Now with these in mind, let me tackle again your initial 5 bullets on your issues with the current cw matchmaking:

1. Not finding matches: Your statistics provided do not directly address this issue. Is it possible that people are leaving due to poor matchmaking? Yes. Is it proven anywhere in your statistics that players cannot find matches? Nope.
2. Separating clans: Again, not mentioned in your data. The statistics on player win rates and days played do not directly confirm this. You would need sth closer to clan vs clan matchups in order to validate your claim
3. Basically all your statistics are for this
4. Your statistics do not touch this issue at all
5. Your statistics do not touch this issue at all

So for your 3rd point my counter arguments are:
- Individual CWR might not represent all facets of a player’s contribution, as they might excel in s2v2 but have a lower individual skill rating
- Implementing a new matchmaking system would, not could, introduce complexities and unforeseen issues
- You hypothesise that individual CWR will prevent island creation, as you phrase it, but it could still occur if players with particular CWRs frequently match against each other
- Upon joining a clan you agree and accept to work on a variety of skill levels. That is a fundamental part of the clan-based system. By pinning the blame on the CWR system or the matchmaking algorithm regarding it, you overlook the purpose of the clan as an entity
- In your random matchmaking proposal, strong players would often get matched vs weaker players, leading to one-sided games. This will demotivate new or casual players from participating.
- In a player-based CWR system there is potential for smurfing or artificially lowering one’s CWR to get easier matchups. Quoting Mylo i think? Someone said it earlier.
- Would rewards be individual-based or clan based, in a player-based CWR system for Clan wars? Genuine question.
- Player-based CWR system would have an absurdly big and unhealthy impact on clan dymanics and would foster competition inside the clan. I would suspect that some players would prioritize their personal growth over clan success, affecting clan dynamics
- Revamping the whole thing is also a technological challenge, so its highly questionable if it is feasible in the first place. + the bugs and server issues coming with it
- A weighted random matchmaking is kind of predictable. If a player can estimate the weighting mechanism, they would manipulate their performance to remain in the same range and avoid better players


And let it be known that i've acknowledged that i don't support the current system. It has its quirks. Matchmaking feels off sometimes, mostly when i lose lmao, and yeah i've seen my fair share of uneven battles. But that doesn't mean i am just going to jump on any change thrown my way.

If you are tossing out new ideas you should expect me and everyone currently playing to question its integrity and pick it apart. And up until now I am very much not convinced your solutions would solve the 5 problems you initially stated. Neither of the 5

Edit:
I didn't want to finish like this as it feels like a personal attack. For the 5 things you mentioned in your initial post for the matchmaking you would need statistics on:
1. A log showing instances when there were sufficient players online but there was a ng. Maybe also what happened to those players. How many are there, from which clans, is there an underline pattern or some distribution. Is the number big enough for the community to care.
2. A match history for how often each clan faces each other over a given period, including template selection.
3. Player score distributions. You say there are players with 0-20 scores prove it. Get the 40 participants from harmony prime or idk throw them in a frequency distribution chart or some histogram and prove to us how many fall in that category and whether its significant.
4. Regarding fw i dont even know what to say you are very oblivious as to how they work and thank god people from other clans already talked about them cause when i did i am a biased Harmony player
5. We all now meta strategies are true but they will occur regardless of the system, will just have a different nature

Edited 9/28/2023 05:29:40

1. you are correlating these results with good and bad matchmaking but you are listing winrate and people that stopped playing. Do you imply that losing a game = bad matchmaking and winning is the opposite?

No. I am implying that if a matchmaker matches people with similar strength, the outcome will be ~50% win rate for all players. Which is a fact. When I say good matchmaker I mean one that matches people with similar strength, so an ideal matchmaker will have all win rates close to 50%. Given our small number of players, it will never actuall reach 45%-55%, therefore some leniency must be calculated.

4. We are missing the sample size. If some of your categories have significantly fewer players than others, the percentages are skewed

The sample size is 702 players outside of the 33-66 range and 1331 players within that range.

