tom wrote: I thought of having an automatic' heads-up elimination between the tied players... the problem as ever with tourney math is making sure it's always conclusive - e.g. what if there is a 3-way tie, how do you have a 3 way elimination battle on a 2 player board?

Whatever solution is come up with it has to guarantee a tie-breaker result in all possible situations.

I don't think tied tournaments are the answer.

Tesctassa II wrote: About RR tournaments, I think we could use a modified version of the Buchholz which count only the score of the opponents the player won against, or maybe those who lost to.

Or maybe board ranking score could be used instead of the win/loss score. That score is harly the same for three players.

I still think these options will work fine for RR tournaments and Swiss tournaments as well.

About tied tournament I don't see them as a bad solution. First because it won't be that common, ad second because, unless you wan't to use the "let's flip a coin" method, it's impossible to find a rule that is simple and at the same time satisfy everyone or every condition you can have. As for scores, since variables here are limited, they are limited too, so it's improbable to find some kind of ranking that will be an optimal tie-breaker.

(=

Tesctassa's last paragraph says what I was trying to say much better than I did.

But, If we wanted to break ties with run-off games you could "run the tournament again" with whoever tied. (Talking round robin in particular here.) If there was a 3-way tie in a 2 person-game tourney then all three would play each other. Still tied? Do it again! 2-way tie in a 3 person-game tourney - they play each other.

I guess since Hugh started with me as being the prime "bad-player" of the tournament I will chime in.

In my opinion, the current system is not broken, but has issues when we use it in a small RR tournament since the order of things becomes more accented there (and can be gamed). The "smoothing out" of things works with the current since the "play bad early/play good late" should be reasonable evenly distributed with larger tournaments (and isn't a huge deal in my opinion anyway).

There is probably more to argue here on why the current system is okay, but I will stop for the following reason. I think is it reasonable to believe that there is no "fair" tie breaker, the same way there is no "fair" democratic election (Arrow's Impossibility Theorem).

For tie breaker's between two players as in the case Hugh started with it seems fair to have the win go to the winner of their head-to-head game. It has been pointed out that with more than two player ties everything gets complicated (see posts/theorem above), so adding in the check in the specific case of two are tied is fine, but it is a strange one-off rule (only applies in this one situation), and then something else must be used. I can see the argument for ties going to the player with higher public ranking score (reward for playing more publicly), but this also punishing new players that haven't played many games or a hot player in the tournament. Sorry I started to return to an argument there.

One last thought:
If score were calculated after the tournament was over, one round at a time, then there would not be a strategy to this order, and thus just part of the luck of the bracket.

A good suggestion from Mongrel, start the tournament with 10*ranking on that board instead of everyone starts with 10000.

Ok so you suggest it's a one off situation that won't happen that often.

http://www.wargear.net/tournaments/view/80

It's about to happen again if I win the next game.

I agree with the tie breaker going to the head to head winner, but using the public rank score as a basis for tournament score is a bad idea that basically punishes people who don't play a lot of ranked games. Again I would suggest that each round start with the base score and then the result is tallied to the overall score.

Sigh:  Looks like it could happen again; however, I expect "Likes-it" to win.

Alpha wrote: I think is it reasonable to believe that there is no "fair" tie breaker, the same way there is no "fair" democratic election (Arrow's Impossibility Theorem).

Nice theorem! Didn't know that (=

Alpha wrote: I can see the argument for ties going to the player with higher public ranking score (reward for playing more publicly), but this also punishing new players that haven't played many games or a hot player in the tournament. Sorry I started to return to an argument there.

I think board ranking score is a good idea if we all agree that it is a good measure of a player's strength. Of course someone will disagree with that saying that a player who just joined, may be the strongest player but will have a low score. Well it could also happen that a player with a high score starts loosing games after games in the tournament, despite being a strong opponent in public games.

