An open letter to the developers about team matchmaking

Age of Empires II: DE II - Discussion
1

Hi guys!

Do you ever play multiplayer team games?
Have you ever felt like that the (ranked) games you play are either too easy or too difficult?
I’m pretty sure at least some of you felt this way before, and I’m sure many threads are talking about this, but I just couldn’t find this exact problem.

In this post - or rather, letter - I’m putting forward a possible reason why this is happening. Yes, it’s the ELO-rating system, and how badly it functions.

An open letter to the Age of Empires II designer team:

Dear Designers,

I feel betrayed after every ranked team game I play.

This has been brewing within me as of late, and I feel like the ranked ladder games – for me at least – are becoming near unplayable.

Why, you ask?

Quite simple. I sit down to play a decently challenging ranked team game every night with my buddy after work and studies. We are mid-ELO 1600-1700,and we play because we love the competitiveness and the vibrant rush of ranked games. They do provide an area where one can improve, one can strive to be better, and where we players can be a challenge to each other.

But recently I’ve noticed a rather annoying thing that’s ruining the game.

Ranked team games are becoming very very unbalanced. Especially around the mid-ELO range.

We’re very often put in a team with random people of differing ratings, which is quite fine. But the issue begins, when I notice the huge gaps between our proficiencies. And oftentimes, the games we get into are either way too easy, or are punishingly difficult. I tend to check the ratings of who I play against after a match, and it’s shocking to me how unbalanced the ELO ratings can be. There are +/-200 ELO differences between players. This, superficially, shouldn’t matter, as skill levels average themselves out, right? Wrong.

The main problem I believe with these games is, that the ELO ratings get averaged out, which is totally ridiculous. A skill gap of +200 ELO in relation to any specific ELO score is is way larger than a -200 ELO skill-wise. Numerically the average might be fine – but I assume there are many who feel how haphazard and often unsatisfactory this can become. Hell, even my win rates would reflect this: One game I win by a landslide, the other I get defeated very very quickly.

As far as I can tell from these 2-3 games per day, skill levels, especially in team games, are applicable exponentially: a high-skilled player can very easily steamroll even multiple opponents – thus even a smallish increase in ELO can mean a rather large gap in knowledge.

I love Age of Empires II – but I hate that the multiplayer environment forces me to play against people not at my level.

One can argue of course, that I should play more, and get better at the game. I’ve been playing rather regularly during the past 3 years with some minor gaps, and I don’t think my reaction times, or my coordination can really improve beyond a point.
And sadly, I feel like a very key aspect from multiplayer games is starting to miss right now.

I’m no expert on the matter, and I see only a limited scope of the ELO mechanisms. But this „averaging out” method seems to be way too simplistic to me. I believe that more data about players should be considered to estimate their skill level. Such as 1v1 ELO. Thinking that someone who is a beast in 1v1s will not be a good player in team games seems misguided at some level.

I’m not willing to play a game more just to enjoy it – I want to enjoy it so I want to play more.

Yours faithfully,

AllegroGuitar81

9 Likes
2

TG ratings are terrible and the devs doesnt seem to care. For more info see:

2 Likes
3

it doesnt help your position that you dont know how Elo is spelt… you understand its an actual word and not an abbreviation?

there’s also more at fault than simple averaging out, like elo inflation, arranged v random teams and so forth.

this is also variable and can be misleading… i might be an ace in 1v1, but i dont use discord or fit in an arranged team so i cant coordinate like they can

and conversely someone might be pretty average in 1v1, but have much better team play and thus do even better, since larger team games are more about how you work as a team

4

It is always nice to see someone noticing the same things as I! As I have suggested in the same topic that @WoodsierCorn696 linked, I think that, because of this issue, both in matchmaking and in Elo point allocation, the average of the Elo should be taken on an exponential scale:

And it really is an issue: When I team up with someone who is way lower ranked than I am, the match is almost guaranteed to be too easy for my team.

As has already been pointed out, though, this is not the only issue plaguing team rankings. The most important problem is team Elo inflation, which @WoodsierCorn696 goes into detail in the topic he linked.

5

Honestly, I don’t know if or how they can really ‘fix’ this. Improve it maybe but removing the issue entirely would be tough surely?

  1. Playstyle

The biggest cause for this is in my opinion playstyles. Some people are just not very good at (or dislike) the early game aggression. It’s boring and annoying. More enjoyable to focus on getting the econ locked down and once it is they turn on military and swarm, this playstyle of course being better for more closed maps like black forest. Other players have a more aggresive playstyle which is generally considered the ‘more skilled’ players style, hurting the enemies economy more than building their own type thing which is good on those open maps like arabia. The elo doesnt consider your playstyle so if you got 3 boomers and 1 aggresive player against the enemy 3 aggresive players and 1 boomer you might wipe the enemy early without difficulty. if they somehow get to late game cause of good walling or a favouable map you may find the boom players stomp the aggresive guys. The only way I could think to mitigate this would be to give the option to queue as an ‘aggressive’ or ‘boom’ player and the game tries to take 2 people from each queue to make a team of 4 with aggresive being flanks and boom being pockets. This would make the queues longer for whichever queue was more populated though.
In short, ELO doesn’t consider playstyle when pairing people and could favour a team of aggresive players with an aggressive map or boom players on a booming map.