5. Correlation does not mean causation. You indicate a correlation between win rate and player retention. We dont know if matchmaking is the direct cause of these dropout rates

I am not claiming that the matchmaker is the direct cause of the dropout rates in total. There might be other factors. However I do say that since it is clear that the dropout rates are hihger in the 'bad matchmaking' batch than it is in the control batch (the 33-66 batch), we can reduce the dropout rate if all players are in 33-66 category. The total dropout can be influenced by anything. The dropout difference between the separate win rates is either a direct result of the win rate differences, or there is a third factor that causes both the win rate difference and the drop out difference. We know that the matchmaker has an influence on the win rates, therefore fixing the matchmaker will at least partially fix the measured difference and improve player retention. It will not magically retain 100% of the players, just slightly higher %.

6. Your decision to categorize percentages as you have currently is arbitrary, for all we, viewers, know. You say that players outside the 33-66% wr have higher dropout rates but is that difference significant enough to be attributed to matchmaking

Actually the categorization isn't arbitrary, it's based on past research done on humans and other mammals, which suggests that mammals stop playing games if they lose more than 2/3 or win more than 2/3. There is other research done by game developers that shows that the best win % for player engagement is 50%. I didn't come up with this, I just verified if it hold true for this specific game that we are playing. I have also shown some of the data in 20% increments. The issue is that the more categories I make, the sample size becomes smaller, therefore the 33/66 is the best as it has the highest sample size per category so outliers will have less influence on the results.

4 and 6 come together in the essence that you provide nothing on the statistical significance of your analysis. The differences in percentages, while somewhat small, might be statistically significant if the sample size is large. If the sample size is small they probably are not. Regardless, we do not know

I gave you the sample size above, plus the sample size is all players that participated, so it can't get any bigger. The difference between the bad and good matchmaker group is 8% lower player retention, that is not statistically insignificant. For the result to be statistically insignificant, the measurement error needs to be larger than the difference. That is not the case here.

1. experience: perhaps some in the 80-100% bracket left because the game is too easy. and vice versa players in the 0-20% found the games learning curve too steep. These have nothing to do with matchmaking

That is exactly the reason I think they are leaving. But because it's PvP, the matchmaker determines how difficult each game is. A player with 0-20 win rate is consistently being matched with players who are vastly better than them, making the game seem to difficuly, while a 80-100 win rate player is consistently being matched with players who are vastly worse than them, making the game have no challenge.

2. longevity: How long were these players active before leaving? Retention might correlate more with longevity and novelty if the people that left in the 80-100 bracket played for years, and in the 0-20 bracket played for some weeks

I have provided both the number of games played and number of days played for all groups before, you'll have to go back to page 3, it's my 3rd post.

- There is selection bias as this is for people who played more than 10 games

If a player played 3 games and won all 3 by accident, they would show up with 100% win rate, but would give us no data.

- Just because players with certain win rates played an average number of games doesn't mean their winrate caused them to stop playing or continue playing. correlation is not causation

This would be true in a standalone case. But we know from past research done in different games that win rates close to 50% do result in better player retention in general. Showing the correlation in this game is enough to show that this game isn't an exception to the rule.

- the 198 days is so vastly different than the rest its either what i presume, being influenced a lot of accounts with 10-20 games only, or is actually statistically significant in which case those players found the game too easy or faced a burnout.

80%-100% - 45 games played on average
80%-100% - 198 days

1. Not finding matches: Your statistics provided do not directly address this issue. Is it possible that people are leaving due to poor matchmaking? Yes. Is it proven anywhere in your statistics that players cannot find matches? Nope.
2. Separating clans: Again, not mentioned in your data. The statistics on player win rates and days played do not directly confirm this. You would need sth closer to clan vs clan matchups in order to validate your claim
3. Basically all your statistics are for this
4. Your statistics do not touch this issue at all
5. Your statistics do not touch this issue at all

We are too early in. I haven't tested different matchmakers at all. I am planning to do it soon, then we will address each of these. Btw, sorry for not going into detail on the rest of the points you make in the post, as I say I am too early in to address them. Until I can what different types of matchmakers will result in, I can't even begin to say how which one will actually affect the outcomes. My current plan is to test different matchmakers and then measure the differences in player win rates, clan win rates, matches found, etc. Then we can see about how each could be abused and how (if) it can be prevented.

