Reconsider Arena Rankings System

Discussions about BGA (all languages)
Forum rules
Warning: challenging a moderation in Forum = 10 days ban
More info & details about how to challenge a moderation: viewtopic.php?p=119756
User avatar
Lady_Jane
Posts: 48
Joined: 05 January 2024, 04:59

Re: Reconsider Arena Rankings System

Post by Lady_Jane »

ChiefPointThief wrote: 15 April 2024, 14:26 So the same person that I discussed before ended up winning arena by a comfortable margin even after playing poor for their 1st 104 games, average for their next 278, and above average only for their last 36 games.
MrsHayden wrote: 15 April 2024, 16:44 I don't think i necessarily have an issue with someone starting the season weak, improving, and ending the season in #1.
Sorry, wanted to clarify one more piece... My comment here is not referring to the numbers you gave. I don't see 36 of over 400 games as representative of skill increasing to the point of being best. This is just one of topics that's been discussed in this thread.
User avatar
ChiefPointThief
Posts: 745
Joined: 14 August 2020, 22:27

Re: Reconsider Arena Rankings System

Post by ChiefPointThief »

This system was designed for chess (an all skill game) but it is being applied here to all games with varying luck factors. If you have a 400elo and a 200elo player the rate at which the 400 elo player is supposed to beat the 200elo player clearly has to change due to luck factor of the game but that isn't accounted for. Another issue is the speed at which the system is reaching these levels due to the increased k factor in arena. If you take two players and one wins both their 1st two games and the other losses they would roughly be around 80pts difference. So just after 2 games into the season player X has to beat player Y at a 62.5% rate to break even. After only 2 games the rate jumps from 50/50 by 12.5%. Now when you consider that this game is also a game with a luck factor which win% are closer to 50% amongst all players that is just a flawed system.

I agree with you though I too am not that invested in trying to figure out a better system but that is more because I believe bga wants it this way and has no intention of changing it.
MrsHayden wrote: 15 April 2024, 16:44 what is strength of schedule?
More of a sports reference but I meant they played tougher opponents. I can several several top players in their game history last season.
1000esqrtn1
Posts: 24
Joined: 15 November 2023, 19:20

Re: Reconsider Arena Rankings System

Post by 1000esqrtn1 »

ChiefPointThief wrote: 15 April 2024, 18:24 This system was designed for chess (an all skill game) but it is being applied here to all games with varying luck factors. If you have a 400elo and a 200elo player the rate at which the 400 elo player is supposed to beat the 200elo player clearly has to change due to luck factor of the game but that isn't accounted for. Another issue is the speed at which the system is reaching these levels due to the increased k factor in arena. If you take two players and one wins both their 1st two games and the other losses they would roughly be around 80pts difference. So just after 2 games into the season player X has to beat player Y at a 62.5% rate to break even. After only 2 games the rate jumps from 50/50 by 12.5%. Now when you consider that this game is also a game with a luck factor which win% are closer to 50% amongst all players that is just a flawed system.

I agree with you though I too am not that invested in trying to figure out a better system but that is more because I believe bga wants it this way and has no intention of changing it.
Hi ChiefPointThief, just for curiosity, what is your opinion on a method similar the one I proposed some posts ago

The general formula would be something like this.
Top Players - (ex 300 + ELO)
Medium Players - (ex 200-299 ELO)
Low Players - (ex 199- ELO)
WT, WM, WL - Win Percentage Against Top/Medium/Low Players respectevily
GT, GM, GL - Value based on games played against Top/Medium/Low Players, This could be something like min(1, games / 20) to ensure a minimum number
KT, KM, KL - Factor to multiply wins against Top/Medium/Low

End of season points = KT*WT*GT + KM*WM*GM + KL*WL*GL


For sure numbers should be studied and some tries should be done, but do you see any clear flaw in that format? I think it mainly covers the main problems that we have been discussing on this post.
User avatar
ChiefPointThief
Posts: 745
Joined: 14 August 2020, 22:27

Re: Reconsider Arena Rankings System

Post by ChiefPointThief »

Can you give an actual example of 2 players using your system? Is their strength being calculated at the time the game occurred or by what their rank is at the end of the season? All games are calculated at the end of the season?
1000esqrtn1
Posts: 24
Joined: 15 November 2023, 19:20

Re: Reconsider Arena Rankings System

Post by 1000esqrtn1 »

ChiefPointThief wrote: 16 April 2024, 09:38 Can you give an actual example of 2 players using your system? Is their strength being calculated at the time the game occurred or by what their rank is at the end of the season? All games are calculated at the end of the season?
Hi, for implementation I think the easier way would be to compute the strength of your opponent when the match occurred. I would be keen to freeze rankings during an Arena, but on a 3 month period it is a little bit too much and would lead to unfair outcomes.
The formula could be updated after every game, as long as we keep track of games and % of wins on every level.

Let's make a simple example of a small arena between players A,B,C,D,E.

A and B start as low players, C and D as medium players and E as a top player.

