I looked at the original algorithm.

0) Join all players in the given template in a pool. Remove them from the pool once they are matched.
1) Update the pool. Let N denote the current size of the pool.
2) If the pool has players just from one clan, give one free win and end.
3) If the biggest clan has at least N/2-1 players in the pool, GOTO 5).
4) Match a player from the highest-ranked clan and 2nd highest-ranked clan. GOTO 1)
5) Match players from the biggest clan and outside that clan, give one free win if at least one player reminds, end.

I re-wrote it as a while loop, as that's more accepted than goto statements. Also, there is an accommodation where largest clan is greater than N/2. I imagine other things to check for as well.

0) Join all players in the given template in a pool. Remove them from the pool once they are matched.
//Initialize
While N>2
1) Update the pool. Let N denote the current size of the pool.
2) Sort by clan size.
3). If largest clan size > N/2 then remove one from the pool from that clan.
Else, if the largest clan has at least N/2 - 1 players in the pool, then Match players from the largest clan and outside that clan.
//. Largest Clan Match
4). Else, match a player from the highest-ranked clan and 2nd highest-ranked clan.
// Top Down Match
End While
5). If N=0, End
6) If N=1, Award free win, End.

This matches the largest clan first, if it is greater than N/2-1.

I point out that a random draw from the pool is just as effective in maximizing games, if that is the goal.

Of course, the "award free win" will create more free wins than today. Note that is severable from the algorithm.

To put it in simpler terms, καλλιστηι proposition was;

1) Every template will have their own player pool [Already is]
2) If half or more players in a pool consist of 1 clan, match players from that clan first
3) For the remaining players, match them according to CWR (top to bottom)
4) If there are players with no match, give them free wins (1 free win per clan in a given template)

Positives:
- This algorithm is more effective at giving everyone match

Negatives:
- This will increase stacking* problem (by offering them free wins)
- This will cause unfairer matches (If avg-CWR clans stack, low-CWR clans will have to match avg-CWR clan instead of other low-CWR clans)

However, we can avoid 1st negative and mitigate the 2nd with slight tweaks. The optimized version of καλλιστηι's system looks like;

1) Every template will have their own player pool [Already is]
2) If half or more players in a pool consist of 1 clan, match players from that clan first (prioritize higher-CWR clans when matching)
3) For the remaining players, match them according to CWR (top to bottom)
4) If there are players with no match, give them +1 participation* (1 "+1 participation" per clan in a given template)

---

*Stacking: Players from a clan intentionally tries to focus on a single template to match easier opponents
**+1 Participation: Player with "+1 Participation" will be able to join 2 timeslots in a day instead of 1 (for 1-time only)

I defined the following relation. In a specific situation, algorithm A is better or equal to B if it matches more players. If both algorithms match the same number of players, then A is better or equal to B, if the delta of A is smaller or equal to B.
My proposed algorithm is better or equal to the current one in all cases (dominates it).

I’ll take the other side

1. It’s a good thing to diversify players across all available templates and all available timeslots so that others have an improved chance at finding an opponent to play their favorite game. The way that free wins and matchmaking are set up today discourages stacking and encourages clans to spread themselves across multiple templates. I believe this is intended behavior.

2. There are limited opportunities for a strong player to play real time games against an equally strong opponent in the base game. This is honestly the single biggest draw of CW to me because this need is absolutely not met by QM. Why should this strong player be denied the opportunity to play a comparable opponent because 20 players who happen to all be in the same clan all demand to play one specific template at one specific time?

3. For competitive reasons, the top clans should play against each other. The seasonal title should be settled by El Clasico, not who beat the shit out of Las Palmas by more goals. You can’t complain about walruses and simultaneously advocate for this change.

4. Clan Wars is set up to be a holistic competition of clans. “Specialist” clans who play exclusively one template shouldn’t be rewarded with free wins. They should be encouraged to learn other templates.

That sounds amazing until you notice that if your clan rating is low enough, two people on a template with 50 opponents is enough to not give one of them a game.

In my suggestion if a bad clan joins a template too much, they will be matched with stronger opponents - it still discourages stacking. Except in a less retarded way.

Edited 4/15/2023 14:10:57

@TheGreatLeon

It’s a good thing to diversify players across all available templates

▶ Yes, still waiting for καλλιστηι to tell why he wants to increase stacking, instead we got "My proposed algorithm is better or equal to the current one in all cases (dominates it)".

There are limited opportunities for a strong player to play real time games against an equally strong opponent in the base game

▶ Yes, again. Moreover, when stackers are the primary focus, they'll have to go against best players more often. Even though καλλιστηι's proposition is designed for stackers getting more games, even they will start to criticize καλλιστηι's proposition.

For competitive reasons, the top clans should play against each other

▶ Another yes, and best way to ensure this, is through current system.

@καλλιστηι

two people on a template with 50 opponents is enough to not give one of them a game

▶ It's very rare (because that happens only to clans with lowest CWR and they are usually free-win clans who likes no game). Again, I would rather see 1-2 players needing to participate another timeslot than 10-20 players automatically lose because they get players above their skill level

it [καλλιστηι's proposition] still discourages stacking

▶ Technically you can argue that your proposition have one aspect that can discourage stackers (if you ignore all aspects that encourages stacking), but even that's through ruining both stacker's and competitive player's game.

Again, I would rather see 1-2 players needing to participate another timeslot than 10-20 players automatically lose because they get players above their skill level

The only time when 10-20 players would be matched with players above their skill level would be if these 10-20 would get no game in the original matchmaking system. It replaces "no game" with "game against stronger clan".

▶ Yes, still waiting for καλλιστηι to tell why he wants to increase stacking, instead we got "My proposed algorithm is better or equal to the current one in all cases (dominates it)".

▶ Technically you can argue that your proposition have one aspect that can discourage stackers (if you ignore all aspects that encourages stacking), but even that's through ruining both stacker's and competitive player's game.

1) the game never tells players, that if they join a template with "20 players" and with "5 clanmates" getting no game is a possibility.
2) Clan rating is not the correct metric to determine players' strength. Clans have stronger and weaker players. The match of 100 rated vs 200 rated isn't that much unbalanced than 100 rated and 101 rated.
3) Players should be allowed to play templates they enjoy, not strategize which template to play to maximize wins.
4) Currently the game denial is based on luck. You don't know how many players of worse clans join CW.

4. Clan Wars is set up to be a holistic competition of clans. “Specialist” clans who play exclusively one template shouldn’t be rewarded with free wins. They should be encouraged to learn other templates.

That is already an issue of the current system. Get a high enough rating and you are guaranteed to have all games. Get enough players on the template and you are matched with noobs *cough* *cough* MB *cough* *cough* MME.

▶ It's very rare (because that happens only to clans with lowest CWR and they are usually free-win clans who likes no game).

There is a difference between a free win and a no game. Free win means, that you get a win for free. No game means, that you don't get a game (and hence no win and are allowed to join the next slot). With no game, you basically take time to be online at a specific time just for the RNG god tells you fuck off and try next time. It happens rather often; I didn't do statistics but it is not hard to find an example of no game in 6 CW slots.

Edited 4/15/2023 15:01:35

I wonder if a potential alternative could be to keep the current system but disclose the clans queueing for each template. This could potentially add more in terms of meta strategy as well.


You’d still have a problem with players backing out of templates in the last second but honestly I think that’s kind of funny. One solution for this problem could be to lock players who have committed for the final 30 seconds or whatever.