In head-to-head matches, upsets happen when the two archers are of similar level

The summary today

That upsets cluster between archers of similar level is more an expected consequence of the format than a measured result. A match is decided over far fewer arrows than the qualification round, and when the two levels are alike a handful of arrows is enough to turn the order round: in the brackets of the first 2013 World Cup stage the higher-ranked archer usually wins, and the upsets show up mainly around ranks 32 and 56, where the bracket brings together archers with close qualification ranks. A 2016 model, checked against 182 major international competitions, reaches the same conclusion by calculation, that with close levels and moderate scores the stronger archer has only a small advantage, and a 2024 analysis of more than six thousand ends finds that the ratio between the two archers' qualification scores is the variable that weighs most on the last arrow of a set. Nobody, however, has counted upsets with a stated definition, and the 2013 study shows that conditions matter too: at the World Championships, with gusts above 20 m/s, roughly the top twenty qualifiers held on in recurve, while in compound many low-ranked archers finished high up, which means upsets between archers far apart in the rankings. That is why the status is contested: the direct evidence is a reading of the graphs in a single study, and it points two ways.

Also phrased as: "upsets in the brackets"; "anyone can win a match"; "the favourite losing to an equal"; "shock eliminations in match play"; "seeds and upsets"

Evidence

• 2014
  For Archers Published later

  Brackets of the first 2013 World Cup stage in Shanghai: the figures show men's recurve and women's compound, and according to the authors the other divisions show the same features. In most matches the archer with the better qualification rank wins, and the upsets show up mainly around ranks 32 and 56, where archers with close ranks meet. The authors attribute this to the small number of arrows in a match compared with qualification. It is a reading of the rank and placing graphs (figures 4 and 5), with no count of upsets.

  Analysis of scores and arrow grouping at major international archery competitions · James L Park, James E Larven

• 2014
  Against Archers Published later

  At the same 2013 World Championships the matches were shot in gusts measured above 20 m/s. In recurve roughly the top twenty qualifiers still dominated the final placings, while in compound many low-ranked archers reached the high places (figures 20 and 21). The authors link this to the different ways compound archers trigger their release aids. These are upsets between archers far apart in the rankings, not of similar level, in the same study as the previous item.

  Analysis of scores and arrow grouping at major international archery competitions · James L Park, James E Larven

• 2016
  Context Archers Published later

  A model that uses the qualification score as the measure of level and calculates the probability of winning each match, checked against every major international competition from 2000 to 2014, 182 in all. At the 2012 Olympics, in men's recurve, nine archers ranked between 33rd and 64th won their first match against 8.1 expected by the model, and across the 182 competitions the top eight qualifiers took on average 4.2 of the top eight final places against 4.3 expected. The author concludes that when levels are close and scores moderate the stronger archer has only a small advantage, because a match is decided over few arrows. It is a model checked against totals, not a count of upsets.

  Winning major international target archery competitions · James L Park

• 2024
  Context Archers Published later

  Arrow-by-arrow data from recurve matches at London 2012, Rio 2016, Tokyo 2020, the 2021 World Championships in Yankton and the 2021 European Grand Prix in Antalya: 6,374 three-arrow ends. In a quasi-Poisson regression on the score of the last arrow of a set, the ratio between an archer's own qualification score and the opponent's has β 0.41 (CI 0.337-0.483; t 11.0; p<0.001), and the authors describe it as the most important variable. The mean ratio is 1.002 (SD 0.057; range 0.648 to 1.543). The study does not count upsets.

  Gender differences in skilled performance under failure competitive environments: evidence from elite archers · Chunhua Li, Yangqing Zhao

Status history

1. 2014 Contested The first analysis to set qualification against the outcome of matches arrives, using the brackets of the first World Cup stage and the 2013 World Championships. At the World Cup stage the higher-ranked archer usually wins, and the upsets cluster around ranks 32 and 56, where the bracket brings together archers with close qualification ranks; the authors put it down to the small number of arrows in a match. At the World Championships, though, with gusts above 20 m/s, many low-ranked compound archers finished high up, which means upsets between archers far apart in the rankings. The same study therefore points two ways and counts once in favour and once against: contested. It is a reading of the graphs, with no count of upsets.
2. 2016 Contested A model built on qualification scores, in which the probability of winning a match depends on the gap between the two archers' levels, is checked against 182 major international competitions from 2000 to 2014 and holds up: at the 2012 Olympics nine archers ranked between 33rd and 64th got through the first round against 8.1 expected. The author concludes that with close levels and moderate scores the stronger archer has only a small advantage. It points the way of the claim, but it is a model checked against totals and does not count upsets: context, and the status stays at contested.
3. 2024 Contested An analysis of 6,374 three-arrow ends, shot in the matches of three Olympic Games, the 2021 World Championships and a European Grand Prix, finds that the ratio between an archer's qualification score and the opponent's is the variable that weighs most on the last arrow of a set (β 0.41). It also shows that in the brackets the two archers are nearly always close: the mean ratio is 1.002, with a standard deviation of 0.057. It does not, however, count who wins the match, so it says nothing about where upsets fall: it is context, and the status stays at contested.

Related

Related articles

• Five-all in archery: why it happens, and to whomOne match in five between equals ends five-all, and not by chance: it happens to archers who are alike. How often, where in the bracket and why.
• The top qualifier is the best archer once in sevenIn a simulated top-level field the top qualifier is the best archer 14.2 times in 100. Yet at the major events analysed, gold mostly went to a top-16 qualifier.
• I scored more and lost: how the set system worksSince 2010 a match has been decided by sets, not by adding up arrows. How it works, how often the lower total wins, and why it happens mostly between equals.

Research lines

• 7 studiesThe arrow that decides
• 8 studiesChoking and competitive pressure in archers

Other claims in this domain

• CorroboratedAbout 150 points on a FITA round separate a club archer from a national-team archer13 studies · 492 archers measuredStatus since 2003
• CorroboratedAn archer's arousal rises with what is at stake, and keeps rising as the competition goes on3 studies · 23 archers measuredStatus since 2011
• CorroboratedArchers are more aroused in competition than in training5 studies · 50 archers measuredStatus since 2025
• CorroboratedAt 70 metres a top archer averages about nine points per arrow7 studies · 72 archers measuredStatus since 2012
• CorroboratedEven in competition at the highest level, most arrows are not tens4 studies · 3,212 archers measuredStatus since 2024

See all claims on competition and scoring