What if they populated the ranked list with some AIs?

Age of Empires II: DE II - Discussion
1

Here’s the problem; AOE2 has an aging population. Many players have played thousands of games, and the average skill level is very high compared to beginners.

This leads to many players joining the ranked queue, and having to lose 10+ times to get to their proper elo. This is, of course, discouraging to many players. But it is also unavoidable, because ELO is just based on average player skill, and you HAVE to start players at the average skill level or the entire system breaks down. If you start players off at 600 ELO, sooner or later everyone is 600 elo and the problem starts all over again - not to mention the abuse the poor players who are ACTUALLY 600 elo will get!

The ideal solution is to have a constant influx of new players. Every time you add a new player at a low skill level, it bumps all the existing players up a bit, and gives other new players a more welcoming environment. Unfortunately, that requires unchecked growth, which is not a feasible long-term solution for a game that’s 20+ years old.

Games like Chess solve this problem by having local clubs with a lower skill level. When you go to play at the YMCA, you aren’t playing against Magnus Carlsen, you’re playing against Joe Davis, your friend since second grade, and you both suck. Unfortunately, AOE2 doesn’t have the sort of cultural weight where you can find AOE2 consoles at every park, so finding fun matches is very hard without using the ranked queue.

Unfortunately, this becomes a self-perpetuating problem; when you join a bunch of unfair games in a row, you will tend to quit, which causes the ranked queue to get even more deflated than before, since the existing average player now have even MORE experience at clobbering poor noobs.

So how can we fix that problem?

What if we added some artificial players?

It’s really pretty simple. You need more players at a low skill level, so the average is lower, and more welcoming to new players. You can’t get enough human players, so why not just add some AI players instead?

Say you add about a thousand computer players, coded to have an ELO between 600 and 900, but starting with the same 1000 ELO as everyone else. This will WIDEN the curve; most 1000 ELO players will beat them and immediately gain rank. The current 1000 will become 1100, or 1200.

As a result, new players will have a much more welcoming experience! They will be less likely to quit, and as a result, the ELOs of players will become even LESS deflated.

I think this could really improve the competitive environment, if done properly, and help make the game more fun, especially for newer players.

What do you think?

2

Oh, yeah, and we could even call this type of games single player mode.

9 Likes
3

I think trying to make the ranked start a better experience for newcomers is a good idea.

Your idea of adding thousands of AIs sounds bad because it would give low Elo players a “Single Player experience”. There shouldnt be more than a 5% chance of playing against an AI at any Elo. Players on the ladder want to play against real humans after all.

I am all for adding a couple of undercover AIs in ranked to have a realistic feeling of how good the AIs are. And every patch the balance team could report the AI Elos (for those who havent figured out or read from reddit).
If we have that, there could be an option to make 5 placement games against AIs, while letting the player choose the difficulty and map.

Without ressorting to AIs, the best alternative I can think about, it to annonymously track where the last 50 new players are after 20 games and set the statimg Elo to this value.

Of course, if the starting Elo becomes different from 1000, the Elo formula ust be updated so that winner and loser gets slightly different Elo change, in order to slowly bring the average Elo back to 1000.

1 Like
4

“sooner or later” it would take a loooong time to significantly change the average elo: Effects of changing the starting elo to 500

TL;DR: with 500 starting elo, adding 10 000 new players only shifts the average by 100 points. I reckon 10 000 new players takes several years

then players between 600 and 900 elo will be fighting bots a lot. You don’t queue up for ranked to fight a bot

I think this could be the solution: put a bot of each difficulty level (standard, medium, hard, extreme) on the ladder to find out what elo they have. then if a new player wants to join they can self-assess (or maybe even make them play against bots) which level they can/can’t beat and make put their initial rating at the level of the corresponding bot

5

If players are being challenged, does it really make a difference? Honestly, the new Extreme AI is quite good, and does many things a decent player would do. Personally, I’d rather have a fair fight against an AI that counted for ELO if I win, than an unfair fight against other players who are much more(or less) skilled than me.

I don’t know, is that really why people play ranked? Or is it to assess their own abilities and get a fair fight? For me, the allure of Ranked is to get a fair fight; whether or not it’s against a human doesn’t really matter. If anything, it’s a negative, since I feel bad for making another person lose.

Well, according to steam metrics, they added over 10k players following the steam page(IE, social players, the kind who would engage in multiplayer) in about 2-3 months. It would happen a lot faster than people think - and then what do they do?

It’s just not a sustainable solution.

I’d be down for starting out with small numbers of bots for testing purposes, though. Heck, with the research Microsoft is doing into AI and deep learning, maybe(and I’m just talking out my butt here) they could train some AIs on real games and make some bots that are basically indistinguishable from real players.