- Would rewards be individual-based or clan based, in a player-based CWR system for Clan wars? Genuine question.

Of course they would be still clan based. I am not proposing changing how clan wars work, all I am proposing is for Fizzer to figure out if he wants to reward participation or skill. If he wants to reward skill, a full on random matchmaker is the way to go. If he wants to reward participation then a full 50-50 matchmaker is the way to go. If he wants to reward both, then something in between would be good. The problem right now is that I don't even know what Fizzer considers as the goal and what most of the players consider as the goal, as no one seems to be answering this when I ask.

If you are tossing out new ideas you should expect me and everyone currently playing to question its integrity and pick it apart. And up until now I am very much not convinced your solutions would solve the 5 problems you initially stated. Neither of the 5

I haven't actually given any solutions yet. So far I was trying to prove that the problems exist.

Here are the different matchmakers I am planning to test:
1) the current matchmaker (CWR based, strong to weak)
2) CWR based matchmaker that prioritized finding more games and distributing free wins randomly instead to the weakest
3) CWR based matchmaker that uses per player ELO if multiple people from the same 2 clans need to get matched
4) fully random matchmaker that only prioritizes finding as many games as possible
5) per player ELO based matchmaker

I am planning to measure the following:
- predicted player winrate distribution
- predicted clan winrate distribution
- predicted clan points
- free win distribution
- percentage of no games found per clan and globally

The reason why I went into finding the data I did was to establish that the ELO I am using is good enough for predicting winrates for the experiment to have any merit. I am not even planning to force any of the matchmakers to anyone, just want to provide all the possible alternatives and see if Fizzer or the playerbase would find any of them more palatable than the current one, or if the general agreement remains that the current matchmaker is the best.

4. Regarding fw i dont even know what to say you are very oblivious as to how they work and thank god people from other clans already talked about them cause when i did i am a biased Harmony player

Great, I have that mistake in OP why does every 5th post have to be about that? I have already acknowledged when the first player pointed it out, and didn't refer it it in any of the later posts in that way. Also, I am sorry to inform you, but we are below you in CWR now, so if anything, I am from a free win clan now, not you anymore.

Edited 9/28/2023 05:59:43

Not much to say, if you have done your math behind the scenes good luck on your journey.

My only note is I don't think its predictable or clear that changing the matchmaking entirely and rebuilding it from scratch would reduce the drop-outs. From Fizzer's perspective its very risky to make large system changes based on correlation. By large system change i mean the revamp of the CWR system. Smaller changes in ng/fw, or in general small changes in the pre-existing system i presume are easier to be implemented.

1. A log showing instances when there were sufficient players online but there was a ng. Maybe also what happened to those players. How many are there, from which clans, is there an underline pattern or some distribution. Is the number big enough for the community to care.

You can see this for yourself, this happens very often. For example here is from the last few timeslots:
https://www.warzone.com/Clans/War?ID=37&Timeslot=5594
Everything good.

https://www.warzone.com/Clans/War?ID=37&Timeslot=5593
Small Earth Commander and Bomb Card
Players without a game:
PuppyManBread
Wildcard
Alphazomgy (Warzone's Best Marketer)
9 games were made, total number of people from TLA was 5. A different matchmaker could have made 11 games and give 1 free win to one person, instead the one we have made 9 matches and gave no matches found for 3 people from the same clan.

https://www.warzone.com/Clans/War?ID=37&Timeslot=5592
https://www.warzone.com/Clans/War?ID=37&Timeslot=5591
https://www.warzone.com/Clans/War?ID=37&Timeslot=5590
Everything good.

https://www.warzone.com/Clans/War?ID=37&Timeslot=5589
Strategic 1v1
3 games were made, total number of people from MB was 4. A different mathcmaker could have made 4 games and all players would play, instead we have 3 matches and 2 no matches found for 2 people from the same clan. You could argue that MB is too strong to be matched against the weakest players, but that already happened in that timeslot. Instead of having:
Python vs OP
MB vs Lu Fredd
MB vs VIW
2x MB no game
We could have
Python vs MB
OP vs MB
MB vs Lu Fredd
MB vs VIW
And everyone getting a game.