The point is that you can never know how a player will play, and hence every score you can find is always a "suggestion" expressing the probability for that player to play well. For this reason, in my opinion, there is no way to find a criteria which works in every situation. And since a tournament is an "isolated" event, either we find an absolute score to rely on, like board ranking score, or only the tournament's variables can be used to determine a tie-breaker.

Tesctassa II wrote: About RR tournaments, I think we could use a modified version of the Buchholz which count only the score of the opponents the player won against, or maybe those who lost to.
Or maybe board ranking score could be used instead of the win/loss score. That score is harly the same for three players.

I still think these solutions will work fine. The only case this is not true is when the tied players won against the same set of opponents (how many chances for this to happen?) and, if they're even in number, they tied each other (in board Five we saw that it is possible to tie a game, but it's possible on any board if the players decide to terminate the game). If these situation occours, then maybe the best solution is a play-off round played with 1-vs-1 games (even if the tournament had 3 players games).

Then "doot" them because of cheating! If they're still tied after all this, it means they're the same person! XD

Another thing: we're talking about a game where luck matters, and since RR tournament are often played with 1-vs-1 games, it counts even more (except for "Five". Great job Alpha and Mongrel!). So why not considering the seats a player started from (like in chess)?

(=

CiscoKid wrote:

Sigh:  Looks like it could happen again; however, I expect "Likes-it" to win.

True.  He's got a good chance..

Tesctassa II wrote:

I think board ranking score is a good idea if we all agree that it is a good measure of a player's strength. Of course someone will disagree with that saying that a player who just joined, may be the strongest player but will have a low score. Well it could also happen that a player with a high score starts loosing games after games in the tournament, despite being a strong opponent in public games.

The point is that you can never know how a player will play, and hence every score you can find is always a "suggestion" expressing the probability for that player to play well. For this reason, in my opinion, there is no way to find a criteria which works in every situation. And since a tournament is an "isolated" event, either we find an absolute score to rely on, like board ranking score, or only the tournament's variables can be used to determine a tie-breaker.

A tournament is it's own seperate entity.  We shouldn't be using external ranking or scores to determine the winner of something that should start with a clean slate.  I do agree that we should only use the tournament's variables to determine a tie break situation.

Tesctassa II wrote: About RR tournaments, I think we could use a modified version of the Buchholz which count only the score of the opponents the player won against, or maybe those who lost to.
Or maybe board ranking score could be used instead of the win/loss score. That score is harly the same for three players.

I still think these solutions will work fine. The only case this is not true is when the tied players won against the same set of opponents (how many chances for this to happen?) and, if they're even in number, they tied each other (in board Five we saw that it is possible to tie a game, but it's possible on any board if the players decide to terminate the game). If these situation occours, then maybe the best solution is a play-off round played with 1-vs-1 games (even if the tournament had 3 players games).

For Swiss tournaments the existing scoring system works pretty much as intended.  How about the following for Round Robin which is a variation of what I already proposed.

Each round each player starts with the base score of 10000 and at the end of each game you calculate that rounds score using the L/W*200 method.  So every game is basically starting fresh, however the score from each game is tallied into an overall score.  Any ties can be settled using the Bucholz or Median-Bucholz system by adding the tied players opponents scores together and comparing or with Median you throw out the top and bottom scores from what I understand.

Thoughts?

Viper wrote:

Each round each player starts with the base score of 10000 and at the end of each game you calculate that rounds score using the L/W*200 method.  So every game is basically starting fresh, however the score from each game is tallied into an overall score.  Any ties can be settled using the Bucholz or Median-Bucholz system by adding the tied players opponents scores together and comparing or with Median you throw out the top and bottom scores from what I understand.

Thoughts?

Wouldn't that result in the same score for players with the same number of wins?  Just  like previous suggestion of a strict +500 or whatever for a win, ie:

tom wrote: That means the tiebreaker becomes completley ineffectual - a 6-1 player will always end up with the same score as another 6-1 player.

scratch that I didn't think that through fully hah.. That's what I get for trying to do multiple things at once.

