Board Game Arena

marvel2010 wrote: ↑22 January 2021, 09:03

Tisaac wrote: ↑18 January 2021, 16:55 This can be changed to mimic what is done in Hanabi : when a game end, we can setup the "opponent" ELO according to what happend during the game.
For instance, in 5p mode, making a perfect 25 is the same as winning against a 2090 ELO player.

What are the options available when computing ELO? Is it possible to assign each player a different fake opponent that they won or tied against? For example, if there are players with ELO 200, 100, 50, can the player with 200 ELO win against a fake opponent with 2090 ELO and the 100 ELO player against a fake opponent with 1800 ELO?

No, that's a coop game so you all won against the same opponent. And we cannot know the current elo of players

I might be a bit late, but I'll suggest the same thing as what I suggested in the 'improvement' request.

Idea: Give all levels some ELO - variable ranking. The levels themselves can get a higher/lower ranking when teams complete them or fail against them. this way every level will reach some kind of equilibrium after enough people have played them. You can win/Lose points for every round that you play based on your ELO and the levels ELO.

Advantages:
- Winning more rounds gives you more points
- Winning harder levels makes you win more points (and vice versa)
- Your Score reflects achievements vs experience
- Possibility to show levels "difficulty" by looking at it's ELO score (nice "extra")

Disadvantages:
- Losing is likelier when playing with inexperienced players and a system like this could lead to experienced players only wanting to play with other experienced players in order not to harm their score.
- ELO-points in cooperative games can lead to ELO-bypass quitting or cheating (e.g. communicating in chat).
- Hard to define initial ELO-score of levels.

Solutions:
- Possibility of working with a "Team Average factor" in case of a loss => If you win, your regular ELO is compared to the Levels ELO. If you lose,
your ELO-point loss is determined by either your ELO or the Team average ELO (whichever ELO is lower) so you wouldn't lose that many points. If you're a high ranked player you can still play with both high-ranked players (with increased success chances) or low-ranked players (with minimized losses) for lower-ranked players it makes no difference.
- Is there any way to bypass the general rule of "when someone left a game for any reason this game is not taken into account by the ELO rating system"? We might want to go higher up and ask Boardgame arena to allow exceptions to this rule for coöperative games. At the other hand side, it's still a boardgame that's supposed to be for fun.
- Not sure if there is historic data but WIN/Lose ratio of each level should be a good indicator for setting initial ELO-score

Other points
- Make sure that the Level's Elasticity is not too high and their ELO points converge.
- Level ranking should NOT go up in case of an early quit (but your points should go down)

matthiasvc wrote: ↑18 March 2021, 12:51 I might be a bit late, but I'll suggest the same thing as what I suggested in the 'improvement' request.

Idea: Give all levels some ELO - variable ranking. The levels themselves can get a higher/lower ranking when teams complete them or fail against them. this way every level will reach some kind of equilibrium after enough people have played them. You can win/Lose points for every round that you play based on your ELO and the levels ELO.

Advantages:
- Winning more rounds gives you more points
- Winning harder levels makes you win more points (and vice versa)
- Your Score reflects achievements vs experience
- Possibility to show levels "difficulty" by looking at it's ELO score (nice "extra")

Disadvantages:
- Losing is likelier when playing with inexperienced players and a system like this could lead to experienced players only wanting to play with other experienced players in order not to harm their score.
- ELO-points in cooperative games can lead to ELO-bypass quitting or cheating (e.g. communicating in chat).
- Hard to define initial ELO-score of levels.

Solutions:
- Possibility of working with a "Team Average factor" in case of a loss => If you win, your regular ELO is compared to the Levels ELO. If you lose,
your ELO-point loss is determined by either your ELO or the Team average ELO (whichever ELO is lower) so you wouldn't lose that many points. If you're a high ranked player you can still play with both high-ranked players (with increased success chances) or low-ranked players (with minimized losses) for lower-ranked players it makes no difference.
- Is there any way to bypass the general rule of "when someone left a game for any reason this game is not taken into account by the ELO rating system"? We might want to go higher up and ask Boardgame arena to allow exceptions to this rule for coöperative games. At the other hand side, it's still a boardgame that's supposed to be for fun.
- Not sure if there is historic data but WIN/Lose ratio of each level should be a good indicator for setting initial ELO-score

Other points
- Make sure that the Level's Elasticity is not too high and their ELO points converge.
- Level ranking should NOT go up in case of an early quit (but your points should go down)

Thanks for your answer. As I explained, I have no access to the player's ELO so what you described as "team average factor" is not possible at the moment.
Also, I like your idea of ELO level but that's also not working with current behavior because even if you 300 ELO and achieve a mission corresponding to 300 ELO, you still win some ELO points. There is no such thing as ELO points convergence sadly.
I don't have any historic data abuot WIN/Lose ratio so feel free to propose initial ELO-scores to see what other players think about them

