And yes, you calculated player elo quite accurately. But people will change behaviour once match making changes. You'll have a player base of 3 parts, legacy greats who have elo in the thousands who can't be bothered to start an alt and win anyway. Normal players who enjoy 50%WR and the worst alt tribe who start at the bottom and try for perfect seasons. If there are enough numbers in the alt tribe it would turn into just random matchmaking.

That wouldn't work. The best players have ~1700 elo and the worst players have ~600 from all the measurements. If you make a clan of alts, they would get to their real elo very fast, unless you lose on purpose, but at that point you're not gaining wins either. Plus making alts and doing stuff like that is what Nono did under the current system anyway.

At the moment, matchmaking is based on a clan-based rating - so the clan is in focus, not the single player.

I don't see why the rating is important, having the clan earn points together, yes. But having one rating for people with vastly different skill levels? No. Show me another game, any game where a group rating system exists. Group points, yes, group rewards, yes. Group rating? Nope.

At the moment, swarming is a good strategy - so clan coordination is in focus, not a player's behavior in isolation.

That is a meta strategy. You're not trying to defeat the opponent, you're trying to defeat the matchmaker. That is bad.

At the moment, rewards are equal for all participants of the clan (given at least 1 real win). Do I only have a 5% win rate? Irrelevant, as I get the full rewards from my clan.

That wouldn't change if you'd change the matchmaker. But if you think that people with 5% win rates are happy with the system, then talk with them.

Rating per player is already implemented well in other ladders, and not necessary for an event like Clan War I think. It would be interesting info, but Clan Wars is about a Clan's production, and not based on any single player.

Which is why the results and rewards would still be per clan basis and not per player basis.

Though knowing everyone's ELO would be nice :) Care to publish such a thing?

Since the game itself doesn't track it, I am not sure if that would be a nice thing, especially towards the players on the bottom part of the list.

If you're able to get this from 5s's dataset, it'd be helpful to see how the shared rating affects player behavior in a few ways:

- likelihood of remaining active in CW in relation to your CW win rate (maybe further disaggregated by clan win rate); do the 20% wr players and 80% wr players stick around? Or do they get bored of lopsided matches and stop playing?

- player migrations across clans (over the seasons). Since CW shared ratings encourage high-skill players to gather in high-skill clans (where the matchmaker can't just offset their wins by lowering teammates' win rates to keep the clan around 50%), my hypothesis is that CW has made it harder than before for mixed-skill clans to retain high-skill players. If this is true, we'd expect to see high WR players move from lower-rated to higher-rated clans.

So far, this discussion seems like a matter of preference- should CW have evenly-matched, exciting games for players, or should it reward clan-level metagaming like swarming templates? Should CW be more dynamic in terms of who's able to compete, or should it effectively have two tiers (clans that can sustain >60% wrs, and clans that don't have consistently high skill) with distinct but overlapping participation contests?

But looking at player behavior- what do more players seem to enjoy (and not just the ones currently participating happily in the top 3 clans)?- might provide weightier evidence that's harder to reduce to different tastes. Also, there's probably been enough seasons by now to come to some conclusions about what drives participation.

- likelihood of remaining active in CW in relation to your CW win rate (maybe further disaggregated by clan win rate); do the 20% wr players and 80% wr players stick around? Or do they get bored of lopsided matches and stop playing?

I have run the numbers for this, there is no need to do migrations as FiveSmith already did that, you can find the link in his profile. Here is the percentage of people in the different win rates with at least 10 games played that haven't played a single game in the last 3 months:

0%-20% winrate - 69.93% stopped playing
20%-40% winrate - 59.81% stopped playing
40%-60% winrate - 58.11% stopped playing
60%-80% winrate - 60.29% stopped playing
80%-100% winrate - 69.04% stopped playing

Apparently people who are in the most fair ELO matchup tend to stay and play more than both people who are getting unfair matchups from the winning or the losing side. Another way to present this data better is to divide people into 2 groups, those that are within 33%-66% win rate (good matchmaking) and those who are either below 33% or above 66% (bad matchmaking):

bad matchmaking leave ratio: 63.24%
good matchmaking leave ratio: 58.38%

Clearly the quality of the matchmaker does have effect on player retention. The more unbalanced games people get, the more likely they are to stop playing. Of course a large number will leave either way, but we don't exactly have infinite number of players, so the more we can keep the better it is.

