[Edit: Realized from reading earlier, that this was an idea already suggested by others]
The knock-out format is exciting but often the winner won’t be the strongest player in the tournament, but you gotta go where the sponsor dollars are, I guess.
It’s interesting, but the most popular format for tournaments worldwide (swiss) has been used in the world championship cycle at the interzonal (or world cup) level in only two different cycles. The 1990 Manilla Interzonal was a 13-SS, with 64 participants, with 11 players qualifying for the candidate matches. (2 with 9-4, 2 with 8.5-4.5, 6 with 8-5). The 1993 PCA interzonal was a 11-SS with 54 participants, with 7 players qualifying for the matches (2 with 7.5-4.5, 5 with 7-5). At the same time, the 1993 FIDE Interzonal was a 13-SS with 74 entries, with 10 players qualifying for the matches (1 with 9-4, 8 with 8.5-4.5, 1 with 8-5).
All of the interzonals before 1990 were round robins. Originally there was only one interzonal each cycle, but starting in the 1972-75 cycle there were 2, and starting in the 1981-84 cycle there were three.
I remember somebody at FIDE, maybe Kirsan, in response to the criticisms of the knock-out formula, suggest that the 128 players be divided into 16 groups of 8 and each group would play a round robin. The winners of the 16 groups would then be formed into 2 groups of 8 and play another round robin. The two winners would then play a 4-game match to determine the winner. This would mean 21 playing days (18 classical games + 3 tie-break days) and everyone would play at least 7 rounds. Nothing ever happened of it though.
Great event, but double elimination would make this event much less random. It’s more than a bit weird for 64 players to go across the world to play a mere two classical games, when a fingerfehler could mean elimination. And with double elimination, the final match(es) could be reduced to two games…
Check out the decisive rook endings in Vitiguov-Morozevich and Caruana-Malakhov: entertaining, but is it serious chess? The first game reminds me of the “Vančura hallucination” in one of the rapid Gelfand-Anand games. And Malakhov could have repeated moves before calculating the win. Great players will do silly things if you give them insufficient time to think.
Since there doesn’t seem to be much hope of FIDE adopting the bracket system used by FIFA for the ‘real’ World Cup, it would be an interesting mathematical exercise to try to define a ‘triple elimination’ event.
Suppose that losing a match cost you 2 demerits, drawing a match cost you 1 demerit, and winning cost you nothing. 3 demerits and you are out. This could produce interesting later round matches when, say, two players with lost first round matches (and thus 2 demerits each) face off against each other. A drawn match means BOTH are eliminated, so there would be serious pressure to win the match.
I leave it to others to figure out how the pairings would work.
A 128-player double-elimination tournament would require 255 matches. There would be 13 distinct rounds that would have to be scheduled:
the first round of the event;
11 rounds of loser’s bracket play that could be held concurrently with rounds 2-7 of winner’s bracket play; and
the final round between the winner of the winner’s bracket and the winner of the loser’s bracket.
Assuming the current match format is retained, each match requires 3 days (2 days of classical, and 1 day of tiebreaks if needed). So, this event would run for 39 days - at the least. And in this format, a player could conceivably have to play for 39 straight days. (There are three players in Tromso right now who have gone to tiebreaks every single round.)
A few other potential issues:
You’d have to play through the winners’ bracket (7 rounds) as quickly as possible, because the losers in those matches need to be seeded into the losers’ bracket ASAP. So, the winner of the winners’ bracket would basically have a two-week layoff before the final match started.
If I’m the winner of the winners’ bracket, I would argue that I should have to lose TWO matches, not just one. After all, one of the luxuries of winning a winners’ bracket is that you get two shots to win the tournament, since you haven’t lost yet. (In many double-elim pool tournaments, they’ve done away with this problem by simply making the final match a little longer, and going to one match for all the marbles.)
If I recall correctly, last summer there was something during Wimbledon about a basketball tournament where to be eliminated, a player would have to be dominated a certain number of times. Maybe I have it backwards, and it was written during the NCAA basketball tournament about the way Wimbledon was once run. It was in the Wall Street Journal sports page. To be dominated, you had to either lose to an opponent, or have someone who has beaten you lose to an opponent. This obviously would have an indeterminate number of rounds.
