Rejected GBG matchmaking and league system changes

[Donna a Macskabolond]

Reason

Actual problems:
- too much guilds in the top league which results too big difference between the strongest and weakest
- active fighter guilds are matched with slow, sleepy, less active guilds
- nothing to fight in 1000 LP diamond league besides the personal rewards

How to identify active, fighter guilds?
- have more strong players than the average
- have the necessary treasury for building
- higher overall activity, organized and clear guild hierarchy
- number of players

These attributes will result in a higher average advance number/player than most guilds. If the guilds have problems any of the above 3 points it will result lower average advance/player. One more thing has impact on the advances what a guild can do in a season - the other guilds in the group. If not one of them bothers to take any sector than the guild can't have as much advances as it capable of.

Details

My suggestion to sort the guilds for matchmaking based on the advances/player of the previous season. For example if a guild have 155000 advances in the season and have 74 members then the average advances will be 2094.

The bigger guilds has more potential fighting capacities, so this number should be multiplied with a number depends on the number of players.
The average guild size lets say is around 40 players, so the multiplier should be:
Number of players Multiplier
70-80 2
50-70 1,5
30-50 1
10-30 0,9
1-10 0,8
It is necessary to push the active, higher headcount guilds above than the active medium or small size guilds. Of course, if we have numbers probably the multiplier needs some adjustment.

Based on this calculation in the next season the guilds will be matched similar guilds in activity, strength and treasury, so it will result a fairer matchmaking. If seems necessary one more multiplier can be added based on the final placement in the group ( 3-4 should 1, above than this higher, 5-8 lower multiplier should be used) or the multiplier can be different in different leagues. I'm not sure if it is necessary, some experiment is needed.

Changing the matchmaking system is not enough on its own, some adjustment is needed in the league system also, further changes seem necessary:
1. A new league is necessary for the biggest, strongest guilds where prestige points can be earned or loose for winning - the 1. place should provide around 2000 PP, 2. place 0, 3. place -500 PP, 4. place -1500 PP. This new league doesn't provide additional league points (just
negative for the 3-4 place), the personal rewards are the same as in diamond. The guild rewards such as guild power and fragments should be higher than in diamond. Maximum 2 groups should start in each season. The league should be exclusive.
2. Change the number of guilds in each groups. The number of guilds should depend on the league as the following:
Crystal league groups have 4 guilds
Diamond league guilds have 4-6 guilds
Platinum and lower leagues guilds have 6-8 guilds
3. Change the provided league points as @Juber suggested - either use a soft cap or use progressive LP (the higher the league the less the earned LP)
4. Shorten the seasons. The actual 11 days way too long, the final results can be predicted usually on the first day. The continous activity takes its toll on the GBG leaders and the members too. 6 days seasons with 8 days rest period would be enough.
5. Add two slots for building on HQ sector. The price should be high - about 50000 guild goods because these building can't be lost. It provides some chance to take some sectors for the smaller guilds.

Balance

The changes does not affect other parts of the game.

Abuse Prevention

The solution shouldn't open any exploit.

Summary

It is obvius now the GBG needs a serious reimplementation. The matchmaking and league system causes lots of problems, imbalance, complains, overall unsatisfied feelings on the players part. The suggested changes result in a more balanced matchmaking and league system while those guilds who has the means to fight get the chance to do so and those who feel a slower pace comfortable are matched similar guilds in activity.

Better matchmaking and more suitable league system increases the overall activity, the shorter seasons prevent burnout and also increase the activity.

Have you looked to see if this has already been suggested?

The matchmaking changes hasn't been suggested yet, some elements regarding the league changes comes from @Juber suggestions.

GBG matchmaking and league system has some serious problems what results in
- uneven groups where 1-2 bigger guilds dominate the map, slower, smaller guilds has no chance to fight
- lots of complaining
- in 1000LP there is nothing to fight for - GBG has no impact to global ranking
- GBG takes too much time which leads to burnout, lower overall activity

I'm happy to incorporate any suggestion which solve the GBG problems and feel free to add any comment regarding the ideas.

 

[Yekk]

Odd. Your battle curve shows that you battled more this past season than the one before. Same with most of the other top players in your guild.

I was in a stronger guild...next

 

[Velvet Thunder]

As polls are not compatible with this new idea software, those who wish to cast their votes may do so using the options at the right of the first post. The upper arrow adds a vote (+1), and the lower one downvotes (-1).

 

[xivarmy]

Didn't vote either way because of too many concepts in one post (some of which sound like a good idea, and others sound pointless/confusing).

Regarding sorting by advances in the previous season, the issue with this is that the new guild coming up from platinum probably had *many* advances the previous season. If trying to do a further sorting, I would suggest instead it be Current MMR > Last Season's MMR > Last Season's Advances. This will pool the 1000s that stayed in 1000 together - and further, the ones that most dominated it. While the new 1000s will be paired with other new 1000s, and probably a few of the less strong 1000s that stayed 1000. And only the following season be thrown into the wringer.

Your prestige stacking has obvious issues in that it offers permanent prestige - so a guild that spends 50 weeks on top of the standings and then implodes, is still stuck in #1 on the server - on one of my servers for instance have GvG guilds that have years worth of "Top 1 days" count, but they're no longer relevant to the standings, which is an important feature. The correction should probably be as soon as you get knocked out of the special "extra-prestige-league", you lose it all. I overall do not think this an elegant solution anyways.

Overall for those two combined ideas purpose, I liked deadpool's suggestion in the recent thread better - remove the MMR cap, but remove 10% MMR from every guild each season. This would make the new effective maximum 1750 - and require a guild to always get 1st to maintain it. Maintaining 1250 would require always 2nd+. Maintaining 750 would require always 3rd+. Leagues would probably need to be redrawn to some extent with this in mind, but it'd do a better job of stratifying guilds and offering more prestige to the actual-best guilds than the also-rans.

