Grand Chess Tour Tiebreaks: A System Than Which None Lesser Can Be Conceived
Sunday, December 13, 2015 at 11:54PM Having concluded my reporting on the proceedings, it's time to vent some spleen. Before doing so, it's important to note that nothing I will now say is intended to blame Magnus Carlsen or to deny that he was a deserving winner of the London Chess Classic. (I certainly don't think he's the deserving winner of the Tour, but again, that's not his fault.)
I've already noted the unfairness of the playoff procedure which forced Anish Giri and Maxime Vachier-Lagrave to engage each other for around three or four hours (including breaks between games) while Carlsen could rest, nap and/or prepare for his tired challenger. For that matter, I don't understand why it should have been a two-stage event. Using the Sonneborn-Berger tiebreak makes sense in a Swiss system event, where players face different opponents; in a round-robin it seems to me without value. Fine, player A beat player C while player B beat player D, where A and C finished in a tie while C outscored D by half a point. Why not criticize A for his relative incompetence in failing to beat D? And what if A beat C because C was fighting for first place and had to take undue risks? Also, maybe A had White against C while B had Black against both C and D. Why is A's performance more noteworthy? Still further: suppose A is higher-rated than B. Then B had a higher TPR than A; again, why isn't that the first criterion? It has the further benefit of not making A's and B's tiebreakers dependent on how C and D perform against players E through J.
So those are two ways - one more particular, one more general - in which Carlsen was (greatly) benefited and Giri and Vachier-Lagrave were harmed by the tiebreak system in the London Chess Classic. Next, let's recap the way Giri and MVL were punished by the Grand Chess Tour's tiebreak system in the Sinquefield Cup while Carlsen was rewarded. That tournament was won by Levon Aronian, and after that there was a four-way tie for second between Carlsen, Hikaru Nakamura, Vachier-Lagrave and Giri (in tiebreak order). Rather than splitting the points for second through fifth places, the points were allocated as if each player had outscored those below him. As a result Carlsen obtained 10 Tour points, Nakamura 8, Vachier-Lagrave 7 and Giri only 6. As was widely noted at the time, the upshot was that Giri, who was undefeated and +3 in the first two Tour events (the first event was the Norway Chess tournament back in May), was behind Carlsen, whose cumulative score was -1. What a crock.
Finally, Vachier-Lagrave got ripped off in his own special way by the Tour and its absurd policies. The London Chess Classic wasn't just important in its own right or even in its own right and for its implications for this year's Tour; it also had implications for next year's Tour invitees. So, you may ask, who gets to play in next year's Tour? The answer is that the top three finishers from this year's Tour, plus the next six players based on the average of their monthly ratings from February through December of this year, with their live post-tournament rating counting as another "month" to average. (As this year, so too next year will include a tenth wildcard spot for each tournament, decided by the organizers.) They are:
- Magnus Carlsen
- Anish Giri
- Levon Aronian
- Vladimir Kramnik
- Hikaru Nakamura
- Fabiano Caruana
- Viswanathan Anand
- Veselin Topalov
- Wesley So
The first three were Tour qualifiers, the last six made it by rating. Carlsen finished with 26 Tour points, Giri with 23, and Aronian with 22. Vachier-Lagrave finished with 21 points, and before you say "hard luck, he just had to win rather than take second", here's some information for you: he took third. That's right: he beat Giri in the playoff and nevertheless took third in the tournament, behind him. (Incidentally, it wouldn't have mattered to Giri if their places were reversed, because Giri still would have qualified by rating, bumping Wesley So off the list.) So Vachier-Lagrave finished tied or better with Carlsen in all three tournaments (not counting the playoff), but somehow finished fourth and off of the 2016 Tour.
There's enough steer manure here to fertilize a small country. FIDE has been guilty of incompetent and unfair practices over the years, but I don't think they've ever managed to pack so many brain-dead and unjust policies within such a small space in their entire history, and that's really saying something. Well done, Grand Chess Tour. Well done.
Reader Comments (25)
A fine tribute to a game of logic! Thank you for pointing out this farce.
Those tie-break procedures are really weird.
Let's take a look at the game points scored over the 3 tournaments:
Giri 16
Nakamura 15
Vachier-Lagrave 14,5
Carlsen 14
Aronian 14
Topalov 13,5
Anand 13
Grischuk 12,5
Caruana 12
Wildcards 10,5 (Hammer, So, Adams)
Carlsen and Aronian finishing joint 4th in this table, yet because of the tie-breaks they finish 1st and 3rd in the tour.