Apparently people who are in the most fair ELO matchup tend to stay and play more than both people who are getting unfair matchups from the winning or the losing side.


Careful, those are extremely dangerous things which you read out of the data.

Edited 9/27/2023 07:22:44

How about: A 0-20 win rate player joined an open clan after starting playing the game, he threw some armies around for a month or so and then quit the game altogether?

Also neither 0-20% win rates nor 80-100% win rates look very sustainable to me, apart from some outliers. So if I have a lucky streak and get an 80% win rate, I have 2 options. I either quit playing and show up as one of the 80-100% win rate players who quit playing or I continue playing and fall into the more reasonable 60-80% bracket.

There is much more going on here.

Edited 9/27/2023 07:22:25

How about: A 0-20 win rate player joined an open clan after starting playing the game, he threw some armies around for a month or so and then quit the game altogether.

That's not really what the most common case is. Here are the average number of games played from each group:

People who stopped playing:
0%-20% - 25 games played on average
20%-40% - 42 games played on average
40%-60% - 54 games played on average
60%-80% - 79 games played on average
80%-100% - 45 games played on average

People who are still playing:
0%-20% - 35 games played on average
20%-40% - 119 games played on average
40%-60% - 177 games played on average
60%-80% - 269 games played on average
80%-100% - 174 games played on average

edited to add:
Here are also the average number of days they have been playing in clan wars per winrate group:

People who stopped playing:
0%-20% - 287 days
20%-40% - 278 days
40%-60% - 286 days
60%-80% - 303 days
80%-100% - 198 days

People who still play:
0%-20% - 376 days
20%-40% - 497 days
40%-60% - 587 days
60%-80% - 608 days
80%-100% - 492 days

Another interesting thing that shows that player retention is best at 33%-66% win rate:

Average number of days played for all players with 10+ games (both who still play and who stopped) per win rate group:
0%-33% - 341 days
33%-66% - 411 days
66%-100% - 378 days

Even though the top winrate group has more days played than the bottom winrate group, the people who on average play the longest are those who get 33%-66% win rate.

Edited 9/27/2023 07:38:08

I don't see why the rating is important, having the clan earn points together, yes. But having one rating for people with vastly different skill levels? No. Show me another game, any game where a group rating system exists. Group points, yes, group rewards, yes. Group rating? Nope.

Due to increasing level of ignorant confidence, I decided to post not one, but two examples (from the top of my head) of real sports, where teamed ratings are calculated from individual matchups and those teamed ratings have direct influence on matchmaking:

1. Tennis - Davis Cup - https://www.daviscup.com/en/rankings/current-rankings.aspx - team ratings influence seeding (i.e. matchmaking)
2. Football - Champions League - UEFA coefficients https://en.wikipedia.org/wiki/UEFA_coefficient (here is the "team" national rating is combined from "individual" ratings of football clubs) - directly influences how big quotas a nation will get in the EU footbal tournaments (i.e. matchmaking)

You are welcome.

Due to increasing level of ignorant confidence, I decided to post not one, but two examples (from the top of my head) of real sports, where teamed ratings are calculated from individual matchups and those teamed ratings have direct influence on matchmaking:

1. Tennis - Davis Cup - https://www.daviscup.com/en/rankings/current-rankings.aspx - team ratings influence seeding (i.e. matchmaking)
2. Football - Champions League - UEFA coefficients https://en.wikipedia.org/wiki/UEFA_coefficient (here is the "team" national rating is combined from "individual" ratings of football clubs) - directly influences how big quotas a nation will get in the EU footbal tournaments (i.e. matchmaking)

Completely irrelevant examples as in both of those cases the teams fight the winners of other groups and in order to actually win you do have to fight the strongest team. Also, in both of those cases they are matching the strongest against the weakest first in order to help the strongest teams get into finale. Most importantly the initial matching has a much lower outcome on the final result because if an outlier shows up it will be corrected by the time we reach the finale. In our case you get 1 match and that is it. Then the next day you get 1 match again and so on. The winners do not go to face winners and there is no correction. If you constantly get bad matchups, you will constantly get the same bad matchups day after day. Which is what produce people with 0-20% win rates and 80-100% win rates.