  1. Maps

Next is maps which ties in to playstyle. They are reguarly shuffled and have a variety of aggressive and booming maps to encourage all styles of play. This is fine for getting everyone to play but unfortunately we don’t have enough bans for all the maps we don’t like in team game (unless you are a premade party) so quite often you will just get a map you aren’t as good at. As someone who prefers to get to castle/imp before i even consider fighting it is tough for me to play maps like arabia where walling is a pain and enemies often attack early. Meaning my skill level on this map would be far FAR lower than my skill level on black forest. Unless they implement a skill level for each map or each category of map (boom, aggresive, water, balanced maps) to get people accurately playing others their rank for any type of map they are strong/weak at you will just have to accept that sometimes an aggressive player on a booming map is confused and doesnt know how to macro a long game because they cant “attack attack attack!” or a booming player on an aggressive map gets bullied out of the game in 15 mins.

  1. Team Play

I have encountered many players who think they are god’s gift to age of empires because they have a decent 1v1 rank but they are tragically bad in team games and vice versa, good team players who have a super low 1v1 rank. Unfortunately 1v1 games or even 2v2 games to an extent are largely about just your skill, if you mess up you cant rely on a pocket to save you and buy you time to recover. These players often focus solely on protecting themselves and winning their fight against the enemy closest to them and will completely ignore allies in team games without adapting to a situation. Just yesterday I had a game where 3 of us team walled (we discussed it in chat) and the 4th guy (one of the flanks) was a 1v1 player and just walled his own tc in. The enemy team just ran right past his walls and into the rest of us because he hadn’t even attempted to team wall before we got markets and realised the problem. This sort of solo play of course causing the team to fall behind economically because we got raided then means this solo player starts acting like he is the super carry and crying that everyone else is useless while we try to finish the team wall and fix our economies. The opposite also happens where a team player who has good macro and is so used to team walling and having trade does poorly when put into a 1v1 where there it is basically just ‘best build order wins’.

You are correct that higher elo people power spike earlier and can take on multiple enemies but this (in my experience) isn’t normally a elo difference problem but more of a playstyle/map matchup problem. I see it happen when an aggresive player goes against boom players that dont really know how to deal with the early aggression and he knocks several of them out of the fight before they are ready. With no safety mechanism like in other online games like league of legends where less skilled players can hide under a tower and play catchup it is very difficult for a less skilled player in aoe to play catchup if they get pressured early. it’s why arena and fortress are so great, gives people a chance to survive early aggression if they arent used to it.

I think i covered everything, feel free to share opinions or god forbid criticisms of my thoughts :stuck_out_tongue:

6

I actually felt this in my last game. So game settings first:

2vs2
Fortress
Lithuanians (me) + Celts (ally) vs Goths + Slavs

Situation: I had bunch of Elite Leitis with 3 relics, when Goths flooded me. 2 castles were lost. 1 was denied. Asked my ally for help. Celts being an infantry civilization had a good chance of countering Goths. My ally also had heavy scorpions (which we was using to counter Slav Halbs). Seeing the Goth flood, I was thinking of switching to my Skirmishers (countered by Huskarls) or Champions (lack armor). Then I thought Celts had Woad Raiders whicu could be great. I asked my ally for help. He started making “HUSSARS”. The situation got worse and he RESIGNED.

Later, I actually did survive the Goth flood. Unless the pressure of 2vs1 got me.

7

are people just not building production facilities? if you would start to create military then maybe you would start to win.

I’ve played like 20 ranked matches now and I feel most comfortable as flank, the enemy focuses on me, my economy is in shambles, my allies make military to clean up the enemy military too late, then I hit castle age, got stone walls down, make an extra two TCs, my economy starts rolling again, and I make 5 to 10 production buildings and spam trash at the enemy, and then we win.
Generally speaking what I’ve seen so far happen is that one of my teams sides gets horribly pushed and one player is down to 20 pop in late castle/early imp, and the same is happening on the other teams side.

and I’m not saying I’m good or anything, but, we had like one 2400 team rating person and the rest of us were all low 1500s and we were generally doing okay, had some great hourlong matches, that swung both ways, first the enemy overwhelemed with gold units, but couldn’t breach our defenses against our trash and then it was our turn to up the ante and produce our gold troops with saved up gold and push them back because they ran out of juice.