Giri outscored Carlsen by 2 points in Norway and in both other events they tied. Yet Carlsen finishes above Giri.
Nakamura and MVL are even outside the tour top 3, although they have obtained more game points than Carlsen and Aronian.
And the results of the London tie-break are weird too. MVL beating Giri and still finishing behind him??
plus, the whole semi-final & final thing doesn't really work with 3 players as you mentioned above. Why couldn't they just play a round robin with 3 players (i know that probably disadvantages the player not playing the 3rd game, but at least who that is would be decided by drawing of lots and the disadvantage is much less than it was now for MVL and Giri)
Ridiculous. Nonsense like this is just unprofessional.
A better system: in each tournament, players scoring equally should share both tour points and prize money. For the Tour, overall score should be first criterion (Giri would be the deserving winner!), awarded tour points the first tiebreaker and if those are equal - share the place and prize.
Also, why is it exactly the february through december ratings + one live rating to decide? It makes more sense to use official ratings for the whole year, January-December, doesn't it?
[DM: On the last point, maybe that's when they set up the tour.]
I'd like to add that I think Maxime V-L should get the wildcard invitation for GCT 2016. Just a big shame there isn't room for Grischuk and Karjakin, too.
[DM: It would be interesting if the sponsors all agreed to have the same wildcard for all three events, but the procedure this year (and in similar circumstances) is to have each sponsor pick a player from the host nation.]
I fully agree with your criticism. However, unless I totally misunderstand something, the "Sonneborn-Berger" would make even less sense in a Swiss format, and as far as I know, it has never been used here (instead, the system used is "Buchholz", which is relatively fair and rational).
I think in a round-robin there is just no sensible tie-break rule. Rating-performance is also an irrational factor, for why should the pre-tournament ratings be used to decide the standings? Why should a player be punished for having been more successful before the tournament?
Maybe the only rational tie-breker is "more black games", as the corresponding player had to perform under more difficult conditions.
Apart from that, one should either accept the tie, set up a real tie-break match, or simply role a dice. The last option would be more sincere than the pseudo-scientific "Sonneborn-Berger"-nonsense.
The two main problems are related. As you note, the playoff system is massively unfair to two of the players in the case of a three-person playoff. Also, the tiebreak procedures are only constructed to produce a single winner. This is more obvious when you look at other situations like four-person tiebreaks, where there's no distinction between the bottom two places (you could add a loser's bracket game, of course). In the three-person case, there does happen to be an obvious ordering, which would place MVL second, but if you were going to require tiebreaks to order the entire group you'd have to make sure to set that up for every size of tiebreak.
I agree that any sort of tiebreak points in a round-robin are silly, except for arguably number of blacks and head-to-head score. Thus I think that everyone tied for second in the Sinquefield Cup should have gotten equal points.
Giri was screwed by that, but not (in my opinion) by the scoring system in general. Whenever you break up scoring into multiple games/matches/tournaments, your previous performance is summarized and you get a fresh start. If you go 6-0, 6-0, 4-6, 4-6, 4-6 in a tennis match, you still lose the match, despite having won more games than your opponent. If cumulative score was to matter, they might as well have just made it a 27-game triple-round-robin played in three cities over the course of a year.
Hopefully the system will be better next year.
Agreed on Sonneborn-Berger and everything else. Poor MVL.
On Carlsen I like Jonathan Rowson's thought: https://twitter.com/Jonathan_Rowson/status/676188582856564737
[DM: I saw that comment as well, and think that it's relevant to some complaints (e.g. complaining about his "good luck" vs. Grischuk and an exercise in missing the point with respect to others (e.g. the tiebreaks in the Tour, or for that matter in the London Candidates).]
Exactly what I was thinking after watching the final standing. Carlsen did not quite deserve this victory.
Maybe someone can ask the organizers to make MVL their wildcard, all things considered.
Yeah, MVL got ripped off, no doubt about it. Hopefully the organizers will do the right thing and give him the wildcard spots. "His fantasy, Tattoo...?"
What amazes me is how consistently stupid the things FIDE does are. It's almost like they're trying to perfect idiocy into an art. Maybe they're trying for a leg up in the World Mind(-less) Games...
[DM: FIDE has an impressive record of incompetence and foolishness over the years, but the Grand Chess Tour has nothing to do with them. This was Kasparov's baby, in connection with the Norway, St. Louis and London organizers. All FIDE will do with this tournament is rate it.]