Then people literally couldn’t tell the difference.

1 Like
6

so the steam group for aoe2de has 237161 members (source: Steam Community :: Age of Empires II: Definitive Edition Member List). is this is what you mean? according to you this group grew by 10k in 2-3 months (how can one find this out?)
assuming that there is a somewhat constant ratio of followers:ranked players (and I admit this is a big assumption) that means that 40k/237161 = new ranked players/10k in 2-3 months. so about 1800 new players per quarter. it would still take over a year to reach 10k new players. I don’t think 100 points of shift in elo over a year would be noticable by most players. The only way you feel that is if you take a long break and return (and are somehow still the same level)

7

If you look at the steam metrics, the steam hub followers increased from 80k in january 2020, to 225k in jan 2024, or about 10k new players every 3 months.

But most of those players don’t stick around. So it really is about 10k new players per quarter, we just only retain a small percentage of them. But the ELO effects of all those players would still be there, and would be quite massive.

8

^ this

Make people play ranking matches against bots.
Only if they beat all the AIs they get 1000 ELO.

If the AI improves enough in the future (without cheating) maybe the ELO for beating all AI can be 1100-1200.

To make it not too annoying you start with being matched against Hard AI and an hidden ELO of 500.
Every time you win your hidden ELO improves by 100 but the AI difficulty increases. If you lose your hidden ELO and the AI difficulty stay the same.
After 5 matches you get your starting ELO.

9

ah thank you. in that case the extrapolation to new players I did is incorrect. however we still don’t know how many of them play ranked. also if a player starts with the current 1000, then loses his way down to 500 that causes inflation instead (we can already see this as the average elo has increased to 1050 or so)

again, this depends on HOW MANY players are joining, and how long they stay.

If the total number of players is more or less constant we need new players to have the same average as the players who are leaving. But I don’t think elo deflation/inflation is a huge problem as long as it’s slow. The player base increases in skill as well, so that shifts elo for individuals already

10

I play chess too, I solve puzzle first, play with AI next then I played with other people online,
You can learn nothing from playing with people in the beginning.

We already have VS AI in quickplay.

According to my experience, I started with VS AI first , joined lobby games next, learned FC from video , then joined ranked without any problem, I didn’t lose my first game, 1.2k ELO and below can’t stop FC knight at all around the the time so I believe people can settle well by knowing only 1 strategy.

11

I don’t think putting AI in ranked games would be ann efficient solution, since it would contradict the proper definition of PvP.

Sure, new player could use AI to understand the basics of the game, improve a bit and evaluate their level.

Understand the basics of the game
I don’t think it’s a good idea to go wild in the ranked without knowing the essentials. Luckily, there are already the Tutorial Campaigns (Wallace and Art of War). New player should at least complete these tutorial before adventuring in the ladder.

Improve a bit
Playing solo vs the AI could help, especially if the game add some sort of advanced tutorial or training mod. I think something like the build order mod, that already exists but could be improved. But I’d like even more the idea of selecting the opponent AI opening or strategy in order to learn how to face it: I want to traint against the scout rush? I could have an option to force the AI to do it and learn how to counteract.

Evaluate the level
New players could then have an evaluation based on the difficulty of the AI they beat, so that they could have assigned a supposed ELO.
This ELO would be hidden and the system could use it to make easier for the player to reach it with the proper ELO withinf the first games. For example by finding matches with player with the supposed ELO, or by increasing the point loss in the first matches until the player reaches the supposed ELO.

1 Like
12

This is a great idea. Like a preELO

13

That could potentially work. You’d need to slowly add in the extra ELO to all other players to compensate or you’d end up with downward ELO drift, but if you just gave an extra point to players here and there, it would average out over time.

You’d need to have some test AIs in the ranked queue to find their effective ELO though.

14

currently we have upward drift. so it might even cancel out

15

I think it is a good idea to put in AIs in ranked score list. Chess.com also has this feature.

1 Like
16

Yes, but only 3-4 of them, just to benchmark (and keep that benchmark updated) the corrispective ELO for easy, standard, hard and extreme AI

1 Like
17

might as well have one of each level: easiest, standard, moderate, hard, hardest, extreme
I think there could also be some rule that you only play against AI at most every X games, so that pvp is still mostly pvp

18

As long as they’re mostly at middle elo, the average player would have plenty of games. That’s where the vast majority of players are, after all.

Good point!

19

The chance should be 0.00000000% repeating of course chance of AI foes

20

why? I think it would be really useful to know what level AIs are and it would allow better placement for new players. Could even have an option ‘opt out of AI elo calibration’ or something like that