Get into discord with some other like minded folk and you will perform better as communication wins more in team games than individual skill.
Text chat just takes too long and can be easily missed.

The only times we tend to struggle with our discord groups(we are all from the official aoe2 discord btw) is against some (insert clan tag here) where the enemy goes for certain dedicated compositions of OP teamgame civs instead of just having fun in team games.
(Franks, Ethiopians, Mongols, Persians) those are the majority of civs I’ve faced against and it gets boring, many of those players were also just plain said bad with those civs and the mentality of them also lacked, they mostly looked like a random 4 stack that got matched together without communications, and they either failed to finish us off properly or didn’t press the advantage, they also had no trade to speak of and no team walls, so once one fell that player was almost certainly completely gone, they played this as a 1vs1vs1vs1vs The enemy where we played it as (2vs2)vs(2vs2) if that makes sense.

I don’t say that the algorithm is perfect for matchmaking, but it’s the best you can get with a basic system like elo, they’d have to make their custom elo system and have more statistics that determine each players ‘potential’ and ‘abilities’ like how they done it in dota 2

8

bad_matchup
bad_matchup1222×271 58.8 KB

I think the matchmaking system is normally pretty good. The problem comes when two friends have vastly different elo. Now, I know playing with your friends is a lot of fun and I do it too, but if the elo gap is around 800 points, it’s a little silly, as happened when we got matched up with the team in the game above and got steamrolled by the 2200. There should be (imo) some reasonable cutoff for queueing together in ranked play, like ~200/300 elo difference. Some people may say, “But then how do we play with our friends?” Well, you still can in unranked play. I think that’s a fair compromise and it allows ranked to be more of a reflection of competition and skill than allowing friends to carry each other.

3 Likes
9

I don’t think I have ever disagreed more with a post on this forum. Instead of banning the possibility to queue with players of different Elo, how about fixing the matchmaking so that this isn’t an issue?

10

1757 vs (1829+1826)/2=1828 + (2212+1446)/2=1829

the team with the higher combined elo won.
So on the right side it matched everyone perfectly together the same can be said about the left side, MAYBE JUST MAYBE, the queue time was over 5 minutes and there wasn’t another team of 4 closer to 1828 in the queue at the same time.

In rainbow 6 siege there’s a restriction on how far apart in rating you can be to team up for ranked, and I dislike it, it does eliminate sort of a ‘boosting factor’ (which in my opinion is the whole reason that one person’s chart looks inflated, which if you apply common sense and understanding of the current system to, makes sense, the whole team elo is inflated)

There was no imbalance because of the 2212 player, because it is assumed that the 1446 just dies to anyone that is 1700.

It was the combined elo that was 72 points above you that made it harder for you to win.
I don’t quite have the equation from SOTL.

the matchmaking isn’t broken there’s just not enough players close enough online at the same time.
which most people don’t seem to understand.

We know a few things, one that many people have way inflated team rating, and secondly that there’s a lot less people playing ranked compared to how many people play the game concurrently.
People compare age system way too much to systems with upwards of 10 to 100 times larger player bases where there’s more players at any given elo playing at a time.

11

I mean yeah we lost a game we should have lost, which isn’t an issue. The issue is that an 2200 is vastly better than an 1800. It’s not really a fair game, and the elo system itself recognizes this because we only lost like 1 point. The point of ranked is to fight against people you have a roughly equal chance of beating, and that’s not the case when you go against someone 400 points ahead of you. Having a 1400 on the team doesn’t really balance this and I’m wondering if you’ve never been in such a situation to suggest it does. This doesn’t happen a lot, but it’s frustrating to waste time on a game the elo system itself suggests we have next to no chance of winning. If you want to play against unbalanced skill levels, that’s what unranked gameplay is for.

1 Like
12

Did you read my thread as well? It is common knowledge that the current elo calculation for team games is bugged. Elo uses two values: Your rating and your enemies rating. In team games, the highest elo on the enemies side is picked. As result elo tend to drift upwards and this will mess up the balance. High elo doesnt just mean ‘better player’ but also ‘someone with many games’.

Just the simple change of not picking the highest elo on the enemy side, but pick the average rating will already make a huge different. It will be a huge improvement. In the end i think a plain elo system like we have is never able to solve all issues. But i do really think most issues will be solved by just this simple tweak.

13

you don’t understand, the match was only unfair in terms of the enemy team being 72 points higher on average than the rest of your team.