If my math is right, this idea would still require 12 separate playing sessions (down from 13 with 128 players). So it would still be a 36-day event, and whoever won the front side would still be looking at close to two weeks between matches before the final.
Round 1 : Pair highest rating against lowest (1v32,2v31…). 2 game matches, no tiebreakers. Enter demerits as WINS into one’s favorite pairing software or cards. Round 2 : Swiss pairings (although highest v. lowest pairings within groups is nicer for this.) Odd-sized groups, lowest in 2 demerit group plays highest in 1 demerit group f’rinstance. Rounds 3,5,7,etc : Withdraw all players with 3 or 4 “wins”/demerits. Give the 0 and 1 groups 0-demerit byes, as well as the highest-rated player in the 2 group if necessary. Rounds 4,6,8,etc. : Withdraw all players with 3 or 4 “wins”/demerits. All three remaining score groups play.
The first by-hand simulation gave this (using players numbered 1 to 32, 1 is highest rated) : After Round 1 :
2 group : 17 19 21 25 27 28 29 30 32
1 group : 2 7 9 10 11 13 15 18 20 22 23 24 26 31
0 group : 1 3 4 5 6 8 12 14 16 After Round 2 :
4 group : 29 30 32 (eliminated)
3 group : 21 23 25 26 27 28 31 (eliminated)
2 group : 8 9 11 12 13 15 16 17 18 19 20 22 24 (8 gets a 0-pt bye round 3)
1 group : 2 4 7 10 14 (0-pt byes round 3)
0 group : 1 3 5 6 (0-pt byes round 3) After Round 3 :
4 group : 11 16 17 19 20 (eliminated)
3 group : 9 24 (eliminated)
2 group : 8 12 13 15 18 22
1 group : 2 4 7 10 14
0 group : 1 3 5 6 (Odd # players left, 1 gets 0-pt bye in fourth round) After Round 4 :
4 group : 12 13 (eliminated)
3 group : 8 10 14 22 (eliminated)
2 group : 3 6 15 18
1 group : 2 4 7 10 14 (0-pt byes round 3)
0 group : 1 3 5 6 (0-pt byes round 3) After Round 5 :
4 group : 15 18 (eliminated)
2 group : 3 6
1 group : 2 4 7 10 14 (0-pt byes round 3)
0 group : 1 3 5 6 (0-pt byes round 3)
and so on. Looks like as the number of players drops, one should spread out the “bye-rounds” more.
Given the substantial prize for finishing 2nd (a trip to the candidates tournament), there are some substantially easier formats where the “winner” of the losers bracket finishes 2nd, and the final of the losers bracket is the:
loser of the winners bracket versus the winner to that point of the losers bracket.
That may actually save a round if i did my math right.
I think this idea would save one round (the tournament final).
I don’t know that it would reduce the priority of finishing the winners’ bracket quickly. You can’t wait too long to get those matches played.
One potential issue: would this format change make the winner of the winners’ bracket the tournament winner? I would not be happy if I played my way through the back side of that bracket, only to be told the best I could do is second place. There’s also some significant moolah on the line; I’d certainly want my shot at the first-place check.
that’s right. i edited my post right before yours as I did the math more carefully.
yes it would make the winners bracket still the true winner. I realize that it’s a pretty big swing, but the difference between 4th place money in the candidates tournament and not being in the candidates tournament is also a pretty hefty sum, so 2nd place is still a heck of a prize and very much worth the time to someone like Anish Giri who is otherwise a very good player but will not make the candidates tournament.
