Tournament module 2020 UPDATES: Swiss system v2

[wehmsi]

Some questions/remarks/suggestions:

1/
Could we please have the "Open" option of adding people even if the tournament already started?
In order to get replacement when a player quits/leaves for any reason.
In a 5 player game a 4 player-table could be in standby during 5 minutes and expecting a late joiner.

2/
Could we parameter the Swiss System and define the pairing rules?
I mean, the IA tries to avoid first people who already have played together and only after gets people by level (total points).
When number of participants is low, it does not matter that much and reverse option is much more competitive.
So, let us chose as competition manager.

3/
Also as a Manager, let us modify the parameters of a Tournament even after players have registered.
There was a bug in the time allowed to players, everybody was negative right from the start and I could not modify it for the next one.
I had to create it new.

4/
If I chose the "Invitation mode", and send invitations to players to register to a tournament, could you please show the list of players being invited in the main page?
We see only those who accepted.

5/
About the clock problems. The time attribution does not work adequately, everybody complains.
Why is it different to the standard game options?

Thanks in advance for any feedback!

[RobRoyDuncan]

We are doing a 10-player Can't Stop tournament, and the Round 2 pairings are slightly off. We have one table with players ranked 7th, 8th, and 9th; one with players ranked 4th, 5th, and 6th; and one with players ranked 1st, 2nd, 3rd, and 10th. I don't believe Swiss pairings should put the last-place player at the table with the top-ranked players.

Tournament link: https://boardgamearena.com/tournament?id=67126

[RobertBr]

I couldn't see a Carcassonne being trialed so I set one up here https://boardgamearena.com/tournament?id=67382

One point - I assume the pairing is working on standard swiss of matching players as closely as possible by their standing in the tournament. That works great if you count wins and the number of rounds matches the number of players (for 3 players: 2 rounds = 9 players, 3 rounds = 27, 4 rounds = 81) because it culls out a winner in a KO style way while keeping everyone playing. But it looks to me like that does not work well if you award points for second place. Then if the number of rounds is greater than is needed to sort the player a second place in the first round gives a player easier matches for the bulk of the tournament and thus potentially increasing their score artificially. I'm not offering an alternative just suggesting its worth considering as a potential issue.

Otherwise this is a very positive step that has brought me back to playing after a hiatus (caused partly by my strong dislike of the new lobby system)

[Draasill]

We are doing a 10-player Can't Stop tournament, and the Round 2 pairings are slightly off. We have one table with players ranked 7th, 8th, and 9th; one with players ranked 4th, 5th, and 6th; and one with players ranked 1st, 2nd, 3rd, and 10th. I don't believe Swiss pairings should put the last-place player at the table with the top-ranked players.

Hi,

true, this is a (bad) consequence of the fact the matchmaker tries not to make players play against the same opponents twice.

After pairing the 1st players in the first table, it tried adding the fourth player with the best following rank (should be any of the 2nd places).

But for each one it detected them as being an opponent already, and finally stopped on the last player (who had zero point) and added him to the table.

The algorithm could be tweaked to prevent this, although it's slightly complicated; also, Swiss System with 10 players on 3-4p tables will have surprising matching, whatever the algorithm.

[wehmsi]

true, this is a (bad) consequence of the fact the matchmaker tries not to make players play against the same opponents twice.
After pairing the 1st players in the first table, it tried adding the fourth player with the best following rank (should be any of the 2nd places).
But for each one it detected them as being an opponent already, and finally stopped on the last player (who had zero point) and added him to the table.
The algorithm could be tweaked to prevent this, although it's slightly complicated; also, Swiss System with 10 players on 3-4p tables will have surprising matching, whatever the algorithm.

This is exactly what I tried to (poorly) explain in my previous message 2/.
I was a chess referee for many years.
The swiss sytem allows several parameters, and avoiding players to play twice against each is not mandatory.
Depending on number of players you should always be able to adapt the system.
I am asking to let us decide if it is relevant or not.
If # of participants exceeds 40 yes it would be nice to respect that rule, but here we would like players to fight against the best.
The longer the duration of a game is the harder it is to play lots of rounds.
So, if we only can play 4 rounds, and have less than 20 or 30 players, we should be able to prioritize the rank and not the "already played" parameter.
Maybe there should be just "avoid repeat same player 2 games in a row".