It's a pity,

I've created a request on BGA Website in General so developpers like you can get more power over this, but I guess that won't get much votes quickly.
We can try however...

https://boardgamearena.com/bug?id=36474

I think something like Habani-based system would be nice, and I agree it should take into consideration the number of levels played.
I just finished the 50 levels starting from #1 with 2 friends, and winning 1 ELO point for that was a bit heartbreaking (I'm not sure I'll ever do that again actually!)
I would have expected to win 50 ELO points for that, for example. Maybe not 50 gazillions. I guess if it can make sense regarding the other ELO rankings on other games (to be at least on a same scale of points) it would be nice. Maybe going from level 1 to level 50 should make you at least an ok player, so give you 100 ELO points? That could mean 2 ELO points per level, or to be divided in another way, 1 point for levels 1 to 15, 2 points for levels 16 to 35 and 3 points for levels 36 to 50 for example (does it add up? I think it might).

It doesn't really give an incentive to go for long runs though. Someone who leaves a game should be awarded 0 ELO point if that's possible. I think the main incentive to continue with the missions one after the other is that when your teammates are playing nice, it just keeps getting nicer. For me you don't really need more incentive than that. When you've succeeded in a mission, you want to try the next one to see if you can keep it up. I think you should make sure there is no incentive in ELO calculations to end a game and start a new one instead of just continuing on with the missions in the same game, and that should be enough on that level.

Some other considerations :

I think generally the difficulty is increasing in the game but not that much, and the players' coordination, ability to communicate with the cards and synergy are by far more important to succeed in the missions. On our long run, we had a lot more trials and fails on earlier missions, and less so at the end, obviously not because the earlier missions were more difficult. And for this game especially, but also generally, when you do a lot of attempts, does it mean you're bad, or does it mean you have more and more experience?

Maybe a special achievement could be created for this game when you do all the campaign from level 1 to 50 with the same players. This is not ELO related but it's linked to recognition of user experience.

Santian69 wrote: ↑19 April 2021, 13:03 I think something like Habani-based system would be nice, and I agree it should take into consideration the number of levels played.
I just finished the 50 levels starting from #1 with 2 friends, and winning 1 ELO point for that was a bit heartbreaking (I'm not sure I'll ever do that again actually!)
I would have expected to win 50 ELO points for that, for example. Maybe not 50 gazillions. I guess if it can make sense regarding the other ELO rankings on other games (to be at least on a same scale of points) it would be nice. Maybe going from level 1 to level 50 should make you at least an ok player, so give you 100 ELO points? That could mean 2 ELO points per level, or to be divided in another way, 1 point for levels 1 to 15, 2 points for levels 16 to 35 and 3 points for levels 36 to 50 for example (does it add up? I think it might).

It doesn't really give an incentive to go for long runs though. Someone who leaves a game should be awarded 0 ELO point if that's possible. I think the main incentive to continue with the missions one after the other is that when your teammates are playing nice, it just keeps getting nicer. For me you don't really need more incentive than that. When you've succeeded in a mission, you want to try the next one to see if you can keep it up. I think you should make sure there is no incentive in ELO calculations to end a game and start a new one instead of just continuing on with the missions in the same game, and that should be enough on that level.

Some other considerations :

I think generally the difficulty is increasing in the game but not that much, and the players' coordination, ability to communicate with the cards and synergy are by far more important to succeed in the missions. On our long run, we had a lot more trials and fails on earlier missions, and less so at the end, obviously not because the earlier missions were more difficult. And for this game especially, but also generally, when you do a lot of attempts, does it mean you're bad, or does it mean you have more and more experience?

Maybe a special achievement could be created for this game when you do all the campaign from level 1 to 50 with the same players. This is not ELO related but it's linked to recognition of user experience.

Have you read my first post ? Because most things you are suggesting are simply not possible with current framework. Not that they are not good idea, but simply not doable at the moment.

I suggest:
(1) Games are always successful (no ELO loss)
(2) ELO of the opponent is the sum for each mission of "2 * N / A" (rounded up)
where N is the number of the mission and A is the number of attempts before success.
(3) Possibly add a multiplier for number of players and other options.

Note:
+This system greatly rewards playing a lot of mission in sequences (rather than individual games). Maybe too much.
+Since there is no significant penalty for failure, we will hopefully not reach the Hanabi problem of "victory or quit".

Hello,

As a big player who loves The Crew, I do not care at all about the current ELO System. WIth one or two rounds, we can easily identify if someone is new or not. And it doesn't even

The issue about having "loss point" would be that noone would want to play with new players, as they would lose points. And for a cooperative games, I do not want that.