Furthemore if we look at the players ELO in that timeslot, we got these matches (I ordered by player ELO, not CWR):
1600 vs 1457 (143) 69.5% vs 30.5% chance to win
1578 vs 1390 (188) 74.7% vs 25.3% chance to win
1464 vs 1298 (166) 72.2% vs 27.8% chance to win


But if we match the 4 mb players based on player ELO against the 4 other player we would get this:
1673 vs 1600 (73) 60.4% vs 39.6% chance to win
1578 vs 1457 (121) 66.7% vs 33.3% chance to win
1464 vs 1390 (74) 60.5% vs 39.5% chance to win
1441 vs 1298 (143) 69.5% vs 30.5% chance to win

Not only would we get games for all 4 players, but those games would be against opponents that have strength much closer matching each other than in the actual output of the current matchmaker.

Unfortunately, I can't just go over all the games and pull up the stats because I don't have the clan compositions for every season, so I don't know which player was in which clan for what season. I do have the data for the last 2 seasons, and I am planning to use the previous season to compare matchmakers, because it is already completed so I will have the most data to compare.

Edited 9/28/2023 06:58:51

Hi! I am Texx! My rating is 1574 (guessing, but would actually like to know). If I joined that slot… no, wait! I wouldn’t join that slot! Because if matchmaking was individual, say goodbye to Texx, say hi to my little brother “xXGelatinousGlizzyGobbler420Xx” who just spent a clan wars season in The Hodopian Dynasty surrendering every game. His ELO is now rock bottom and he’ll spend the season in Harmony crushing every noob he gets matched up with. He’ll end the season 24-0 and never play again. Because the following season, say goodbye to xXGelatinousGlizzyGobbler420Xx, say hello to my other little brother “OooOOooBABYaTriPle!!!” who also just spent a season in THD surrendering every game.

Sounds like Hodopian Dynasty is a great training clan!

and how come this is not a prove of rigging cw..
sometimes I wonder on your thoughts-processor

pirtzugno is right @Texx, you'd be playing with two accounts in one season, what you need is a roster of 80 players and rotate them each season.

Hi! I am Texx! My rating is 1574 (guessing, but would actually like to know). If I joined that slot… no, wait! I wouldn’t join that slot! Because if matchmaking was individual, say goodbye to Texx, say hi to my little brother “xXGelatinousGlizzyGobbler420Xx” who just spent a clan wars season in The Hodopian Dynasty surrendering every game. His ELO is now rock bottom and he’ll spend the season in Harmony crushing every noob he gets matched up with. He’ll end the season 24-0 and never play again. Because the following season, say goodbye to xXGelatinousGlizzyGobbler420Xx, say hello to my other little brother “OooOOooBABYaTriPle!!!” who also just spent a season in THD surrendering every game.

Hi Texx with 1477 rating who would win against me and my measly 1349 67.63% of the time. There is a thing called K in ELO calculation. It is used to measure how certain the system is about your rating. If you try to abuse the system, your K number will grow and you'll end up at your real elo very quickly. For example with K = 320 (which is 100% uncertainty in my system), you would get from 600 ELO to 1477 within 5 games. It would also take 5 games for your K to grow that much, so enjoy your 8-9 wins, hope making an account and losing ~40 games was worth for it. The system will also easily detect alts like that because their predicted number will vastly vary from their actual results. Banning abusers would be very easy.

pirtzugno is right @Texx, you'd be playing with two accounts in one season, what you need is a roster of 80 players and rotate them each season.

You do realize that you can do exactly that in the current system? Clan CWR stays from season to season. If you have 40 people that move over from clan A to clan B each season, you can have a clan that loses every single game and then next season wins a lot. All you need is to have clan A be the winner clan this season and then B be the winner clan next season. You just move over the accounts that need the rewards from the designated winner clan to the designated loser clan each season.

So you're saying your system doesn't fix anything?

I'm mostly kidding, and I believe if the old timers aren't really invested in your proposals is that we've been here before. Frenzy about how the matchmaking should be changed happens every few months and never went anywhere. Plus It's easy to poke holes in the current system, finding an acceptable alternative isn't.

Plus your solution is for Fizzer to implement a rating system without messing it up? Good luck with that.