Since I am ignorantly confident, please enlighten me as to what the actual purpose of the current matchmaker is and while you're at it please address the following:

1) is the purpose of the matchmaker to reward skill? If so then why isn't it matching weak players against strong players or just randomly?
2) is the purpose of the matchmaker to reward participation? If so then why doesn't it keep the win rates at near 50%?
3) is the purpose of the matchmaker to provide games where both players have a chance to win? If so then how does it achieve it?
4) is the purpose of the matchmaker to provide games to most possible players? If so then why do we have so high number of no games found?

I am completely puzzled as to what the matchmaker is actually supposed to do in it's current state. It doesn't seem to achieve anything.

Edited 9/27/2023 09:23:18

Completely irrelevant examples as in both of those cases the teams fight the winners of other groups and in order to actually win you do have to fight the strongest team. Also, in both of those cases they are matching the strongest against the weakest first in order to help the strongest teams get into finale. In our case you get 1 match and that is it. Then the next day you get 1 match again and so on. The winners do not go to face winners.

These matchmaking is specially made to increase the amount of games played by top teams. You may disagree with the purpose of it, but hardly you can deny that these are examples of "another game, any game where a group rating system exists".

Edited 9/27/2023 09:23:26

These matchmaking is specially made to increase the amount of games played by top teams. You may disagree with the purpose of it, but hardly you can deny that these are examples of "another game, any game where a group rating system exists".

Which is what I wrote as well. Are you saying that the purpose of the warzone matchmaker is the same? If it is, then how does it achieve this? If it isn't then how are your examples relevant? Also, before we go into more of this, can you explain to me what the purpose of the current matchmaker is? I still can't figure it out.

Edited 9/27/2023 09:27:07

So many people told you already, that the current matchmaking if focused on matching clans because Clan Wars is a team competition, so that I am not sure, that I should tell you the same thesis again.

So many people told you already, that the current matchmaking if focused on matching clans because Clan Wars is a team competition, so that I am not sure, that I should tell you the same thesis again.

This is not a purpose. What is the purpose? Here are some possible purposes:
1) find fair matches
2) reward participation
3) reward skill
4) avoid "no game found" situations
Since it doesn't do ANY of those, there must be a 5), 6), 7), 8), 9) that outweighs those that I am missing somehow. What are those? What is the actual thing it does that is good for the players?

edit:
Also, while clans get points together, the matches are very much NOT A TEAM COMPETITION as we don't play any of the games as a team. Instead we play a bunch of 1v1 games, or 2v2 to a lesser degree. If you think that a clan mate who goes 0-10 doesn't care because another clan mate went 10-0 so the clan is still at 50% win rate, then I am not sure what to tell you.

Edited 9/27/2023 09:45:35

You can’t have both fair matches and no game found if not enough players join.
Keeps giving games even if there is a big rating difference.
Maybe there should be some sort of threshold for matching and deciding a random matchup based on similar ratings? Obviously this could be abused by deliberately losing to get easier matches

Edited 9/27/2023 09:54:26

1) find fair matches -> Yes, it finds fair matches between clans.

2) reward participation -> No, to be honest, from the top of my head, I cant imagine a matchmaking that rewards participation? Like giving extra points for just playing the game? Should that be shifted from matchmaking to the "rewards system" (in that case CW reward system does indeed reward participation"

3) reward skill -> No. In CW skill is rewarded by the "rewards system".

4) avoid "no game found" situations -> It does that poorly. Yet, Fizzer says, that it is the lower skilled clans' issues, and they just need to get better. I think that this aspect needs to be improved.

You demonstrate very strong opinion on what a team competition is supposed to be.
When there is a weak player in a team of Tennis, Judo, Figure skating, Warhammer 40k, etc ... (basically any team competition, that consists of individual 1v1 matches or sub-competitions) and he gets 0-20, then he is not provided with some easier matchups. But you suggest that weak players should have easier matchups in the team competition of Clan Wars.

