Both the Open and Women sections of the Batumi Olympiad were decided by tiebreaks. Like just about every two years, this led to criticism on chess forums. According to grandmasters Peter Heine Nielsen and David Smerdon, the time has come to actually do something about it.
In the open group of the Olympiad, China and USA went into the last round with 17 match points, and were paired against each other for the final round. A dream scenario, you would say—but not without a decisive result.
As we know, the match ended in 2-2 after all games were drawn. Russia beat France to join the two teams in first place, with all finishing on 18 match points. Complicated tiebreak calculations were going to decide on the medals.
It was the same thing in the women’s tournament, where a nail-biting game between Alexandra Kosteniuk and Ju Wenjun was won by the latter. China scored 2-2 there as well, and tied with Ukraine on 18 match points.
Adding to the drama, Kosteniuk-Ju Wenjun included an (incorrect) claim for a threefold repetition. | Maria Emelianova/Chess.com.
This board-one game was the very last of the tournament, so here the tiebreaks could have been calculated already. Nonetheless, when it ended, the Chinese delegation didn’t dare to celebrate too loudly yet.
The most important team tournament in chess, held only once every two years, had finished, but nobody knew who won.
Olympiads, like for instance the European Individual Championship, use the Swiss pairing system, which means each round teams are paired against teams on the same number of match points. As a result, teams who tied for first place after the last round have played against different opponents. Tiebreaks are meant to determine which team performance has a higher value, usually based on how the respective opponents scored.
The main tiebreak system for Olympiads, which is being used since 2008, is “Olympiad-Sonneborn-Berger-Tie-Break without lowest result,” as mentioned below the Chess-Results standings page. What defines the tiebreak score is the sum of all the match points a team scored against each of the opponents, multiplied by the number of board points made against them.
For example, China’s tiebreak score was 372,5 points, based on:
3-1 vs Morocco (12 points) — 3 x 12 = 36
3.5-0.5 vs Colombia (12) — 3.5 x 12 = 42
3-1 vs Peru (13 points) — 3 x 13 = 39
3.5-0.5 vs Croatia (14 points) — 3.5 x 14 = 49
1-3 vs Czech Republic (16 points) — 1 x 16 = 16
2.5-1.5 vs Iran (15 points) — 2.5 x 15 = 37.5
2-2 vs Ukraine (16 points) — 2 x 16 = 32
2.5-1.5 vs Netherlands (13 points) — 2.5 x 13 = 32.5
2.5-1.5 vs Azerbaijan (15 points) — 2.5 x 15 = 37.5
3-1 vs Poland (17 points) — 3 x 17 = 51
2-2 vs USA (18 points) — 2 x 18 = 36
To reach the final number used for standings, the opponent that scored the lowest is removed from the calculations. In China’s case, their opponent Morocco finished 69th, the lowest of all their opponents. Therefore, 36 points got removed from the total of 408.5.
In 2016, USA won gold after having a better tiebreak than Ukraine. In Baku, like in Batumi, the decisive tiebreak numbers depended on the scores of the teams’ respective opponents.
Two years ago, particularly Germany vs Estonia was a critical after the USA and Ukraine had already finished. If Germany had lost, Jordan would have been counted for Ukraine’s tiebreak, and in that case Ukraine would have won gold. Germany won, and made USA the champions.
Danish grandmaster Peter Heine Nielsen provided Chess.com with some more details on what happened back then.
Ukraine beat Germany 2½-1½ and Jordan 4-0. Jordan ended up on 12 match points. Before the last round Germany had 11 match points and played Estonia. They won 2½. Lets see the effects for Tiebreaks:
As Germany won, Ukraine got 2½ x 13 tiebreak points: 32½. Had Germany only drawn, they would have had 12 match points like Jordan, and Ukraine would thus be allowed to use Jordan in their tiebreak, meaning 4 x 12 = 48.
These extra 15½ tiebreak points would have been enough for gold. Note the dynamic: Ukraine loses gold because a team they played wins(!) a game. By any kind of logic you should never be worse of because a team you played improves their score. It goes against any kind of ideas of fairness in a tiebreak system.
A similar scenario occurred in Batumi. Danish grandmaster Peter Heine Nielsen pointed out on Twitter what was going on.
Also for Batumi, Nielsen gave us the specifics:
USA had beaten Panama 4-0 and Georgia 3 2½-1½. Before the last round they both had 11 match points. USA was leading the tiebreaks ahead of China as Panama was calculated in their tiebreaks meaning 4 x 11 = 44. However, crucial was that Georgia 3 would not overtake Panama. The pairings was Georgia 3-Singapore and Panama-Bangladesh.