I think the right level of expectations for FIDE is zero. It's what I do with certain relatives. Sometimes they *STILL* manage to disappoint me by going negative, but it's an improvement over hoping for better-next-time...
One point worth noting is that Kramnik was an invitee to this year's Tour as well and declined. It may be that he or one of the other participants (Anand notably has missed major events for "other commitments" in recent years) declines their spot, which would give the organizers an out for re-inviting MVL.
[DM: If you checked out the ratings link, you'll see that it's not just Kramnik who would need to drop out, but also Grischuk, Ding Liren and Sergey Karjakin. It's not gonna happen; at least not that way.]
I suppose we could also just kick out all the people who play the Berlin and invite MVL and others in their places...
I kid.
[DM: That would pretty well empty out the Tour. Carlsen, Giri, Nakamura, Grischuk, Caruana and Topalov all used it in London, and while Anand didn't it's (probably) only because no one played 1.e4 against him! Kramnik uses it, So uses it (occasionally), Karjakin uses it (occasionally)...and on and on it goes.]
Dennis, I agree with your points, but regarding the three-way tiebreak - the alternatives to Carlsen waiting for his opponent are even worse. The problem with three-people round-robin tiebreak is that in case of a 3-way tie in that mini-tournament, there is no way to break it with armaggedon. One would have to fall back to pre-tiebreak coeffiecients like S-B, but then why play the tiebreak in the first place?
While certainly it was a goof up but harsh criticism is uncalled for. It is the first time and the concept was good. If they repeat it again then everyone has the right to criticize. They had rules and regulations before the tournament and it cannot be changed in the middle.Organizers seem to have realized that and hope they will put a better system in place for the next tour.
[DM: This goes way beyond a "goof", and I'd add that when making rules one should try to think through implications beforehand, not afterwards. There have been Grand Prix events in chess and elsewhere before, so it isn't as if the organizers were discovering fire or inventing the wheel. But you're right that everyone makes mistakes, so let's wait and see if they fix the problems.]
One of the good things about the system was that there was a lot of drama right until the last game, with players constantly moving in and out of (realistic) contention for the overall Tour. Regardless of any other minuses, excitement for spectators is surely a very good thing for any sporting event, especially one with a permanent sponsorship problem.
[DM: There still would have been drama even if my implicit proposals had been in place. Moreover, the system used doesn't guarantee drama any more than having had the players split the points from the Sinquefield Cup - it depends on the overall placements. Let's suppose Carlsen had won the Norway Tournament; in that case the 10 points for second in St. Louis would have put the Tour Championship to bed regardless of how he did in the playoff, and probably even without the playoff.]
As others point out, one could have a prize for best overall score, but that is unlikely to create much additional drama, and just turns the whole Tour into a very long round robin. Dfan makes a nice analogy with tennis. Any tennis fan would agree that the scoring system gets a good balance between rewarding the best player while maintaining tension and excitement and realistic room right until the end for a comeback.
One natural solution would be to have a grand final between the overall winners of the three tournaments. If the same person wins all three (unlikely), that person is just declared the winner; otherwise there is, say, a high stakes rapid, blitz and armageddon playoff, and some similar method should be used to settle the winners of individual tournaments. All the tie break methods are arbitrary, and these rapid, blitz etc. playoffs are great fun for spectators, which is surely what it's all about.
Andrey points out that 3 person round robin tiebreaks don't guarantee a winner, but if you go from say rapid to blitz to bullet or whatever, it is a practical certainty that you will get a winner.
Dennis, on another note, thanks for your really great coverage of this event.
[DM: You're welcome.]
Dennis your assessment appears absolutely spot-on, but it seems strange to criticize so harshly without mentioning what your own improved system would be.
[DM: It can be helpful if a critic offers a positive proposal, but it isn't necessary. One can recognize that a system is poor without having a handy alternative. (Does a citizen of a corrupt dictatorship [probably a redundancy] have to propose a sophisticated political theory in order to recognize and justly complain about the dictator's malfeasance?) Indeed, presenting my own alternative might only serve as a distraction or, in the unlikely event that anyone associated with the Tour bothers to read this blog, as an opportunity to create a straw man by criticizing my proposals rather that acknowledging their own mess.
That said, I did offer a couple of suggestions, one explicitly and one implicitly. Explicitly (in the post on the Giri-MVL playoff) I proposed a double round-robin playoff with all three players involved; implicitly, my complaint about the use of the Sonneborn-Berger system applied to the second-place tie in the Sinquefield Cup at least hinted at my view, which is that ties for places lower than first shouldn't be broken, but that Tour points should be allotted equally.
