). Also this fix will stop the inflated, but they mess of the previous calculation is still present. The devs can improve their formula, but if they wont clean up their mess, nothing will really help to improve the situation.

For more info about all these issues of the past, see Analyses of the ratings - Spotting the issues.

That being said, let us have a look at your suggestions.

New players dont really understand what to pick, smurfs will probably pick the lowest, so they can smurf even more. I dont see any benefits from this. I also have no idea this will improve the match making. You will gain or lose already more elo in you first games. So your elo will be adjusted quickly after the first games, no matter what you pick. I have no real hope that picking a rating at the start will help at all.

Inflation on the ladder is really messing up this one. Normally you start at average player on the ladder and you will move up or down based on your first games. But due to the inflation you start at 1000 elo, which is about bottom 5% or something like that. As result you have to climb over 45% of the ladder before you are at the normal starting point. This is one of the reason why just fixing the calculation isnt the solution. The devs really need to clean the ladder.

I can see improvements in how much elo you win if you are on a winning streak. In the thread that is already mentioned earlier (Analyses of the ratings - Spotting the issues) i also suggests using TrueSkill or Glicko-2 as rating system over Elo. It looks to me these systems will improve the rate to catch up. They look more flexible systems.

Winrate shouldnt matter. Match making will make sure that everyone will be around 50%, if it functions well. Only exceptions are really the real top and bottom of a ladder. These high winrates are the result of the mess of the past. This is again a reason why i think the devs need to clean up the mess at the TG ladder. I would even go that far as just resetting the full TG ladder.

Hi Woodsier

My post was based on the hope that the community in its majority are not smurfs and seek good matchmaking as opposed to OP teams and inflating their scores.

Im not sure how many smurfs are there in relation to the ammount of players, i didnt know there was an actual “documented issue” (if you will) with smurfing.
I agree that if in the first few games your elo would adjust rapidly the different starting points have no use.
But in the last game before i made this post i played against a 1434 13W/1L 1646 27W/1L guy who was in a lobby/teamed up with 2.8k rmtg rated players. For refference my team was 2.2 with 100 deviation.
As you saw lower in my post i mentioned that taking into account the W/L, the streak and how few games a player has in order to adjust faster for his elo could be improved. But it clearly not yet done as 28 games should be way more than enough.

Completely agree with cleaning up the ladder, fresh start using another rating system or even improving their own. Win rate shouldnt matter, also agree, but as you said, ppl should be close to 50, not 100.

Do you know if they have considered just raising the baseline TG rating while they come up with a solution? For new players to begin at 1700 i.e.?

Heres some pics for reference, past 24 hours

elo is a flawed system
it takes hundreds of games to move people to where they belong

Currently the elo change is based on the average elo. So if you have a lower rating smurf in your team, then you will in the first place get more easy opponents, and have an higher change to win. And everyone on the team gets the same elo. As result some people let friends use a smurf account, so they can boost the main account elo.

https://aoe2.net/#aoe2de-leaderboard-rm-team

If you see the leaderboard, then you see one player having more then 500 more then the number #2. He is really exploiting this smurfing issue.

Based on the pictures i would most likely think you are playing against a smurf account, so other accounts can best their rating. I say most likely, since you cant really be 100% sure. So i wont guaranteed these are smurf accounts, but it is really likely they are. Both teams look like a 2.7k team, that is rated as 2.2-2.3 because of the smurf account. Easy win for them. Note: This is just my conclusion based on the pictures you showed. I havent really looked into their profiles.

Smurf will break almost every system, so you dont really need to look at those accounts.

The issue is that the ladder is a completely mess because the previous calculation. So unless you somehow clean that up (and i think a reset is the only real way to clean the mess), you will still have issues with unbalanced teams. Having a higher starting elo doesnt really matter.

If we start over, then i would assume most of these 2.8k players might be around 2k and 1.6k is already much closer, if the average is back at 1000 elo. So this will be a smaller issue.

You might even increase the elo win / losses for the first games, so people go up or down even more quickly. That should be the best reason to even nullify this issue and boost these players to the top even quicker. Or the devs have to stop with elo and use a system like TrueSkill or Glicko-2. Based on the theory, both are better for a quicker estimate of the true rating of someone. I wouldnt mind if the devs change the system to TrueSkill or Glicko-2 based ratings.