Beating the bookies with their own numbers - and how the online sports betting market is rigged
Open accessThe DOI matches its DataCite record; 28 of the 28 references it cites were found in a registry or an index; none of the 19 sources its claims cite is checked for support yet.
Abstract
The online sports gambling industry employs teams of data analysts to build forecast models that turn the odds at sports games in their favour. While several betting strategies have been proposed to beat bookmakers, from expert prediction models and arbitrage strategies to odds bias exploitation, their returns have been inconsistent and it remains to be shown that a betting strategy can outperform the online sports betting market. We designed a strategy to beat football bookmakers with their own numbers. Instead of building a forecasting model to compete with bookmakers predictions, we exploited the probability information implicit in the odds publicly available in the marketplace to find bets with mispriced odds. Our strategy proved profitable in a 10-year historical simulation using closing odds, a 6-month historical simulation using minute to minute odds, and a 5-month period during which we staked real money with the bookmakers (we made code, data and models publicly available). Our results demonstrate that the football betting market is inefficient - bookmakers can be consistently beaten across thousands of games in both simulated environments and real-life betting. We provide a detailed description of our betting experience to illustrate how the sports gambling industry compensates these market inefficiencies with discriminatory practices against successful clients.
Overview
CoChat Summary
TL;DR
The authors designed a football betting strategy that uses bookmakers’ aggregate odds as probability estimates and bets when the maximum offered odds exceed fair value. It was profitable in historical simulations, paper trading, and real-money betting, but bookmakers limited their accounts, which the authors interpret as evidence that the market is inefficient and rigged against successful bettors.
Key findings
- In a 10-year historical simulation using closing odds (479,440 games, 818 leagues, 2005–2015), the strategy had 44.4% accuracy and a 3.5% return across 56,435 bets, equivalent to $98,865 profit for $50 stakes. A random-bet strategy averaged −3.32% return and −$93,563 profit; the authors’ return was 10.82 SD above the random mean, with probability <1 in a billion.
- In a 6-month simulation using continuous odds from 5 h to 1 h before games (31,074 games, Sep 2015–Feb 2016), the strategy had 47.6% accuracy and a 9.9% return across 6,994 bets, equivalent to $34,932 profit for $50 stakes. The random strategy averaged 0.2% return; the authors’ return was 4.80 SD above it, with probability <1 in a million.
- Paper trading (06/03/2016–19/04/2016) yielded 44.5% accuracy, 5.5% return, and $1,128.50 profit across 407 bets. Real-money betting (23/04/2016–18/09/2016) yielded 47.2% accuracy, 8.5% return, and $957.50 profit across 265 bets; combined paper and real betting gave 45.5% accuracy, $2,086 profit, and 6.2% return across 672 bets.
- After real-money betting began, bookmakers limited accounts: examples include William Hill limiting a bet to 2,428.33 yen, Interwetten to $11.11, Sportingbet to $1.25, and Betway to $10.45. The authors report Doxxbet closed its website and they could not withdraw €90.
Methods
- The strategy used the consensus probability from the mean odds across bookmakers, adjusted by an intercept α, as a proxy for true outcome probability. Bets were placed when maximum odds exceeded 1/(p_cons − 0.05); α = 0.05 was chosen after testing 0.01–0.1.
- Data included historical closing odds from 32 bookmakers for 479,440 games (2005–2015), continuous odds for 31,074 games (Sep 2015–Feb 2016), and paper/real betting with fixed $50 stakes. Consensus probabilities were binned from 0 to 1 in steps of 0.0125 with at least 100 games per bin.
- Analysis included linear regression of consensus probability against outcomes (R² = 0.999 home, 0.995 draw, 0.998 away), and bootstrap comparisons against a random-bet strategy repeated 2,000 times.
Limitations
- The paper states that closing odds are not a perfect real-life estimate because bookmakers close odds shortly before games; it also notes a time delay where about 30% of dashboard odds had already changed at bookmakers, creating sample bias, though randomly discarding 30% of games still left the strategy profitable.
- The paper states that the real-money sample was small (265 bets) and the combined paper/real result had p = 0.089, above the conventional 0.05 threshold; it also notes no dedicated operator bet 24 h/day, so many opportunities were missed, and bookmaker limits created non-random sampling bias that could negatively affect performance and forced the experiment to stop.
