Why are ELOs slowly rising on average in team games

tl;dr how is ELO computed for TGs and why is mean ELO rising over time?

I’ve been pulling ELO stats from the aoe2.net API and making some plots and graphs. I noticed the mean ELO in the DE 1v1 ladder is just over 1000. That makes sense: when a player gains X ELO, another gets -X ELO, so it should balance out.

However I noticed that the mean TG ladder ELO is higher than 1000 (it’s like 1200!) and growing! How is ELO being computed for team games? Seems problematic to have ELOs that grow over time. Someone searching for “tips for 2k ELO” in a year will be a noob looking for basic build orders! :rofl:

1 Like

People play some ranked games, lose their elo and leave the game. There’s always those players. So the elo has only 1 way and that is up. There might be other reasons too but that’s one that came to mind straight away.

Even if they rage quit and uninstall the game, they still own it on their Steam account and aoe2.net picks it up in their statistics.

Does it pick them up even before they reach 10 games?

I have actually no idea how ELO works specifically, I only know the basics. But elo medium rises in all games, yes?

That’s a good point - I don’t think it counts you with less than 10 games.

I still think something is up just because the offset is so large. For the mean ELO to be 1200 it means half the player base lost 200 ELO within 10 games and just gave up after.

I also thinks it’s strange that this phenomenon occurs for TGs but not 1v1s.

Yea I got nothing. Can you show the graph? Did it happen all suddenly or is it just continuous since launch?

Btw, what about microsoft store users?

I think aoe2.net knows about non-steam since it tracks separate steam IDs and “profile” IDs.

Now that I plot it, it’s actually worse than I thought. 1500 TG ELO means you’re average.

1v1 mean
tg mean

COVID has left me with lots of time on my hands :sweat_smile:

1 Like

Apparently it seems the elo calculation is broken on TG as the winning team will sometimes (often ?) gain more points than the losing team will lose.

It seems that the elo gain is calculated based on the rating of the highest opponent in the losing team instead of the average of the team.


Ahh that system seems like a recipe for problems. Devs please fix?

I’ve heard in some competitive shooters they look at your K/D and that influences how much your rating changes (like if you lost but carried your team, but lose less rating). Would be cool to have your final score be a factor in AoE.

There’s a other reasons the mean wouldn’t be 1000, although I don’t think this explains what you’re seeing, something else must be going on, as this would be the same on 1v1 and team games

Your placement games can earn (or lose) you a lot of Elo, I see people get 50+ and I’m pretty sure the other guy didn’t lose 50
After 28 days players are hidden from the leaderboard
Every 14 days, rating decays, I think it’s only only those with 1400+

I didn’t know about rating decay! So high ELO players lose rating while inactive? Does the leaderboard hiding happen for everyone or just players who are yet unranked? If ratings decay then wouldn’t the mean ELO start falling?

I did know your placement matches give amplified rating changes. I assume most early players won their placements as people moved from Voobly -> DE, but now most people lose their placements since they’re not regulars. So maybe that explains the spike and leveling off of 1v1 ELOs.

After one month of inactivity, AOE2.net doesn’t show your profile. I took a month break and went to search for myself and didn’t see my profile until I played a match.

See Analyses of the ratings - Spotting the issues for more information and a much longer discussion about this topic.

There is definitely an issue with the calculation of the team ELO ratings, other than “people quitting”. After playing enough games, even average players get a 2K+ rating. In an Elo rating system, playing more games should not automatically lead to a higher rating. You see many players with a 1000-1200 1v1 rating and a 2K+ team rating nowadays. I’m over 2300 TG myself, just because I played many games, but I don’t consider myself an expert.

Team ratings are not representative of a team player’s skill anymore.

True, TG “ratings” are just a number to be ignored

And even if they fix it retroactively for players, the concept of TG rating itself is not useful to judge one single player in AOE2

tbh these so-called TG "ELO"s should be removed from the game and TG matchmaking should be done on 1v1 ELO.

1 Like

What ? Are you saying a team of 1200 players has decent chance to win against 2000 players ? I don’t see where you’re going here.

No, there’s a massive difference between 1v1 and 3v3/4v4 play, especially pocket. Maybe 2v2 should be merged with 1v1 but 3v3/4v4 deserve their own ratings unless your plan is to ruin the matchmaking for no reason. They just need to fix the inflation issue and TG ratings will be fine.

1 Like

This is actually true. A team of 1200 players could be of equal strength as a team of 2000 TG rated players in some cases. Especially if the lower rated players played a lot less games (let’s say 20, vs. 400 of the higher rated team). If the win rate of the teams are equal, you will almost always see that the players with more games played have a higher rating. And that is not because their skill improves so rapidly, it’s how the current system works.

It’s not uncommon for average players to have a 2K team rating.

Exactly. Good hosts balance games according to 1v1 elo, not TG elo which provides fake data about the player’s skill.

and Viper:

1 Like

I’m not opposing that there is an obvious inflation issue, just pointing out that “ratings are irrelevant in TGs” as OP claimed is a blatantly false exaggeration. 2000 players will still smash 1200 players in 90% of cases despite all the problems that TG ratings have…

That’s a problem of devs’ inability to program TG Elo properly, not TG Elo itself. Using 1v1 ratings instead of TG Elo would be a bandaid, what they truly need to do is fix the inflation and everything will just work as expected.

The bottom line is that we can’t tell if it makes sense to have separate 1v1 and TG ELOs because the latter is broken and effectively hands out ELO for participation. The thread mentioned has been open since May. It’s like the devs don’t realize…

1 Like