A study of thousands of tennis matches between 2011 and 2013 has found that almost two dozen matches a year have unusual betting patterns that could point to match-fixing.

The Journal of Prediction Markets published an analysis of 6,204 men's and women's tennis matches between 2011 and 2013, conducted by a professor of Sports Law at Florida State University and an Indiana-based tennis gambler. It compared betting patterns to the prices suggested by two statistical models based on previous meetings between the players, finding that 23 matches a year on average have betting patterns that do not match the models.

Three of the matches were at Wimbledon, and one each at the French Open and Australian Open. Overall, 20 differed from one of the models by as much as 16-29 percent before play started. However, the statistics did not account for any other influences on wagering, such as reports of injury or recent form.

The players involved were not identified.

The findings are among several investigations that have found regular indications of some match-fixing in tennis, particularly in lower-profile matches. A report by Belgian betting organization Federbet to the European Parliament this year found unusual betting patterns during a match between Olivier Rochus and Vincent Millot at the Guadeloupe challenger in April.

The issue became a public topic when betting exchange Betfair cancelled wagering on a match involving Nikolay Davydenko in 2007. The ATP, WTA, and ITF subsequently established the Tennis Integrity Unit to monitor potential cases, and its 2008 report suggested 73 matches during the previous five years had been fixed.

Assistant professor Ryan Rodenberg, who co-wrote the study, wanted to see whether things have changed. "Our results are fairly similar, they are in the same ballpark," Rodenberg was quoted as saying by Bloomberg.

Tennis authorities have not commented on the findings published by the journal.