- Evident from the text: the real-money test is short and small, and the strategy depends on bookmakers continuing to offer the selected odds, which the authors’ account restrictions directly undermined.
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
0 Valid · 1 Not valid · 8 Not checked · 82 Own results and interpretations
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“if every bet placed was $50 our strategy would have generated $34,932 in profit across 6,994 bets (Table 1, Figure 2B).”
§ Beating the bookies with their own numbers -and how the online sports betting market is rigged > Strategy Outcome
hedgedstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“In comparison, the return of our strategy was 10.82 standard deviations above the mean of the distribution of returns of the random bet strategy.”
§ Beating the bookies with their own numbers -and how the online sports betting market is rigged > Figures and Tables
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“The probability of obtaining a return greater than or equal to ours with a random strategy that bets on the maximum odds across 31,074 games is less than 1 in a million.”
§ Beating the bookies with their own numbers -and how the online sports betting market is rigged > Figures and Tables
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Our strategy proved profitable in a 10-year historical simulation using closing odds, a 6-month historical simulation using minute to minute odds, and a 5-month period during which we staked real money with the bookmakers .”
§ Beating the bookies with their own numbers -and how the online sports betting market is rigged > Abstract
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Rests on an assumption not shown here.
None of the sources cited for this claim have been checked yet.
In the paper
“Our analysis shows, however, that the implicit information contained in the average odds across bookmakers provides a highly accurate 4”
§ Beating the bookies with their own numbers -and how the online sports betting market is rigged > Strategy Outcome
What the sources say
- Not checked yetcites Forrest, Goddard, and Simmons 2005 matched to “Odds-setters as forecasters: The case of English football” (2005) source
- Not checked yetcites Boulier, Stekler, and Amundson 2006 matched to “Testing the efficiency of the National Football League betting market” (2006) source
- Not checked yetcites Spann and Skiera 2003 matched to “Internet-Based Virtual Stock Markets for Business Forecasting” (2003) source
boostedstates an association- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
40 more claims
Create a free account to see all 91 claims and to verify the 8 references still open.
- It may state two separate claims as one. 36
- The quote only partly states this claim. 4
- Probably not a substantial claim. 11
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
Each contradiction between two passages listed here was confirmed by an independent check.
Rated minor: it would not touch a result — a value no result rests on, a p-value that stays on the same side of the significance level, or a conflict that is clerical, resolved elsewhere in the paper, or outside its own findings.
In the paper
§ Beating the bookies with their own numbers -and how the online sports betting market is rigged > Strategy Outcome
“after three months of paper trading our strategy obtained an accuracy of 44.4% and a return of 5.5%, earning $1,128.50across 407 bets”
§ Beating the bookies with their own numbers -and how the online sports betting market is rigged > Figures and Tables
“Paper trading 06/03/2016 to 19/04/2016 1,128.5 5.5 % 407 44.5 %”
Rated minor: it would not touch a result — a value no result rests on, a p-value that stays on the same side of the significance level, or a conflict that is clerical, resolved elsewhere in the paper, or outside its own findings.
In the paper
§ Beating the bookies with their own numbers -and how the online sports betting market is rigged > Strategy Outcome
“considering games from April 2015 to July 2015 (the period used for paper trading and real betting)”
§ Beating the bookies with their own numbers -and how the online sports betting market is rigged > Figures and Tables
“Paper trading 06/03/2016 to 19/04/2016 1,128.5 5.5 % 407 44.5 % Actual betting 23/04/2016 to 18/09/2016 957.5 8.5 % 265 47.2 %”
Rated minor: it would not touch a result — a value no result rests on, a p-value that stays on the same side of the significance level, or a conflict that is clerical, resolved elsewhere in the paper, or outside its own findings.