Using the buchholz won't work with RR tournaments as Hugh said, since everyone plays against everyone, and hence tied players played against the same set of opponents.

My suggestion was to modifie the buchholz score so are counted only the scores of opponents a player as won against. But again if tied players won against the same set of opponents, this score will be equal.

(=

Ok then while completely wrong on using Bucholz apparently I still think the base score should be applied every round since in a Round Robin the games can and are played in a random order of which you have no control and those lucky enough to play the losing players early have the advantage.

Perhaps the simplest method would be easiest here and make the head to head game results be the tie breaker. A 3 way tie while possible is still fairly unlikely and for the most part this is only going to effect 1 on 1 games.

Exactly. I think that the first criteria should be head-to-head matches, than the modified Buchholz and then play-off round.

Guess it's fair enogth (=

It's way more fair than the current and basically random method that's in place now.

Amidon37 wrote: 
But, If we wanted to break ties with run-off games you could "run the tournament again" with whoever tied. (Talking round robin in particular here.) If there was a 3-way tie in a 2 person-game tourney then all three would play each other. Still tied? Do it again! 2-way tie in a 3 person-game tourney - they play each other.

I like this; I think we should use this for round robin tournaments on 2-player boards when "head-to-head" doesn't determine it.  The only thing people might have against a playoff system is having to play more games than they expected to.  Perhaps it could be a tournament setting option.

I finally had some thoughts about our Swiss tiebreaker versus others.  The way we do it, the winner gets points according to how the opponents did prior to the round when the winner played them.   The Bucholz and others systems (see http://en.wikipedia.org/wiki/Tie-breaking_in_Swiss_system_tournaments) use information about how the opponents did in the rest of the tournament.  The Bucholz system sums the total number of wins of all opponents played in a Swiss.  Both our system and the Bucholz attempt to break ties using information about who played the stronger opponents, but the Bucholz system is using the additional data about how our opponents played in later rounds.

I'm not a fan of this style of tiebreak, but it appears to be a necessity for Swiss because the number of people tied need not be divisible by the number of players per board.  Also, for both our system and the Bucholz system, I can create scenarios where a tie remains.  For that, I offer the iterated Bucholz:  sum the opponents' sum of opponents' scores.

Viper wrote:
A 3 way tie while possible is still fairly unlikely and for the most part this is only going to effect 1 on 1 games.

Not sure about that, and not sure it effects only 1 on 1 either...

Although I'm still in the boat of the system is fine as is by using the score with the exception of ties going to the winner of head-to-head instead of score if applicable.

Yertle wrote:

Although I'm still in the boat of the system is fine as is by using the score with the exception of ties going to the winner of head-to-head instead of score if applicable.

Why?  What are your reasons for this?  Do you believe that the system is selecting who had the harder schedule?  What do you think the system is selecting objectively that gives this system merit?

Posts 19-22 are pretty much why.

In RR everyone plays everyone, so they have the same schedule. Is it better to play noobs first and get their ranking and thus play good players last and have a better chance of losing more? Or losing first and ensuring a late run of wins versus the noobs? I don't know if there is a good way to "predict" which order you play your games in a tournament, therefore I think it's a moot point.

Score fluctuates based upon the score of your opponents, seems like a decent way to determine your "skill".

As has been mentioned before it's easy to work the current system by playing all your RR games at once using the multiple tournament games option and then you play the bad players quickly and take the largest amount of time playing the good players by letting the games sit in queue until just before you turn timer is about to expire. That's not exactly a good or fair representation i think.

Correct. The current system is not fair. If it was obvious to everyone that playing the worst players fastest was advantageous, MAYBE an argument could be made saying that it adds to the strategy. However, it's not obvious at all and took everyone quite a while to figure out. Leaving this system in place would only serve to frustrate players who should have won but didn't because they didn't read this thread and control who they played when.