Match Fixing in Tennis

The BBC and Buzzfeed are jointly reporting on match fixing in tennis. Their story is based partially on leaked documents and partly on data analysis.

BuzzFeed News began its investigation after devising an algorithm to analyse gambling on professional tennis matches over the past seven years. It identified 15 players who regularly lost matches in which heavily lopsided betting appeared to substantially shift the odds – a red flag for possible match-fixing.

Four players showed particularly unusual patterns, losing almost all of these red-flag matches. Given the bookmakers’ initial odds, the chances that the players would perform that badly were less than 1 in 1,000.

More details of the analysis here.

EDITED TO ADD (2/11): This is also a problem in sumo wrestling.

Posted on January 18, 2016 at 10:50 AM21 Comments


Who? January 18, 2016 11:37 AM

Game, set, match, I presume?

A serious example of how much information can be extracted with appropriate data analysis techniques.

Clive Robinson January 18, 2016 11:57 AM

It only goes to show that no matter how much money somebody has some will always “play” for more.

Like drugs, match fixing, is part of “sporting life”, and like all corruption it’s rot spreads quietly but quickly.

You only have to look at what is going on with “Professional Football” (Soccer to those in the US). With Sepp Blatter and Co passing brown envolopes, loans, fees, ticket rights and heaven alone knows what else around to “fix things”.

en there is “World athletes” with Seb (now Lord) Coe apparently not seeing corruption etc etc.

Remember that bloke Lance and his push bike…

Daniel January 18, 2016 12:20 PM


A serious example of how much information can be extracted with appropriate data analysis techniques.</>

The larger issue is that we are under surveillance all the time, whether we know it or not and whether we like it or not. So it is easy to pick examples where this type of big data seems to be “for the good of society” and likewise easy to pick examples where it seems “bad for society”. This leaves us with three possible options. Option one is to get rid of the surveillance entirely, accepting the cost that in our ignorance a certain amount of evil will flourish. The second option is to make surveillance as great as possible, knowing that privacy will eventually vanish. The third option is to try and get the mix correct, as the 4A does in the USA with its test of “reasonable searches and seizures.”

I’d argue that the history of the 4A in the USA shows that societies, or at least American society, does a poor job of getting the mix correctly and that judges have over the years shifted “reasonable” heavily in favor of option #2, leading to what Bruce has called the “golden age of surveillance”. In short, history would suggest that in practice option #3 only pays lip service to privacy but in reality just option #2 in disguise. So this leaves us with a choice between option number one and option number two.

The question is not a choice because I’d argue that as a practical matter we have already chosen number two. The question is whether we are willing to face that act squarely and want are privacy advocates going to do about it. And it is here where Bruce and I part ways. Bruce thinks that we can reform policy and law to get a better mix or even move society over to option number one. I don’t think so. I think the inevitable result is a technological arms race. The reason I think this is because it is impossible to have a rational discussion in a climate of fear. All someone has to do is yell “terrorist!” and decades of balanced policy go out the window. So in the long run what is the use of talking? What is the use of policy when screaming defeats intelligent discussion? I don’t any purpose in what Bruce is doing, to be honest. He’d be far better of going back to working on the math than staying at Harvard.

J January 18, 2016 12:53 PM

They are currently already installing real time heart rate monitors on players etc.
So the option to cheat would be to push heartbeat over limit, until you get clumsy.

albert January 18, 2016 1:14 PM

Frankly, I don’t give a RSA about sports, sports betting, or even cheating athletes. It’s a trivial pastime, meaningless in the Grand Scheme. Don’t even bother about ‘fairness’; nothing is fair. If you don’t like to lose, don’t gamble. You pays your money, you takes your chances.

Why shouldn’t sports mirror the real world? It’s all theatre anyway. Let’s try to limit lead poisoning instead.

Simple statistical analyses don’t impress me either. In tennis, the outcome of a match can be easily determined by one player, whereas in team sports, it’s a little more difficult. I suspect the amount of cheating in pro sports is orders of magnitude more prevalent than is commonly assumed.

Ironically, the the tennis match ‘study’ exactly mirrors what we have been discussing in the mass surveillance posts. You end up with a list of suspects, but no demonstrably guilty parties. And this doesn’t even include the very real possibility of faked data. Prosecution requires evidence, not statistics.