 

[DEADP00L]

Being very fond of mathematics, I tried to find a solution to MMR that is adapted to all situations.
I only found 2 that are perennial:
- make an average of the meetings made on the 4 preceding GBG to group the guilds within a league.
- no longer limit to 1,000 LP but remove 10% from each guild on all leagues at the start of each GBG.

These two different calculation methods make it possible to regroup, in all leagues, the guilds according to a similar activity after 3 or 4 seasons.

 

[.Chris]

I still think that a fixed number of spots in Diamond, Platinum, etc. (possibly as percentages of total active guilds) in combination with a Multiplayer Elo rating system would be the best approach.

True skill by Microsoft is more complicated but - in the long run - more accurate:

Spoiler: TrueSkill

The TrueSkill ranking system is a skill based ranking system for Xbox Live developed at Microsoft Research. The purpose of a ranking system is to both identify and track the skills of gamers in a game (mode) in order to be able to match them into competitive matches. TrueSkill has been used to rank and match players in many different games, from Halo 3 to Forza Motorsport 7.

An improved version of the TrueSkill ranking system, named TrueSkill 2, launched with Gears of War 4 and was later incorporated into Halo 5.

The classic TrueSkill ranking system only uses the final standings of all teams in a match in order to update the skill estimates (ranks) of all players in the match. The TrueSkill 2 ranking system also uses the individual scores of players in order to weight the contribution of each player to each team. As a result, TrueSkill 2 is much faster at figuring out the skill of a new player.

Ranking Players​

So, what is so special about the TrueSkill ranking system? Compared to the Elo rating system, the biggest difference is that in the TrueSkill ranking system skill is characterized by two numbers:

• The average skill of the gamer (μ in the picture).
• The degree of uncertainty in the gamer’s skill (σ in the picture).

The ranking system maintains a belief in every gamer’s skill using these two numbers. If the uncertainty is still high, the ranking system does not yet know exactly the skill of the gamer. In contrast, if the uncertainty is small, the ranking system has a strong belief that the skill of the gamer is close to the average skill.

On the side, a belief curve of the TrueSkill ranking system is drawn. For example, the green area is the belief of the TrueSkill ranking system that the gamer has a skill between level 15 and 20.

Maintaining an uncertainty allows the system to make big changes to the skill estimates early on but small changes after a series of consistent games has been played. As a result, the TrueSkill ranking system can identify the skills of individual gamers from a very small number of games. The following table gives an idea of the minimum number of games per gamer that the system needs to identify the skill level:

Game Mode	Number of Games per Gamer
16 Players Free-For-All	3
8 Players Free-For-All	3
4 Players Free-For-All	5
2 Players Free-For-All	12
4 Teams/2 Players Per Team	10
4 Teams/4 Players Per Team	20
2 Teams/4 Players Per Team	46
2 Teams/8 Players Per Team	91

The actual number of games needed per gamer can be up to three times higher depending on several factors such as the variation of the performance per game, the availability of well-matched opponents, the chance of a draw, etc. If you want to learn more about how these numbers are calculated and how the TrueSkill ranking system identifies players’ skills, please read the Detailed Description of the TrueSkill™Ranking Algorithm or find out in the Frequently Asked Questions.

Source:

https://www.microsoft.com/en-us/research/project/trueskill-ranking-system/

https://www.microsoft.com/en-us/research/wp-content/uploads/2007/01/NIPS2006_0688.pdf

A simpler version is the Simple Multiplayer Approach:

Spoiler: Simple Multiplayer ELO

The Elo rating system is designed for two-player games, like chess.

It would be nice to have a similar rating system for games with more players. Unfortunately, generalization of the Elo system to n-player (multiplayer) games has proven difficult.


Simple Multiplayer Elo (SME)

Generalization may not be necessary, though. I've come up with a simple and effective way to apply the two-player Elo system to multiplayer scenarios:

• At the end of a game, make a list of all the players and sort it by performance.
• Think of each player as having played two matches: a loss vs. the player right above him on the list, and a win vs. the player right below him.
• Update each player's rating accordingly using the two-player Elo equations.

I call this method "Simple Multiplayer Elo" (SME) and am making it public domain.

If you decide to use SME, I would appreciate an e-mail!


Performance

To measure the performance of SME, I did the following experiment:

• Imagine 10 players with "true" Elo ratings of 1100, 1200, ..., 2000.
• Each player is assigned a random initial SME rating of 1500 ± 1.
• Multiple rounds of an imaginary game are simulated:
  • The players are assigned random game scores according to their true ratings. The scores have normal distributions with μ = true rating, σ² = 200.
  • The predictive ability of the SME ratings is calculated by enumerating the 45 possible pairings between the 10 players, counting the number of times the higher-rated player beat the lower-rated player, and dividing by 45.
  • The predictive ability of the true ratings is calculated using the same method. This is the ideal predictive ability.
  • The SME ratings are updated according to the SME method, described above.

I ran this experiment millions of times and averaged the results:

After round...	SME predictive ability	Ideal predictive ability
	50.0% (random)	84.5%
1	62.3%	84.5%
2	72.5%	84.5%
3	77.3%	84.5%
4	79.2%	84.5%
5	80.3%	84.5%
10	82.2%	84.5%
20	83.3%	84.5%
1000	83.9%	84.5%

You can see a graph of this curve here.

This experiment shows that SME only needs a small number of rounds to achieve near-ideal predictive ability, and is thus a strong candidate for any application that requires a multiplayer rating system.

Spource:

http://www.tckerrigan.com/Misc/Multiplayer_Elo/

 

[Velvet Thunder]

Due to a lack of support it has been rejected.