So you're saying your system doesn't fix anything?

I'm mostly kidding, and I believe if the old timers aren't really invested in your proposals is that we've been here before. Frenzy about how the matchmaking should be changed happens every few months and never went anywhere. Plus It's easy to poke holes in the current system, finding an acceptable alternative isn't.

Plus your solution is for Fizzer to implement a rating system without messing it up? Good luck with that.

I keep saying this, but here we go again. I don't have a solution yet. I want to measure the outcomes of different matchmakers and see which one would be the best for what purpose. I wrote up these 5 that I will test:
1) the current matchmaker (CWR based, strong to weak)
2) CWR based matchmaker that prioritized finding more games and distributing free wins randomly instead to the weakest
3) CWR based matchmaker that uses per player ELO if multiple people from the same 2 clans need to get matched
4) fully random matchmaker that only prioritizes finding as many games as possible
5) per player ELO based matchmaker

But I am actually interested in any other suggestions for testing. Once I am done, I will publish what each matchmaker can do, and if one actually seems better than the rest with the limited player base we have (and this is a huge issue for any matchmaker), then I will start writing mails to Fizzer. Not before that.

Fun fact: with 1 exception, all the people defending the current system in this thread are from the 33%-66% win rate group, and that one exception is in the 66%-100% win rate group. How funny that the people for who the current system works as a system should work for everyone thing the system is good.

This is not a smart thing to say. You will struggle to design any matchmaking system where I wouldn't win more than 66% of the games.

The match making system itself might not need to change, only the rating system. Player-based ratings show better reflection of skill than a overall clan rating so matches are inherently fairer. Match making could be done same way (highest to lowest), but instead of ordering clans, orders players instead and clan mates dont get games with each other (at least for 1v1s).

Edited 9/28/2023 10:25:24

WarTog am always puzzled by suggestions and how to improve on Clan Wars, because they almost always come from players who don't seem to understand what Clan Wars is. And it is hard to know what to do with suggestions primarily aimed at turning CW into a completely different thing that it never set out to be in the first place.


First, and most important, one must understand that Clan Wars is clearly not intended to be the Olympics of WarZone. There are plenty of highly-competitive Clan Ladders and Clan Tournaments and individual events that are specifically designed to weed out the chaff and crown the Greatest Athlete(s) of WZ.

Clan Wars is not that. At all.

As peculiar (and sometimes puzzling) as Clan Wars is, it is actually a surprisingly well-designed game when you understand that Clan Wars was specifically designed to create an activity that ALL MEMBERS OF A CLAN can play together as a team. A single Clan Wars team can include one of the All-Time Great Champions of WZ and the newest of noobs who just joined two days ago - and everyone in between - and they can all play together and contribute to their team's success in their own way.

Clan Wars is NOT intended to be an Olympian event where only the world's greatest athletes should even think about competing.

Clan Wars is much more akin to The Boston Marathon. Where thousands of people compete. And yes, it does take the form of a competition. And yes, they do announce a winner and give them a prize. But that is NOT why thousands of people take part. Most runners (and spectators) are just there to have fun. Challenge themselves a little. Maybe break their own personal record. Maybe beat their best friend's time. MOST Bostonians probably could not even tell you who actually won the marathon. Because determining-the-greatest-marathoner-in-Boston is *not* the point of the Marathon. Quite the contrary, it is simply a fun activity that the whole town can take part in. Nobody actually cares who the winner is. (Except maybe the winner. And good for them!)

So if someone were to start suggesting changes to the Boston Marathon that would make it "better" at accurately determining a winner - (Special lanes for the pro runners? Give the elite athletes a 10-minute head start so they don't have a bunch of randos bumping into them at the starting line? Maybe establish restrictions to eliminate runners who aren't "serious"? (I mean, seriously, why are there wheelchairs out there?)) - such suggestions would be a complete waste of everyone's time, because they completely miss the whole point of having the Marathon in the first place.


And Clan Wars is very much the same way. It is NOT about the Winners. Yes, some winning happens, cuz it stokes a fun competitive spirit. But Fizzer did not create Clan Wars because he felt that there just weren't enough ways in WZ to honor how awesome Flexus is.