Instead of trying to limit illegal surveillance (which appears to be impossible), why not eliminate the use of electronically-obtained evidence in courts? Just kidding…

. .. . .. — ….

BoppingAround January 18, 2016 3:59 PM

It’s not even about ‘allowing a certain amount of evil to flourish’, for such (or greater) amount will flourish no matter how much surveillance you throw at it. Both the abuse and the plain criminal elements.

Yes, it’s a full-fledged #2. I’m still hesitant to call it as ‘ours’ choice. Ignorance precludes choice, if there was any choice whatsoever in the beginning. More like a wacky clusterfuck of indifference, ignorance and carelessness combined with adversaries that so far have been organised better as well as having more power. Plus anything I haven’t noticed.

Jesse Thompson January 18, 2016 4:07 PM


So in the long run what is the use of talking? What is the use of policy when screaming defeats intelligent discussion?

I think that screaming only defeats intelligent discussion so long as we’re dealing with environments and tools so new to us that virtually no people feel truly confident predicting what they will do.

Compare, scenario 1: you are getting your groceries in a Piggly Wiggly. A hundred other shoppers are milling about half asleep behind their carts, getting their own chores out of the way beneath chirpy PA announcements about fresh fish.

A child screams “Let go! let go!” You look down the brightly lit aisle to determine the commotion, and it’s a 4 year old fighting it’s parents to try to get ahold of some chocolate bars.

Do you freak the fsck out and call 911 about child assault, or does the screaming fail to drown out your common sense?

Scenario 2: You are out camping late at night, several RV’s with other families parked within a few hundred feet from you in all directions. You’ve pitched your tent, doused the fire, and curled up into your roll to sleep for the night. Come 2pm, you are knocked rudely awake as your entire tent is torn away from you and you are flipped end to end in your roll. You hear your tent-stakes and tarp jostle against trees a dozen feet away, and a rhythmic fullumping sound of large padded feet galloping away into the distance grunting gutterally.

The overcast sky leaves the world basically pitch black. You fumble around for your torch but only find your phone instead. You aim the screen into the darkness but that only offers you a few feet of view. Your campsite is torn up pretty badly and you can’t work out a bearing to get back to your jeep.

While trying to get the LED-light of the phone to turn on for greater visibility, you hear a child scream “Let go! let go!”. You crane your eyes into that direction but cannot penetrate the night. You hear a roar, and then a shriek that is suddenly cut off by a thud.

Do you freak the fsck out and call 911 about a wild animal attack in the woods, or do you calmly take the time to regain all of your information gathering tools and make a definitive assessment of what’s going on in favor of presuming the worst?

All technology starts new and that makes crimes related to that technology potentially very frightening: especially when it’s your ass on the line every time something negative happens due to abuse of the system, or when people who have only scifi to draw from cannot feel safe about what new powers the supervillians must have this week.

If you’d like examples, look at the attempt to regulate trains in the late 1800’s and automobiles 1900-1914. People were honestly concerned that trains exceeding mind-numbing speeds like 40mph might startle passing people or livestock completely to death, and pedestrians wandered backwards into streets in front of moving cars as if the cars were merely other people (or, to be precise, as if they were quite pedestrian-accustomed horses instead of distracted human pilots of 1 ton deathtanks).

It’s only after we’ve become more accustom to a technology’s capabilities and limitations do we feel “enlightened” about that part of our lives, and less likely to freak the fsck out every time a child screams.

tyr January 18, 2016 7:13 PM

You might want to look at trains in the 1900-1914
period for a real eye-opener on the conditions of
safety and security the ancients considered normal.

Try Railroad Signatures across the Pacific Northwest
for a fascinating look at things like wooden rails
with a steel cap that would cobra through the floor
of a passenger car if they became detached.

“Railroading is more nearly akin to warfare than any
other humane profession, The number of casualties to
persons on our railways for he year ending June 30,
1907 was 122,855, of which 11,839 represented the
number of persons killed and 111,016 the number
injured … the passengers killed were 610 and those
injured 13.041. The personal injuries, fatal and
non-fatal, assume much the same aspect to the general
officers in command of an army. “

JeFF January 19, 2016 2:37 AM

I just took a (very) quick look at the code, and I don’t get the ‘simulation’ part.