That’s easy to understand and much less random. And one could reduce the final to two games. I don’t care who wins the World Cup, I care who the two world championship qualifiers are. Cinderella stories (e.g., Hübner 1970) are still possible…
With 64 players in a standard bracket, and double elimination, you could have W=winner, L=loser):
Round 1: 32 W, 32 L, days 1-3
Round 2: from the winners bracket 16 WW, 16 WL and from the losers brack 16 LW and an eliminated 16 LL, days 4-6 (the 16 WL paired against the 16 LW)
Round 3: from the winners bracket 8 WWW, 8 WWL and from the losers bracket 16 (WWL) and an eliminated 16 (WLL), days 7-9 (the 24 WWL paired with each other)
Round 4: from the winners bracket 4 W4 and 4 W3L and from the losers bracket 12 W3L and an eliminated 12 WWLL, days 10-12
Round 4: winners 2 W5, and 2 W4L, losers 8 W4L and dropping 8 W3LL, days 13-15
Round 5: winners 1 W6 and 1 W5L, losers 5 W5L and dropping 5 W4LL, days 16-18
Round 6: winners 1 W6 gets a break for 3 days, losers 3 W6L and dropping 3 W5LL, days 19-21
Round 7: winners 1 W6 vs bottom (on tie-break) 1 W6L while the other 2 W6L play each other, days 22-24
Round 8 (if the W6 won): W7 vs W7L, days 25-27 - possible follow-up match days 28-30
Round 8 (if the W6 lost): W6L sits (reward for winning the winners bracket) and the 2 W*7L play each other, days 25-27 with the last two playing round 9 on days 28-30 and maybe also 31-33
I don’t know how that would make it less “random”. The format introduces the heightened luck factor. The only way to reduce the luck factor is to lengthen the classical portion of the matches, which seems implausible from a scheduling point of view, or to abandon the knockout method entirely.
Also, even after FIDE gets its 20%, there’s still $32,000 on the line in the final. While we might not care who wins the World Cup, I can think of at least two people who will care very much.
In a 64-player, double-elimination bracket, there are 10 rounds in the losers’ bracket alone. (Losers don’t seed back in every round on the back side - after the first two rounds in the losers’ bracket, losers come back in every other round.)
The first round of the event has no losers’ bracket, and the final involves the winner of the losers’ bracket, so that’s 2 more rounds. There will be 12 rounds of play. Assuming the current match format, that’s 36 days - possibly 39, if you play a true double-elimination final where the winner of the front side gets a rematch for all the marbles if he loses to the winner of the back side.
(I ran pool tournaments for a while in and around Champaign, as well as a couple of ACU-I regional 9-ball championships. Spent a lot of time staring at double-elim brackets. LOL)
This is an interesting topic - but I don’t want to derail discussion of the ongoing World Cup event, either. So, I moved the posts discussing format changes for the World Cup to this thread.
In both cases, the higher rated player (Moro and Caruana) survived those matches. I was rooting for Vitiugov too but that stability is what makes a guy like Moro a long-termer in the world elite. Correspondingly, the precious experience he’s gotten as a result of getting those top tournament invitations has surely helped him become the tough customer he is. The young 2700 players have repeatedly mentioned the precious opportunity they have here to play against the recognized elite. Even at 2700, they are generally shut out of the “elite” cycle of tournaments.
Of potentially greater concern I think is the large number of world elite who have been beaten by kids. But looking at those matches and watching the games, they all seemed to be decided on the merits! Maybe other players will disagree with me here. Objectively I guess one could look at the move by move sequence of computer evaluations of the positions. I think this would support my view.
The existing format is looking like genius to me. Stunning results but fully decided by merit. We have a tendency in chess to protect the existing oligarchy a bit too much, and once someone gets into the Candidates cycle, they tend to be able to stay for a while. The Grand Prix depends on getting the right invitations. It’s important that the young generation be given a fair chance to kill the old in at least one place, and that’s the World Cup.
This year, I see the elite being pushed out with extreme prejudice, on live webcast. There’s nothing random about these outcomes.
Dmitry Andreikin has won only one regular-rated game while drawing the rest. As a result, the World Cup semi-finalist has a rating performance below 2700 and is poised to lose rating points.
It is one of the quirks of the rating systems that one can win an event and lose rating points. Andreikin can do that if he continues to draw the regular games and wins in the quick or blitz games.
