Page 1 of 3

Adjustments on quitting players

Posted: 27 April 2015, 20:13
by sourisdudesert
In last January, we changed the way we handle game clock and quitting players on BGA (unique clock, skip turn function, and so on).

A lot of you react about this new system in the forum. Thanks to your help we identified and fixed some bugs, and removed/updated some of the new features.

Now, the new system is simpler than the old one, and we have still an average of 98% of "games finished with everyone on board" on BGA, so we are quite happy with this evolution.

However, we are still concerned about some of your remarks, and decided to do some minor changes to address them.

Here is the list of changes:
  • Like yesterday: there is a unique clock for each player, and a "skip turn" button appears as soon as the clock is negative (no change).
  • Addition: you can't skip the turn of a player who have more time than you, even his time is negative. So if you choose to allow your opponent to play with a huge negative clock, he/she cannot skip your turn when you are 1 second below zero.
  • Update for turn-based mode: when you skip the turn of a player, all his following turns are skipped instantaneously until he eventually join back the game - just like we do in real-time mode. Note: as it was until today, you was able to only skip 1 turn, and then you had to wait that your opponent clock decrease below zero again to skip another turn.
  • Addition: as soon as at least 1 turn has been skipped, we display a banner at the top of the game that say the following: "Player XXX left this game (at Y% of the game progression) and lost this game. All other players will be considered 1st. You can continue to play if you like, or [quit] this game without any penalty".
  • Note: to be 100% clear: this means that you are to continue to play "just for fun", but it counts for nothing. In any way, player that left the game are considered as the last player in results (with associated ELO losses in addition to a "leave" penalty), and the other players will be considered as winners (all tied in 1st place).
  • Note: ... and this is true whatever your choice is: continue the game until the end or quit it without penalty.
  • At the end of the game, the quitter has a "leave" penalty (in addition to the "clock" penalty), plus a loss of ELO corresponding to his last position.
  • At the end of the game, the others players have a ELO gains corresponding to a 1st position (tied), multiplied by the percentage of progression of the game where the quitter left the game. Example: if the quitting player left the game at 95% of the game progression, you will get 95% of the ELO you would have won as a normal winner. This way, this is always more interesting to continue a game until the end than skipping turn of an opponent that is just below zero.
We think that this new system combines the advantages of the old system (possibility to leave a pointless game after a player left + incentive to not using the "expel player" button) and the new system (simple clock + discourage "rage quit" one move before the end). In addition, it is even more simple than the current one (and it works exactly the same with realtime and turnbased).

Re: Adjustments on quitting players

Posted: 27 April 2015, 20:44
by jimbobo
The above is good and I agree with it. However,in game 12412220,nothing happens when I hit the skip player button.
His clock is still spinning and the rest of us are still waiting on a player who has not played in over a month and probably
won't play on BGA again.. Thx.. Jimbobo

Re: Adjustments on quitting players

Posted: 27 April 2015, 20:47
by Oso
Nice adjusts :3. Congratz!

Re: Adjustments on quitting players

Posted: 27 April 2015, 21:10
by stupka
I am not a fan of the % based ELO.

This is completely unfair to a player who has a large advantage early in the game and would likely win 100% of the ELO points.

It would also allow rage quitters to adversely affect their opponents.

Re: Adjustments on quitting players

Posted: 27 April 2015, 21:41
by hiddendragon
Is it possible to lose elo when tie for first?

Re: Adjustments on quitting players

Posted: 27 April 2015, 21:46
by sourisdudesert
hiddendragon wrote:Is it possible to lose elo when tie for first?
In theory, no. Please report the problem if you encounter it.
stupka wrote:This is completely unfair to a player who has a large advantage early in the game and would likely win 100% of the ELO points.
With the 2014 system the winner would have win 0 points. With the previous system he would have win 100% of the points. There is no good system, but this is the best tradeoff we could found.

Re: Adjustments on quitting players

Posted: 27 April 2015, 23:39
by LisaLisLi
Great work! Thank you!

Re: Adjustments on quitting players

Posted: 27 April 2015, 23:48
by stupka
sourisdudesert wrote:
hiddendragon wrote:Is it possible to lose elo when tie for first?
In theory, no. Please report the problem if you encounter it.
stupka wrote:This is completely unfair to a player who has a large advantage early in the game and would likely win 100% of the ELO points.
With the 2014 system the winner would have win 0 points. With the previous system he would have win 100% of the points. There is no good system, but this is the best tradeoff we could found.
Fair enough. Thanks for the response.

Re: Adjustments on quitting players

Posted: 27 April 2015, 23:59
by NomDeRNG
Nice changes there. Have you considered upgrading the overtime marks that players get to reflect how much time has been spent over the clock, or whether the overtime was caused by server lag? I've played games where each click took 30s to a minute to get a response, and wound up 15 seconds overtime, and it seems silly that kind of thing marks profiles the same as when someone takes 10 minutes to decide each move in that game, or someone who goes unresponsive altogether. Maybe non-dropout overtime could be marked as a percentage (i.e. 4 minutes of overtime over ten 30 minute games = 4/(10*30) = 0.013 = 1.3% overtime)?

Re: Adjustments on quitting players

Posted: 28 April 2015, 01:09
by qwertasdfg
Interesting changes. Hope it works well overall.

But certainly doesn't for Hanabi.
see #13719545 as an example