He is testing the odds with a random check ; so at the end he is just calculating the odd again (or testing his randomizer) ? So why this ‘simulation’, apart from hiding other non publishable data source

Avid Tennis Fan January 19, 2016 10:32 AM

Any Schneier fans who have watched a decade or more of tennis will immediately suspect certain players. Totally anecdotal and subjective biases aside, one of these players has gotta be……

Nikolay Davydenko

Has had some amazing flops / drop sets horribly & suddenly in matches where he was completely dominant. Also prone to ‘injuries’ when absolutely pummeling certain opponents e.g. up 2 sets and a break in the 3rd and retires with some dubious injury. Odds suddenly changing and wagering amounts hitting obscene multipliers also present in historical cases.


The ATP launched a match fixing investigation of Davydenko’s match against Martín Vassallo Argüello in Sopot of 2 August 2007, after several large bets were placed at an online British gambling company, Betfair, in Argüello’s favour after Davydenko had won the first set 6–2. Davydenko withdrew from the match during the third set with a foot injury. Although Davydenko had suffered three first-round defeats in his last three tournaments, was injured in an earlier-round match, and showed signs of injury in the second set, it did not make sense to Betfair that such a heavy betting volume would go in Argüello’s direction at that point of time in the match. Per its agreement with the ATP, Betfair notified the Tour. It has since been revealed that nine people based in Russia had bet US$1.5M on Davydenko losing while two unknown people would gain US$6M from the loss. A total of $7M was wagered on the match, ten times the usual amount.

Further controversy also surrounded Davydenko after one of his matches at St. Petersburg Open in October 2007. During his 1–6, 7–5, 6–1 defeat by Marin Čilić, he was given a code violation by umpire Jean-Philippe Dercq for not giving his best effort. He was later fined $2000 (£987) by the ATP, but the fine was rescinded upon appeal. The following week, he lost 6–2, 6–2 to Marcos Baghdatis at the Paris Masters. This generated some controversy, as Davydenko was cautioned by the umpire to do his best during the match.

JB January 19, 2016 11:56 AM

Who are the inveterate gamblers who are betting on single-set performance of semi-elite tennis players?

The thing I find the most shocking is that there are enough marks to take the losing side of these bets.

Unless it’s 100% money laundering where people put up dirty money on the losing side, and clean on the winning side, and nobody is actually honestly betting.

casey January 19, 2016 12:13 PM

I am not convinced that the statistical analysis is evidence of cheating, but am I convinced that the steps taken to protect the anonymity of the players is weak. The sample has 1500 players and based on giving up the year and the odds of the matches picking out the 4 crooked players by name is possible. Sure the data is from public sites, but not naming the players is not particularly ethical when you publish the raw data along with an accusation backed up by math. They should have published outcomes of the algorithm with variables changed. For example, in stead of limiting odds movements to 10% they should have published at 6% 8% 10% 12% etc to prove they are not just using a ‘sweet-spot’ in the numbers to bring the claims into relief. I am always suspicious of these types of messages- watching sports will generate many apparent trends that evaporate when you tweak a few assumptions…

Fascist Nation January 19, 2016 2:24 PM

“…. the chances that the players would perform that badly were less than 1 in 1,000.”

I am a little uncomfortable with this view. Yes, I will accept for argument a <0.1% chance of ANY player. But there is a lot of tennis and a lot of matches and a lot of players over a lot of time. What are the odds over years that a player or handful of players WOULD perform that badly? Probably not high, but a lot better than 1/1,000. Coincidences and shit happen. You can investigate but numbers are not evidence.

Patrick Star January 23, 2016 9:15 AM

Now, I’m no tennis expert or anything, but can’t it just as well be the case that there are hard-to-measure reasons why a player might perform worse against another player than would be expected from their general performance? Such as how various elements of their playing style etc. interact.
Some large betters would, without a doubt, be aware of this and thus take it into account when placing bets.

jomki June 16, 2016 7:12 AM

Obviously tennis is easier to fix since it involves two players unlike a team sport.. its also very hard to detect the fixing..

Leave a comment


Allowed HTML <a href="URL"> • <em> <cite> <i> • <strong> <b> • <sub> <sup> • <ul> <ol> <li> • <blockquote> <pre> Markdown Extra syntax via

Sidebar photo of Bruce Schneier by Joe MacInnis.