As it happened, Georgia 3 won and USA got 2½ x 13 = 32½ match points. Had Georgia 3 lost, they would have gotten 4 x 11 = 44 match points, which was exactly not enough for gold. However, had both matches ended 2-2, it would have been enough for gold!
According to Nielsen, an important issue with the tiebreak system that’s being used is that it can work against its own purpose: “In both examples it stands out how a team you played, improving their score, lowers you tiebreak.”
This indeed doesn’t sound logical at all. “It would be much more fair and much less random,” says Nielsen, “to at least count all 11 teams for tiebreaks. Still last games could decide, but with much smaller margins than suddenly removing a 4-0 instead of a 2½-1½.”
The Danish GM, who is also the second of World Champion Magnus Carlsen, thinks there is another flaw in the current system: a 4-0 victory over a weaker team can be more valuable than beating a strong team 2½-1½. This is relevant for yet another Olympiad: the one in 2012 in Istanbul.
“When Armenia won ahead of Russia in Istanbul, they won the tiebreak based on a 4-0 against Thailand which gave them a much bigger tiebreak number than Russia beating Ukraine 2½-½,” said Nielsen. “In my opinion this goes against our feeling of justice. So you could say that my criticism of this part of the system is specifically aimed at the current strength-level of the teams participating.”
Peter Heine Nielsen. | Photo: Maria Emelianova/Chess.com.
Nielsen can be considered an expert from experience, as he was involved in a similar situation in 2011. At the European Championship, also an 11-round Swiss, he just missed out on qualifying for the FIDE World Cup with the second tiebreak being performance rating. The European Chess Union eventually agreed with him, but Nielsen’s protest came too late to make changes in the final standings. The affected players, including Nielsen, did get a wild card.
“The similarity to my 2011 case is that also there they removed players from tiebreak calculations based on one criteria, without realizing that they used two criteria for awarding tiebreak points,” said Nielsen. “This way it becomes extremely important how you performed exactly against the players removed, instead of what we aim for: How you performed generally in the whole tournament.”
Australian grandmaster and economist David Smerdon agrees with Nielsen. “All of the main tiebreak systems are both mathematically sound and mathematically flawed, in the sense that you can always find an exception where each system isn’t ‘fair’. The trick is working out which system more often works best in which sorts of tournaments.”
But, as Smerdon notes, such considerations can get lost behind a high-profile example of where things went wrong. “I get the feeling that past policy changes have been largely reactionary to isolated events. One system gets replaced by another just because it fixes exactly that situation and calms the noise. The most difficult problem is that it’s much easier to say ’This system is better than that system for this one tournament’ than to think about all possible scenarios in a comprehensive way.”
David Smerdon. | Photo: Peter Doggers/Chess.com.
Both Nielsen and Smerdon recommend the new FIDE administration to start working towards a better tiebreak system. This process should start with comparison of the main existing (and new) tiebreak methods, “especially by mathematical robustness,” says Smerdon. For this, 2017 a study by Roberto Ricca (here in PDF) can serve as a starting point.
Based on computer simulations of different types of Swiss tournaments, Ricca provided a list of five recommended tiebreak criteria. The study, which has been discussed in the October 2017 meeting of the FIDE Technical Commission in Antalya, should definitely be taken into account in the process of establishing an improved tiebreak system for, say, the 2020 Olympiad. It is not the definite answer yet, says Smerdon:
“I hope it could be expanded to discuss fairness, which is a big topic in economics these days, and to analyse which systems do best for different types of tournaments. For example, systems that do best in the Olympiad, with many rounds and many teams, might not be as suitable for shorter events with fewer teams. What we need are transparent, evidence-based guidelines that are convincing and acceptable to arbiters, organisers and players alike.”
The normal procedure would be that the Technical Commission sends recommendations to the World Championships & Olympiad Commission. It is this commission that has the mandate to change regulations for the next Olympiad. However, it is also this specific commission that hasn’t been very effective in recent years, having met infrequently and having made few concrete decisions.
The question is whether the new FIDE administration can change this. What is clear is that current flaws should be changed, and specifically the adaptation of removing the lowest result.
This specific part of the current tiebreak is intended to do something against the randomness of first-round pairings, in which strong teams are paired against (very) weak teams and the result of these early matches can have long-lasting effects in the standings.
Nielsen: “This is a reasonable concern, but obviously not dealt with properly in the current system. It unintentionally makes it much worse.”