Edited 9/27/2023 10:00:28

1) find fair matches -> Yes, it finds fair matches between clans.

First of all this is demonstrably untrue. If you find fair matches, ie matches where both clans matched have an equal or close to equal chances for getting points then the clan winrates would sit near 50%. They don't, so it doesn't.
Secondly I haven't seen a single game where a clan was playing against a clan, all I see is thousands of games where players play 1v1 games. Most of them horribly lopsided.

2) reward participation -> No, to be honest, from the top of my head, I cant imagine a matchmaking that rewards participation? Like giving extra points for just playing the game? Should that be shifted from matchmaking to the "rewards system" (in that case CW reward system does indeed reward participation"

You just contradicted point 1). If you have a 'fair' system, it leads to near 50% win rates therefore rewarding participation. If a system is not rewarding participation then the matchmaking isn't 'fair'. Even if you perfectly match clans against clans with similar strength, it will lead the clans to have close to 50% win rates. The clan win rates will fluctuate more if the matchups are more lopsided.

3) reward skill -> No. In CW skill is rewarded by the "rewards system".

The matchmaker is part of the CW. As proven before the number of points per clan are predetermined by the matchmaker and are predictable using player ELO once the matches are made. Therefore the matchmaker is what ultimately decides who gets rewarded and for what. A fully random matchmaker that just matches people randomly rewards skill because high skilled players will get higher win rates. A fully fair matchmaker will reward participation because if you match players (or clans) with similar strength against each-other they will all achieve ~50% win rate.

You demonstrate very strong opinion on what a team competition is supposed to be.
When there is a weak player in a team of Tennis, Judo, Figure skating, Warhammer 40k, etc ... (basically any team competition, that consists of individual 1v1 matches or sub-competitions) and he gets 0-20, then he is not provided with some easier matchups. But you suggest that weak players should have easier matchups in the team competition of Clan Wars.

All the above mentioned have active players in hundreds of thousands or millions. We have hundreds. Keeping as many players as possible should be the highest priority for the game. This is why I think that we shouldn't have a system that frustrates players too much.

Edited 9/27/2023 10:22:45

You can’t have both fair matches and no game found if not enough players join.
Keeps giving games even if there is a big rating difference.
Maybe there should be some sort of threshold for matching and deciding a random matchup based on similar ratings? Obviously this could be abused by deliberately losing to get easier matches

I was actually thinking about a system that keeps track of previous matches and stops players from being constantly on the same side of advantage/disadvantage. If a clan matching system is kept in place, it too should look for things like that. Matching a clan constantly against a stronger clan because they are #2 in every single time slot might reduce their win ratio in an unfair way.

let's try remember the hundreds players, that we have lost in trying to fix it in the past...

so the few can just do what they do, without any improvement

Anyway, as a bad player myself, I am all in to change the CW system so that it would give me easier matchups.
I can also vote for introducing easier newbie friendly templates (like "Duel Lotto") to the CW pool.

Appreciate your effort, thank you.
Looking forward to the results of your better matchmaking MVP.

@5s i really hope that wasn't a serious suggestion...
SEAD is newbie friendly enough

@5s i really hope that wasn't a serious suggestion...
SEAD is newbie friendly enough

It is serious in the way, that it would really improve my personal experience of CW.
It will make games easier and will improve my winrate (or make it easier to get the same winrate as now in SEAD).

Cant say for all the others, but I think that the expected winrate that would make me personally most happy and engaged is like 75-80%. I see that the stats show the most engaging WR for others to be ~50%. Maybe I am special.
For the consistency of overall matchmaking (it is impossible for everyone to have 80%), I would agree to be matched with players with high boot rates (I think it is fair, that those who boot a lot will have lower winrate to allow me to get my desired 80%).

Maybe Fizzer should code smth like this

if player_1.name == 'FiveSmith#1':
    player_2 = sorted(player_list, key=lambda x: x.boot_rate, reverse=True)[0]

It is in Python, but if Fizzer agrees to implement it, I can rewrite it to any other language.

Edited 9/27/2023 11:23:51