Arimaa Forum - Print Page


    
      
        Arimaa Forum
        (http://arimaa.com/arimaa/forum/cgi/YaBB.cgi)
      

        Arimaa >> Events >> Strange tiebreak in 2005 Computer Championship
        
(Message started by: Fritzlein on Jan 4^th, 2005, 12:51pm)

Title: Strange tiebreak in 2005 Computer Championship
Post by Fritzlein on Jan 4^th, 2005, 12:51pm

The rules for the computer championship say:

-----
A win will count as 1 point, draw as 0.5 and a loss as 0. The program with the highest score from the playoff games will be selected as the best program. In case of a tie, the program with the fewest number of moves per game averaged for all wins and draws will be declared the winner.
-----

Now draws seem almost never to happen, so it doesn't matter, but it strikes me as odd that the number of moves in a draw should be relevant. Why is it better to have drawn in 20 moves than in 120?

Let's take a case of program A and program B splitting their two games, each winning in exactly the same number of moves. Then each of them beat program C twice in exactly the same number of moves. Then program A wins one and loses one against program D, while program B draws two against program D. Which performance is judged better? By the current rules, if program B had fast draws, it will win, but if program B had slow draws it will lose.

I suggest first tiebreak be number of wins, so that drawing two isn't quite as good as winning one and losing one.

For the second tiebreak I suggest number of moves in all losses minus the number of moves in all wins. Losing slowly seems exactly equal to the accomplishment of winning quickly.

Only for the third tiebreak would I resort to the fewest number of moves in all the wins. (Maybe this was promoted to the first tiebreak accidentally, because in a two-player match the first two tiebreaks will necessarily be equal.)

I wouldn't use the number of moves in a draw in any tiebreak, because I can't see why drawing quickly or slowly is better than the other.

Title: Re: Strange tiebreak in 2005 Computer Championship
Post by omar on Jan 4^th, 2005, 8:26pm

You're right. I should not have included the draws in breaking ties. I've made a note to change that for next year. But Im considering just having one more game to break the tie; with the lower rated player selecting the color.

Title: Re: Strange tiebreak in 2005 Computer Championship
Post by MrBrain on Jan 5^th, 2005, 7:24am

Why not have head-to-head performance as a first or second tie breaker?

Title: Re: Strange tiebreak in 2005 Computer Championship
Post by Fritzlein on Jan 5^th, 2005, 8:33am

I've never understood the use of head-to-head performance as a tiebreaker. It seems to me that the winner should be based on best performance overall. Also you get weird situations whereby there are two teams tied for first and one is winning head-to-head, but if a third team joins the tie, it changes the balance of the head-to-head because you have to include the games all three played against each other, etc.

I must admit that head-to-head results is a popular tiebreak, and if it were used I'd be in the minority that didn't like it.

Anyway, if there is the option of a mini-playoff, that's the best tiebreak of all. It's always better to settle things on the board than via a formula if possible.

Title: Re: Strange tiebreak in 2005 Computer Championship
Post by MrBrain on Jan 14^th, 2005, 9:13am

Settling things on the board is exactly the reason for the head-to-head tiebreak. If A and B are tied in the general pool, but A beat B, then A did beat B on the board.