The 2200 and the 1400 balance each other out. that’s how elo works. the 1400 is not even close on the level of skill as is a 1700 player.

image
image1467×465 81.9 KB

here we see two of my last games, the last I lost because our side had an average of (1774+1882)/2= 1828, so about what your opponents had of average skill, with the other two on our team having 2365 average skill making our team have 2096 average elo rating, the enemy team meanwhile had 2211 rating, which is 115 points above, they did win, I only lost like -2 rating from that game.

the below game left side had 1957 elo vs us having 1966 rating, which is fairly even, and it was a fairly tight game if I say so myself.

but this is just to showcase that despite these ‘uneven’ games occurring, there’s actually nothing much going to change about your personal elo rating, as in those unlike for you to win games you as a result won’t lose much elo at all, on the other hand if you were to actually win that game, you will be swimming in elo points.

I also can’t really remember how long the upper game’s queue was but it was at least 3 to 5 minutes, so there weren’t a lot of players in that range queueing at that time.

14

@NoraLoralei You make some valid points, but just because yes, sometimes you do get matched against higher rated opponents because there aren’t many in the queue, that does not mean that there are no issues with the team matchmaking system that can be solved easily.

First of all, you assume that a 2v2 game with a 2200 player + an 1800 player versus two 2000 players is balanced. It is actually not; the team with the 2200 player has an advantage. No, unfortunately I do not have stats to back this up, but I have teamed with players with wildly different Elo in the past and noticed the trend.

Second of all, team Elo inflation is a real issue. @WoodsierCorn696 has linked to a thread where this problem is discussed extensively.

15

I did not say that elo inflation isn’t an issue, I just said that it’s a thing in every game and to be expected when players of vastly different elo are allowed to match up together.

That’s the main reason why the elo system only really works for 1vs1 as team games have different factors to go for them.

in 2vs2 yes the game is more balanced in favor of the 2x 2k people ganging up on the 1.8k and destroying them turning it into a 2vs1 effectively and then the solo player I’d argue can’t really do much more, with their skill they’ll likely put up a bit of a fight but if the 2 players boom enough have stone walls up and make enough units it’s 200 army (with 200 vills on eco behind them) vs 100 military and 100 eco.

Had this happen before, that’s usually the strat, or if you’re really brave you go double team the higher player to cripple them.

2vs2 has a lot of other issues outside of uneven elo on one team side.

16

Not at all. Neither Voobly nor HD had team Elo inflation. So no, it is not ‘expected’, it is a DE bug.

I disagree with the rest of your post too, but it seems both of us only have anecdotal evidence so it is not the most important part of the discussion.

17

I feel for your partner, doing so much work for so little appreciation.

18

I’m not trying to be rude, but be honest and look at KvaKva’s elo graph, they’re getting hard carried up to the point the high elo can’t carry as consistently anymore. That’s the problem, even tho it’s “balanced” by average elo, it’s not actually balanced in skill. You don’t go from losing to 1400’s to beating 2000’s in 5 days. Skill doesn’t rise that fast. If it does, i guess i should watch Hera’s guide to 2k again :rofl:.

For real, the game matchmakes on average elo, but it awards / subtracts elo based on the difference between your elo and the highest rated opponent. They do this because it’s inherently imbalanced in favor of the higher rating.

I did watch the ghost lake win from KvaKva’s perspective. The opponents fought 2v1 (and lost) against the 2300, while a single opponend went even 1v2 against the two lower rated players. That’s not balanced gameplay and was almost entirely carried by the 2300.

I’m genuinely trying to be helpful when I say KvaKva should practice some build orders, because it’s super weird to have a dark age like that and be near 2k. I’m not going to pretend mine is perfect every time, but having the basics down is important.

I’m glad you love the game and keep up the energy, but you’ve gotta realize why imbalanced teams are so frustrating to play against.

19

We dont expect elo inflation. There is no reason why we would expect this. The median elo would stay about the same. It wont go up like we have for team games. That just shows something is currently off in the calculation. That is really the main issue why team games are currently are unbalanced.

The elo system is meant to be for 1v1, but can be expended to team games as well. There is nothing really wrong with this. Yes, it wont never be perfect, but a non inflated elo system would solve most of the current issues in the TG rating. Pretty sure about that statement.

1 Like
20

We do. Why? Well lets assume that the winning team always gains as much as the loosing team looses, that should be at least way better than the current system. But we still would have inflation, extremely slow and not worth to mention, but still here. Why? Well we all start with 1000 elo, and after some time one gets up to 1200 while someone else drops to 800. All fine, all good. But I think there are way more low elo players quitting the game than high elo players. So lets assume the 800 guy stops playing, but the 1200 guy stays in → we slowly gain more elo in the system → inflation.
(Ofc, every >1000 elo player stop playing causes deflation, but still its more unlikely imho).

This very slow, but still mentionable (at least after a few years) inflation I have seem in some other games that always do the 0-sum elo calculation. So we are not absolutely inflation free with the 0-sum, but tbh: it would be great to have that inflation of the last year over the timeperiod of several decades