There may be problems with those suggestions, which keep the current system mostly intact. But maybe some clever readers have far better ideas - it wouldn't surprise me a bit. This is something that Tour organizers should consider slowly and carefully.]
@several people: MVL as tenth regular is very unlikely to happen. Norway might do without Hammer (I doubt it), Sinquefield Cup could well do without a fourth American player (IMHO Kamsky, Robson, Sevian, ... would all be completely out of place). But London will insist on an English player (Adams or maybe Howell) - be it only to keep a trace of what the event used to be: an opportunity for English players against the world elite.
[DM: Hammer belongs, but Kamsky, who has been in the world's elite for 20+ years, played in a world championship final and made it to a Candidates' final as recently as 2010, would be "completely out of place"? He hasn't played with a great deal of ambition the last couple of years, and his results have shown it, but as long as the players are given some prep time I would still take Kamsky in a match against Hammer, hands down. Hmm...maybe this should be arranged!]
@Andrey: They could first do a rapid round robin, then a blitz round robin. Rather unlikely that both will end in a three-way tie - and if this is the case, they could either share first place (we tried everything ...) or indeed return to S-B. Of course this should be held the day after the event. Players were still in London anyway, some would have been obliged to skip the pro-Biz cup (interestingly, nobody was willing to pay 3000GBP to be paired up with Carlsen).
Hard to imagine, but it is possible to do worse than London: Gibraltar managed in 2014. Three players were tied for first after the Swiss open. Ivanchuk and Vitiugov had played on the top boards throughout the event, Cheparinov had played "Swiss gambit", came from way behind and logically had a very inferior Buchholz, Then "london-style" tiebreaks, but the initial bye was determined by drawing of lots and favored ... Cheparinov (who won the final to become an IMO paradoxical winner of the event).
Thank you Dennis for your fine analysis of the brain twisting tie-break rules of the Grand Chess Tour, which remains, in spite of those absurd rules one of the most entertaining chess events ever All chess lovers probably reckon that MVL's fighting spirit, courtesy and, above all, results at the board should have allowed him to be invited next year. He is currently number 7 on the Live Chess ratings list and definitely higher if you consider his rapid and blitz performances.
[DM: Rapid & blitz have their own rating lists - rightly, I think. But perhaps there could be a "little tour" elsewhere for rapid & blitz?]
By defeating Giri in the play-off, he deserved the second place that would have qualified him for next year Tour. Hopefully, the organizers will invite him next (at least in Saint Louis where I cannot imagine a fourth US player joining the field). If they don't, we will miss the witty Frenchman.
As to Hammer vs. Kamsky: Personally I think Hammer doesn't "belong" either.
[DM: You might be missing the point of having a wildcard: it's to reward sponsors and drum up local/national interest, not to find a backdoor for those who came heartbreakingly close to qualifying for the permanent nine.]
But a point could be made that he is still young (while 25 isn't that young), has room for improvement and can benefit from supertournament experiences - which he gets only in Norway, but in due course he might be good enough for something like Wijk aan Zee or Biel. Kamsky already played many supertournaments (so did Adams, but he is still competitive) but - to me - seems well past his prime and hinting at retirement. He played the 2010 candidates final based on winning the World Cup in 2007, that's long ago!!?
But the essence of my comment was: Who is more likely to give a wildcard to a foreign player - Norway (one local), Sinquefield Cup (already three locals) or London (no locals)? It might be unlikely also in St. Louis for another reason: The Chess Tour, or Sinquefield as part of the Chess Tour, would thereby admit that they made a mistake. Are they willing/capable of doing so? I am skeptical.
[DM: If that was your main point, you shouldn't have brought in the red herring about Kamsky not belonging. :) Anyway, the answer, in my view, is that none of them should. If the organizers want to admit a mistake re MVL, they should add two spots to the tour and invite him and another permanent member, keeping the wildcard policy intact for the reasons I mentioned in my previous interjection.]
I'm not sure that anything is written about replacement spots being in rating order. For the first tour they mentioned to invite the top ten and when Kramnik declined they didn't invite Nr. 11 (imho Karjakin at the time), but MVL instead.
Btw. I'm not totally sure Kramnik would decline the offer again.
Generally, I like the Grand-Prix-like point system for holding the tension till the last tournament.