In the paper
§ Beating the bookies with their own numbers -and how the online sports betting market is rigged > Strategy Outcome
“after three months of paper trading our strategy obtained an accuracy of 44.4%”
§ Beating the bookies with their own numbers -and how the online sports betting market is rigged > Figures and Tables
“Paper trading 06/03/2016 to 19/04/2016 1,128.5 5.5 % 407 44.5 %”
Citations
Cites
Géron, A. (2017) · Cited by 1,443
Modelling Association Football Scores and Inefficiencies in the Football Betting Market
Dixon, M. J., & Coles, S. G. (1997) · Journal of the Royal Statistical Society Series C: Applied Statistics · doi:10.1111/1467-9876.00065 · Cited by 438
Modelling association football scores
Maher, M. J. (1982) · Statistica Neerlandica · doi:10.1111/j.1467-9574.1982.tb00782.x · Cited by 373
Boredom Proneness in Pathological Gambling
Blaszczynski, A., McConaghy, N., & Frankova, A. (1990) · Psychological Reports · doi:10.2466/pr0.1990.67.1.35 · Cited by 327
Internet-Based Virtual Stock Markets for Business Forecasting
Spann, M., & Skiera, B. (2003) · Management Science · doi:10.1287/mnsc.49.10.1310.17314 · Cited by 282
The five-factor gambling motivation model
Lee, H.-P., Chae, P. K., Lee, H.-S., & Kim, Y.-K. (2007) · Psychiatry Research · doi:10.1016/j.psychres.2006.04.005 · Cited by 220
Odds-setters as forecasters: The case of English football
Forrest, D., Goddard, J., & Simmons, R. (2005) · International Journal of Forecasting · doi:10.1016/j.ijforecast.2005.03.003 · Cited by 212
Spann, M., & Skiera, B. (2009) · Journal of Forecasting · doi:10.1002/for.1091 · Cited by 190
Aggregated knowledge from a small number of debates outperforms the wisdom of large crowds
Navajas, J., Niella, T., Garbulsky, G., Bahrami, B., & Sigman, M. (2018) · Nature Human Behaviour · doi:10.1038/s41562-017-0273-4 · Cited by 183
Information and efficiency: an empirical study of a fixed odds betting market
Kuypers, T. (2000) · Applied Economics · doi:10.1080/00036840050151449 · Cited by 172
Loroz, P. S. (2004) · Psychology and Marketing · doi:10.1002/mar.20008 · Cited by 145
Platz, L., & Millar, M. G. (2001) · Journal of Leisure Research · doi:10.1080/00222216.2001.11949950 · Cited by 110
Vlastakis, N., Dotsis, G., & Markellos, R. N. (2009) · Journal of Forecasting · doi:10.1002/for.1085 · Cited by 97
Sports betting: Can gamblers beat randomness?
Cantinotti, M., Ladouceur, R., & Jacques, C. (2004) · Psychology of Addictive Behaviors · doi:10.1037/0893-164x.18.2.143 · Cited by 85
Constantinou, A. C., Fenton, N. E., & Neil, M. (2013) · Knowledge-Based Systems · doi:10.1016/j.knosys.2013.05.008 · Cited by 69
Profiting from arbitrage and odds biases of the European football gambling market
Constantinou, A. C., & Fenton, N. (2013) · The Journal of Gambling Business and Economics · doi:10.5750/jgbe.v7i2.630 · Cited by 54
EFFICIENCY IN BETTING MARKETS: EVIDENCE FROM ENGLISH FOOTBALL
Deschamps, B., & Gergaud, O. (2012) · The Journal of Prediction Markets · doi:10.5750/jpm.v1i1.420 · Cited by 32
Are Sports Betting Markets Prediction Markets? Evidence From a New Test
Kain, K. J., & Logan, T. D. (2012) · Journal of Sports Economics · doi:10.1177/1527002512437744 · Cited by 31
The home field advantage revisited: a search for the bias in other sports betting markets
Gandar, J. M., Zuber, R. A., & Lamb, R. P. (2001) · Journal of Economics and Business · doi:10.1016/s0148-6195(01)00040-6 · Cited by 27
Testing the efficiency of the National Football League betting market
Boulier, B. L., Stekler, H. O., & Amundson, S. (2006) · Applied Economics · doi:10.1080/00036840500368904 · Cited by 25
Nonlinear modelling of European football scores using support vector machines
Vlastakis, N., Dotsis, G., & Markellos, R. N. (2007) · Applied Economics · doi:10.1080/00036840701731546 · Cited by 15