They play 10 games against each other, A always lose, B beats C 60%, D beats C 80% of time and D beats B 60% of time. E beats B 80% of time, C 90% of time and D 80% of time

After this initial period they played a minimum amount of games and we have the following situation:
A vs low - 10 games 0 wins
A vs med - 20 games 0 wins
A vs high - 10 games 0 wins
A current score = 5 * 0 + 3 * 0 + 1 * 0 = 0 (considering as factors 5,3,1 for your wins against top, medium, low)
B vs low - 10 games 10 wins
B vs med - 20 games 10 wins
B vs high - 10 games 2 wins
B current score = 5 *0.2 + 3* 0.5 + 1*1 = 3.5
C vs low - 20 games 14 wins
C vs med - 10 games 2 wins
C vs high - 10 games 1 wins
C current score = 5 *0.1 + 3* 0.2 + 1*0.7 = 1.8
D vs low - 20 games 16 wins
D vs med - 10 games 8 wins
D vs high - 10 games 2 wins
D current score = 5 *0.2 + 3* 0.8 + 1*0.8 = 4.3
E vs low - 20 games 18 wins
E vs med - 20 games 17 wins
E vs high - 0 games 0 wins
E current score = 3* 0.85 + 1*0.9 = 3.45

At this point we could extract some situations. A and C are clearly underperforming, hence their last position in rankings. B and D have done pretty solid, winning games during the whole season both in their level and on higher levels.

Even if E has been the best performer, it could be considered that he has been farming, that's why he is not leading, even though that will be for a short period. E is penalized for not playing anyone on top level, however if B and D keep playing at a consistent win rate, they will increase level and E will compute games against top players, and if he keeps that win rate he will easily win the tournament.

Clearly to avoid the problem of E being exploited, low, medium and top levels should be defined in a way that ensures that players will play many games against each level.

Note that some big points are being avoided:
1 - Order of wins only matter when there is a player changing ranking, but in general WL and LW are equal
2 - Players B/D are not incentivized to stop playing, as their "Top Player" score is quite low and will easily be surpassed
3 - Farming against player A won't produce much extra points as the wins against low players have a limit on points awarded

Just a rough idea of how it could be

Thank you
User avatar
ChiefPointThief
Posts: 745
Joined: 14 August 2020, 22:27

Re: Reconsider Arena Rankings System

Post by ChiefPointThief »

How are players starting off as low, med, high? Are you using their elo or was that just for purpose of an example?

At 1st glance this seems flawed. It assumes that everyone can get equal matches vs every tier player and that is not the case. I play in some games with small communities in which there are only a few players on my level. Meaning that all the rest of the players have an extreme advantage by being able to play us and they only have to win a minimal # of games to equal the max I can get for playing them. What a strong player is also varies from game to game. How do players break off into med/high players in your system if they start as low/med players?

Player D having a .85 higher score than Player E is the 1st red flag. Sure player D played 10 games against hard opponents but they only won 2 of them. That doesn’t justify them being ahead of player E who has a better win% vs low/med players and definitely not by .85 score.
1000esqrtn1
Posts: 24
Joined: 15 November 2023, 19:20

Re: Reconsider Arena Rankings System

Post by 1000esqrtn1 »

ChiefPointThief wrote: 16 April 2024, 12:08 How are players starting off as low, med, high? Are you using their elo or was that just for purpose of an example?

At 1st glance this seems flawed. It assumes that everyone can get equal matches vs every tier player and that is not the case. I play in some games with small communities in which there are only a few players on my level. Meaning that all the rest of the players have an extreme advantage by being able to play us and they only have to win a minimal # of games to equal the max I can get for playing them. What a strong player is also varies from game to game. How do players break off into med/high players in your system if they start as low/med players?

Player D having a .85 higher score than Player E is the 1st red flag. Sure player D played 10 games against hard opponents but they only won 2 of them. That doesn’t justify them being ahead of player E who has a better win% vs low/med players and definitely not by .85 score.
Thanks for your analysis.

Sure, the tricky part of the system would be on how to categorize players to make a fair outcome. I made a quick example to avoid overcomplicating where it precisely shows what would be the problem to focus (I had another example adding player F as a top player as well). If every level has a certain number of active players it shouldn't matter that you can play n-1 top players while others can play n, because win percentage is what matters. Sure if you are the only player on that level then there is the problem.

The initial idea was to determine the strength based on the ELO they start with. An alternative to avoid exploiting this problem of a few top players could be that everyone starts valued as a low player and you only move to medium or top based on outcomes that occurred on arena, if you promote, you also get some bonus points, making it impossible for someone who didn't promote to surpass you due to winning you while you are on top.
Ceaseless
Posts: 1303
Joined: 12 November 2022, 17:06

Re: Reconsider Arena Rankings System

Post by Ceaseless »

ChiefPointThief wrote: 15 April 2024, 18:24 Another issue is the speed at which the system is reaching these levels due to the increased k factor in arena. If you take two players and one wins both their 1st two games and the other losses they would roughly be around 80pts difference. So just after 2 games into the season player X has to beat player Y at a 62.5% rate to break even. After only 2 games the rate jumps from 50/50 by 12.5%. Now when you consider that this game is also a game with a luck factor which win% are closer to 50% amongst all players that is just a flawed system.
Where are you getting these numbers from by the way?
Shadow00Fox
Posts: 31
Joined: 05 June 2024, 18:15

Re: Reconsider Arena Rankings System

Post by Shadow00Fox »

Found this thread today after getting tired of the imbalanced wins/losses in Can't Stop, and wondering why when I have the filter set to only play my level or higher, I keep getting low levels like bronze, and losing to one of them sets me back 2-3 games worth of elite points.

Sure, might mean more games overall will be played trying to get to high elite status but it extremely decreases the enjoyment factor of bga.
Post Reply

Return to “Discussions”