|
|
|
Topic: Tournament group selection
| |
|
|
Author
| Message |
|---|
|
|
Tournament AUTO-ADV-592 has just started. The group arrangement is very unfair as the groups vary dramatically in strength: the average ratings of the players in Groups 1, 2 and 3 are 1827, 1884 and 1739, respectively -- an average difference of 155 points per player between group 2 and group 3.
Arranging the players by rating, we see (where, e.g., `3rd' means `the player with the third highest rating'):
Group 1: 3rd, 6th, 7th, 9th (avg 6.25)
Group 2: 1st, 2nd, 4th, 5th (avg 3.00)
Group 3: 8th, 10th, 11th, 12th (avg 8.25)
Four of the five strongest players are in group 2 and four of the five weakest are in group 3! A much better way of arranging the groups (for any tournament of twelve players) would be:
Group 1: 1st, 4rd, 9th, 12th (avg 6.50)
Group 2: 2nd, 5th, 8th, 11th (avg 6.50)
Group 3: 3rd, 6th, 7th, 10th (avg 6.50)
For the tournament I'm playing, this would result in the three groups having average rating 1833, 1812 and 1805 -- these are much closer together.
By choosing the groups based on the actual ratings of the players, rather than just by their position in the rating list, one could probably get an even closer match but that sounds like it ought to be computationally quite expensive. (Which is to say, I thought about it for five minutes and it seems to be very close to various NP-complete problems and I couldn't see a way of doing it by dynamic programming.)
|
|
|
AUTO-OPEN-797 is even worse! The average ratings of the players in the five groups are 1889, 1770, 1463, 1238 and 1668, respectively. That's a difference of 651 between the highest and lowest!
The highest- and second-highest rated players in the tournament (whom one would expect to be duking it out in the final round) are both in the same group. The third- and fifth-highest are in the same group; ditto the fourth and seventh. The two unrated players are in the same group, in which the other two players are rated 1100p and 1253.
This leads to an extremely lopsided second round, where two of the five players rated over 1990 have almost certainly been eliminated (unless they tie at the top of groups) but at least one player currently rated under 1300 is guaranteed to get through.
Please, Miguel, could you put some effort into arranging tournaments so that the groups are of approximately equal strength? The current system produces some shockingly poor groups. Even just ensuring that the top-rated players are in different groups would be a start.
(I've assumed that unrated players are rated 1300, since that's the rating that is given to two unrated players who draw a game.)
|
|
|
|
|