When I see a game with 3 people under 5 points, I still want to play with them, explain some tricks, and so on, so they have a good experience of the game, etc. and may want to play again later! With the elo, noone would even try to play with new players, and as explained in other post, if it forces external communication just to get elo, it would ruin all the purpose of the game, so no thanks. We must not forget that new players must be helped so they can enjoy the games, to get a good experience, and maybe play again later

So please, for cooperative games, do not put a way to lose points at all! We need to accept new players as they are so they enjoy the game, and play again later! No need of tryharding everything!

In my opinion, the only good thing that could be done would be :
- if you win 0 to 3 missions => +0 (no more farming from people who just do one mission, then leave...)
- if you complete 4 to 15 missions in one session => +1
- if you complete 16 to 27 missions in one session => +2
- if you complete 29 to 35 missions in one session => +3
- if you complete 37 to 43 missions in one session=> +4
- if you complete 44 to 49 missions in one session => +5
- if you complete all 50 missions in one session => +6

- One extra point for some missions (and if you complete also at least 4 other missions, to avoid farms of these missions specifically) : 17 ,26, 34, 41, 50 (let's say some "checkpoints" for "hard missions" - but we must keep this list "small")

Pros :
- people who wants to make short game, or discovering the game, or other, would still get some points and will not be frustrated
- people would still accept to play with new players
- it doesn't force people to start from the beginning, some people really enjoy playing from 40 to 50 (= +3 points)
- distress & losing are accepted
- we are not getting 500 ELO points in one day...
Cons :
- the only bad thing would be if someone leave before completing a "checkpoint"... some people could get frustrated, but from all the people I play with, noone would care. We're not playing here for points, but for fun!

And just one last thing: for this game, if you want to rework Elo, please listen to people with a current High Elo (like PurpleSunday / LaViajera etc.) and not people with "30 current ELO that just wants it to get higher faster" Right know, the "Apprentice/Medium/Good/Master players" ranking is fine. We still need to give more points for "long runs", and that's all!

Adavayn wrote: ↑17 July 2021, 15:18 Hello,

As a big player who loves The Crew, I do not care at all about the current ELO System. WIth one or two rounds, we can easily identify if someone is new or not. And it doesn't even

The issue about having "loss point" would be that noone would want to play with new players, as they would lose points. And for a cooperative games, I do not want that.

I agree (and argued almost the same is multiple other posts, mostly about Hanabi).
The core of the issue here is that an Elo system is not supposed to be an experience bar. It's not something your should progress regularly, and a perfectly designed Elo system would make you win a lot of points for your first few games, and then you will stagnate forever (well, until you actually get significantly better at the game). The goal of Elo is not for peoples to feel good or bad about their performance, it's just to help the matchmaking algorithm to match peoples of even strength together.

In fact, the whole point of Elo is to ensure that weak players only ever face weak players, while strong players only ever face strong players. So the fact that high Elo players want to reject games against new players is a feature, not a bug. If your system does not do that, it should probably not be called an Elo ranking system.

But it follows that:
(1) Elo loose a lot of its purpose in a cooperative game (where you're not matched AGAINST someone, but matched WITH someone). Especially on a platform like BGA which is supposed to be welcoming to new players.
(2) Elo does assume that peoples play enough to quickly reach their actual level and that they will be content with oscillating around the number of points corresponding to their "true level" instead of constantly progressing (like it would be the case in an XP system). I'm pretty sure that's not the case for most of the BGA players.
(3) Elo does assume the player base is large enough so that peoples can always be matched with peoples of similar level. And this is very wrong for most games on BGA.

=> Elo is not a good ranking system for BGA, except maybe for the few games popular enough to truly have a competitive scene on BGA.

MoiMagnus wrote: ↑17 July 2021, 15:57 In fact, the whole point of Elo is to ensure that weak players only ever face weak players, while strong players only ever face strong players. So the fact that high Elo players want to reject games against new players is a feature, not a bug. If your system does not do that, it should probably not be called an Elo ranking system.

I Agree. This is true for not-cooperative games (so 99% of all games) ; and is completely the opposite for coop' games! When we love a coop' games, the players want to see new players everyday so they discover their fav' games!

MoiMagnus wrote: ↑17 July 2021, 15:57 (1) Elo loose a lot of its purpose in a cooperative game (where you're not matched AGAINST someone, but matched WITH someone). Especially on a platform like BGA which is supposed to be welcoming to new players.
=> Elo is not a good ranking system for BGA, except maybe for the few games popular enough to truly have a competitive scene on BGA.

&

MoiMagnus wrote: ↑17 July 2021, 15:57 => Elo is not a good ranking system for BGA, except maybe for the few games popular enough to truly have a competitive scene on BGA.

Exactly.

An ELO system in coop isn't worth a thing. I would just go for "number of games done" instead, and that's all.