To clear up one of the biggest misconceptions I see over and over: CLAN WARS is absolutely NOT designed to honor the team with the BEST CWR!

If it *was* then First Place would go to the team with the best CWR! Python would win every season, and probably solved. But clearly Clan Wars is not about obtaining the highest CWR, so all these are idea about how to create a more 'accurate' CWR are ultimately pointless.


It seems pretty clear that the CWR exists solely as a mechanism for ensuring that Clans are *generally* matched with opponent Clans of comparable average skill level.

But it is meant to be a blunt instrument. Not a surgical tool. This appears to be intentional.

The current CWR system is designed to *mostly* match Clans with comparable competitors - so we don't see Kittens getting constantly stomped by MythBusters - but it is also intentionally designed to mix things up a bit. That's why the templates are random and the names of the opponents are hidden. CW *wants* the occasional mismatch or to force a player to play an unfamiliar template.


So as well-intentioned, and well-meaning as many suggestions for improving the Matchmaking appear to be, on the surface, they almost all miss the point that CW is *not* shooting for smooth and fair and predictable matches. It's meant to be a little playful and rough-and-tumble.


One of the most wildly misguided of the well-meaning suggestions is the idea that CWR should be by Individual Player rather than by Clan. This sounds unquestionably fair, but is also, I'm sorry to say, an incredibly terrible idea.

The obvious reason is that if CW switched to an individual CWR - you would see a lot more of the same players matching each other over and over. (Which, see above, is not the goal of CW)

The second, perhaps less obvious reason, is that it would be devastating for all of the smaller clans. There are many many clans that are essentially built around one experienced player (usually the Clan Founder) and a couple of their friends plus a bunch of noobs that they are training to get better. If you strip out a fledgling Clan's only reliable winner and dump them into a talent pool where their WinRate will automatically drop to 50%, you will inevitably crush any hopes of a bright spot in that clan's day.

You want all the players on a team to be competing in roughly the same talent pool, so that your team has something in common. My Clan (Harmony) is currently in a close race with Prime (who also have a similar CWR) So very often, when we go out to play, our whole team draws Frogs as their opponents. This is a good thing, because we get to be a little competitive with each other to see who beats our rivals, or who beats them fastest, or etc. The sense of camaraderie and shared purpose simply wouldn't be the same if Texx was always fighting an MB and Old Yeller was fighting a Prime and WarTog was fighting a Clanzilla.

If you take a microscope to CW matchups on a case by case basis, it's easy to argue that there is something unfair about this pairing or that pairing. But when you take a step back and look at the results in the aggregate, the current CW system works surprisingly well.

Thanks Wartog. Straight to the point!

This is not a smart thing to say. You will struggle to design any matchmaking system where I wouldn't win more than 66% of the games.

Super easy. Barely an inconvenience. A player with 1407 ELO will have 33.38% win rate against you. If I only match people who are above that, you can't have over 66% win rate. There are currently 107 players that are active in clan wars with ELO higher than that. With 6 timeslots per day and less than 4 maps per timeslot on average, that means 4.45 or more of them in every single map in ever single timeslot. If the average participation is 25% or more, that means you'll have 1 such player every time you go into clan wars. All the matchmaker needs to do is match you with that guy. Want a less than 50% win rate? Can do that too. There are 43 players that have over 50% chance of defeating you. There you have a chance to avoid them by playing at odd maps in odd timeslots, but I could in theory make a matchmaker that would give you no game, unless one of those are present and there you are sitting below 50%. It's not a matchmaker I would do nor once I would think would be good or fair, but I wouldn't struggle to make it.

@WT
Yes its for fun.
But if you're going to have fun, you don't want to always obliterate your opponents. Individual CWR fixes that. And by ensuring CW doesnt match same players more than a fixed % of the time, its fine... just my immediate thought lol
For those who find it ok already, nothing wrong with switching to individual cwr? it wouldn't affect your own winrate too much...


then again im not an expert on this and i haven't even bothered to read the whole thing yet


and is there not something wrong with players not getting matches when they easily could have?

Edited 9/28/2023 11:09:49

if it's all fun and games you should get rewards. Oh is there something on the line players are competing for? that makes sense to consider this an issue.