[DM: Okay, but as I noted in an earlier comment this system didn't do that. If Carlsen had won in Norway, it would have been more or less over unless he collapsed in London.]
But instead of sharing points and prize money for equal standings, I would propose an alternative tiebreaker. Therefore one can take a closer look on drawn games. We all know about the German term `Remisbreite`.
[DM: We do? :) I don't.]
If there exists a fair and objective evaluation of a drawn game, one could give extra tiebreak points for the stronger side. Here are some propositions for extra tiebreak points:
-stalemating your opponent
-the drawn position is a tablebase win, without the 50 moves rule
-the player who accepts an offered draw gets an extra tiebreak point (which makes only sense without Sofia rules, but makes them obsolete, I guess)
-may be black should generally get an extra tiebreak point for a draw
If there are still equal standings after applying the tiebreaker, money and points were shared.
[DM: That would certainly change things - disastrously, in my view, but YMMV. I'm looking forward to all the 150-move games where players try absurdly to force stalemate in dead drawn endings - endings where they can't offer a draw. (Btw, why should Black get an extra tiebreak point in a position where White has a king on a8 and a pawn on a7, Black a king on c8?) What happens in this Dantean hell when there's a repetition - does the player who claims it lose a tiebreak point? Do they repeat the position 600 times until one player or the other cracks, dies of boredom, or resigns because his bladder is going to burst?
I might be with you on the last point, and sometimes having more black games, or more wins with the black pieces, is already used for tiebreak purposes. That makes sense and doesn't do anything weird with the current rules and conduct of play.]
My point is really about the many drawn games, where one side is kind of outplaying or dominating her opponent, but the game is still within the drawn range - according to rhe rules. This drawn range is relatively broad in chess. In German we have the word 'Remisbreite' to describe this situation. Specially strong players might want to make some further differentiations within that range. I am sure that would generally make sense, but of course there might be also pitfalls and it has to be thoroughly developed. But using it only as a tiebreker might not be that risky.
"I'm looking forward to all the 150-move games where players try absurdly to force stalemate in dead drawn endings - endings where they can't offer a draw."
"What happens in this Dantean hell when there's a repetition - does the player who claims it lose a tiebreak point?"
OK, there is definitely a need for decent arbiters, but I assumed, a player still can ask the arbiter for a draw in a dead drawn position.
"..why should Black get an extra tiebreak point in a position where White has a king on a8 and a pawn on a7, Black a king on c8?"
That is true. Then the stronger side of the draw is without question Black. Hence she gets the extra tb point.
A good idea as a tournament, but maybe too much of a closed circuit. In Norway I expect Hammer to play Tari in a
play off for the wild card spot. Wild card should go to young players, Adams played well but is not a name for the future.
[DM: It's a thought, but in general, the wildcard should go to whoever the people putting up the $$$ want it to go to. No money, no Tour.]
The tour got the points system wrong completely. Also, is chess the only 'sport' with arbitrary time control? Ridiculous. Imagine a football match (real football) being played anything other than 90 minutes).
[DM: I think you're confused about which sport is "real" football, but your point works either way as the (real) American one also has a fixed time limit, albeit of 60 minutes. (It has to be shorter than that used for soccer, as the pain and injury sustained by football (not soccer) injuries can't be cured by magic spray.)]
Maybe tour should include top ten last three months, or include top three under 20 or under 23 to create some interesting pairings between the old and the new?
[DM: Maybe one or two outsiders/young talents, but the idea of having the best of the best seems right to me in principle, if the idea is to draw media attention and have something like an informal tournament world championship.]
My $0.02:
Remove all play-offs. In all ties: split points and prize money. In ties after the 3 tournaments split prize money as well. This is a classical chess tournament series, (not action chess, not blitz) and should be decided playing classical chess. Establish similar action or blitz tournaments (with the same or different participants) with similar rules, i.e. no play-offs, to decide who is better in action or blitz chess
Tasos
[DM: Works for me, though I am sympathetic to the idea of breaking first-place ties for those who value tournament victory. Maybe a hybrid is possible there: a playoff (or even just tiebreaks) for the tournament's first-place trophy, but neither counts for tour points. Would that work for you?]
This gives new meaning to the phrase "A good Player is always lucky". Carlsen was good enough to fight back from a disastrous start in the GP to create all this conundrum. Again off the board the Tie-Break Gods smiled on him and awarded a title that belongs to him by birthright.
I have a nasty feeling the result would have been the same has it been Carlsen fighting his ways to meet a well-rested Giri or MVL.