[SaraCV]

There is an issue with the way expired time is handled that I think needs to be addressed.

I was just playing in a 4 player tournament game where I ran out of time, received a time penalty and was kicked out of the game by one of the players. Another player also ran out of time but was not kicked out and ended up sharing the victory when the game ended on time.

So the end state was:
Player ...... Points ..... Time ...... Finish
Player B ... 2 points ... + time ..... 1st place
Player C ... 2 points ... + time ..... 1st place
Player D ... 0 points ... -6 hours ... 1st place
SaraCV ..... 1 point .... -3.5 hrs ... 4th place

I regret running out of time, and agree that I should lose because of it. But I think it's important to have consistency in a tournament setting, so all players who run out of time should lose, or be ranked by time remaining, and not have the game results determined by other players deciding who to kick from the game.

Table in question: https://boardgamearena.com/table?table=127841943
Tournament: https://boardgamearena.com/tournament?id=67078

[RobRoyDuncan]

We had something strange come up in a live Can't Stop swiss tournament. We had 10 players register, but one failed to show up for the first round. They were correctly dropped, which gave us 9 players left. Instead of running the rest of the tournament as 3-player games, each round had a 4-player game, a 3-player game, and a 2-player game. This made some games go longer than others; it also did strange things to points, since the 4-player games would pay out 6-4-2-0 for the four players (with some sharing in case of ties), while the 2-player games paid 6-4, so being in those not only gave you a much better chance of victory, failing was better too!

Tournament: https://boardgamearena.com/tournament?id=67462

Also - is this still the correct place to post these, or should they be reported as bugs somewhere? Thanks!

[Draasill]

it also did strange things to points, since the 4-player games would pay out 6-4-2-0 for the four players (with some sharing in case of ties), while the 2-player games paid 6-4, so being in those not only gave you a much better chance of victory, failing was better too!

This is a choice we made, the default configuration trying to never penalize players :
* who are skipped (they earn max. points, because they cannot pretend to the first place as they don't play)
* who are in smaller games. Points attribution is '6/4/2/0' starting always from the first place, because it wouldn't be fair being for players in 2p tables, if points were '2/0' !

Of course, "not penalizing" players means that some may be advantaged.

Customisation of these parameters are being developed, you then will have more control on this.

Also - is this still the correct place to post these, or should they be reported as bugs somewhere? Thanks!

At the moment, this is a good place for discussions as the Swiss System is still in Beta

Thank you for your input !

[Draasill]

The swiss sytem allows several parameters, and avoiding players to play twice against each is not mandatory.
Depending on number of players you should always be able to adapt the system.

Configuration system is being developed, this option may be added in the future.

Maybe there should be just "avoid repeat same player 2 games in a row".

This is a good idea, although it would not prevent the case which triggered this discussion ¯\_(ツ)_/¯

[RobRoyDuncan]

it also did strange things to points, since the 4-player games would pay out 6-4-2-0 for the four players (with some sharing in case of ties), while the 2-player games paid 6-4, so being in those not only gave you a much better chance of victory, failing was better too!

This is a choice we made, the default configuration trying to never penalize players :
<snip>
* who are in smaller games. Points attribution is '6/4/2/0' starting always from the first place, because it wouldn't be fair being for players in 2p tables, if points were '2/0' !

I agree that a win should be a win, regardless of size, but I don't think that all 2nd places should be equal. I'd like to see last always score 0, with the places between evenly spaced. If possible, I'd like to propose making a win worth 12 points, and paying out the following by default:

4p: 12/8/4/0
3p: 12/6/0
2p: 12/0

This makes the rankings between first and last more equitable, in my opinion, and I think using 12 points for first means that ties will always pay out in round numbers.

Thanks for the response, and for your work!