Beating the bookie: A look at statistical models for prediction of football matches
Langseth, H. (2013) · Frontiers in artificial intelligence and applications · doi:10.3233/978-1-61499-330-8-165 · Cited by 14
Inter-market Arbitrage in Sports BettingPreprint
Franck, E. P., Verbeek, E., Stephan, N., & Nü, S. (2009) · RePEc: Research Papers in Economics · Cited by 10
PROFITING ON INEFFICIENCIES IN BETTING DERIVATIVE MARKETS: THE CASE OF UEFA EURO 2012
Cortis, D., Hales, S., & Bezzina, F. H. (2013) · The Journal of Gambling Business and Economics · doi:10.5750/jgbe.v7i1.597 · Cited by 7
Lock! Risk-Free Arbitrage in the Japanese Racetrack Betting Market
Ashiya, M. (2013) · Journal of Sports Economics · doi:10.1177/1527002513493630 · Cited by 7
Cortis, D. (2016) · Figshare · Cited by 6
Forecasting football match results: Are the many smarter than the few?Preprint
Villar, J. G., Pérez, L., & Rodríguez, P. (2016) · RePEc: Research Papers in Economics · Cited by 1
Aggregated knowledge from a small number of debates outperforms the wisdom of large crowds
Navajas, J., Niella, T., Garbulsky, G., Bahrami, B., & Sigman, M. (2017) · arXiv · doi:10.48550/arxiv.1703.00045
Research works printed but not linked: 1
- Daunhawer, Imant, David Schoch, and Sven Kosub. 2017. "Biases in the Football Betting Market." Available at SSRN: https://ssrn.com/abstract=2977118 .
Cited by
Sports prediction and betting models in the machine learning age: The case of tennis
Wilkens, S. (2021) · Journal of Sports Analytics · doi:10.3233/jsa-200463 · Cited by 86
Structural characteristics of fixed-odds sports betting products
Newall, P. W. S., Russell, A. M. T., & Hing, N. (2021) · Journal of Behavioral Addictions · doi:10.1556/2006.2021.00008 · Cited by 75
The Gamblification of Investing: How a New Generation of Investors Is Being Born to Lose
Newall, P. W. S., & Weiss‐Cohen, L. (2022) · International Journal of Environmental Research and Public Health · doi:10.3390/ijerph19095391 · Cited by 67
Who uses custom sports betting products?
Newall, P. W. S., Cassidy, R., Walasek, L., Ludvig, E. A., & Meyer, C. (2020) · Addiction Research & Theory · doi:10.1080/16066359.2020.1792887 · Cited by 27
A machine learning perspective on responsible gambling
Hassanniakalager, A., & Newall, P. W. S. (2019) · Behavioural Public Policy · doi:10.1017/bpp.2019.9 · Cited by 26
Metadata
- Authors
- Kaunitz, L., Zhong, S., & Kreiner, J.
- Affiliations
- Research Center for Advanced Science and Technology, The University of Tokyo (JP)
- Monash University (AU)
- ORCID
- Zhong, Shenjun
- Venue
- arXiv
- Publisher
- arXiv
- Published
- 2017
- ISSN
- 2331-8422 (ISSN-L)
- arXiv
- 1710.02824
- OpenAlex
- W2763626739
- Cited by
- 33
- References
- 27
- Citations by year
- 2026: 7 · 2025: 5 · 2024: 3 · 2023: 6 · 2022: 2 · 2021: 5 · 2020: 3 · 2019: 1 · 2017: 1
- Type
- posted content
- Language
- en
- Open access
- green
- arxiv.org
- Locations
- arXiv (Cornell University) (submittedVersion)
- arXiv (Cornell University)
- Funders
- Japan Society for the Promotion of Science (P15048)
- Topics
- Social Sciences › Economics, Econometrics and Finance › Economics and Econometrics › Sports Analytics and Performance
- Social Sciences › Psychology › Clinical Psychology › Gambling Behavior and Treatments
- Social Sciences › Decision Sciences › Management Science and Operations Research › Forecasting Techniques and Applications
- Keywords
- online sports betting, bookmakers, arbitrage, football betting, market efficiency, betting strategies
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
A language model drew the claims from the paper on 2026-11-19, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).
A second model labelled each claim on 2026-11-19 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.