Peer-reviewed study
Can betting be profitable with models and strategies?
Authors
Open accessPál József Gergő · Bíró Csaba
Published: Annales Mathematicae et Informaticae 61 (2025), 202-214
Evaluating profitability in sports betting using probabilistic models and betting strategies
What the paper examines
It asks a single question: can statistically significant profit be shown in football betting when probabilistic forecasting models are combined with betting strategies? The question is not trivial, because the bookmaker's margin is part of every price, so a model does not need to be good: it needs to be more accurate than the market price.
Data and method
Every pairing of six forecasting models and five betting strategies was run on data from 539 real matches, across 300,000 simulations in total. The models include two heuristics developed for the paper (Veto and Balance), and the strategies cover flat staking, Martingale, Fibonacci, value-based staking and the Kelly criterion.
Result
Among the combinations examined, the Veto model with Kelly staking produced the best outcome. Lasting profit independent of the settings did not appear: most model and strategy pairs stayed in the red once the margin was deducted.
What happened since
The result was later re-measured on a much larger sample, more than 50,000 matches, with walk-forward evaluation. That measurement did not confirm the advantage of Veto with Kelly as durable. This is why LabOfMetrics does not sell tips: the refutation of our own earlier result is the starting point of the platform.
Where to read it
The article is the same in all three places, in a freely available version.
- Publisher version (DOI)https://doi.org/10.33039/ami.2025.10.016
- REAL repositoryhttps://real.mtak.hu/228847/
- University repositoryhttps://publikacio.uni-eszterhazy.hu/8836/
Citation
Pál, J.G. & Bíró, Cs. (2025). Evaluating profitability in sports betting using probabilistic models and betting strategies. Annales Mathematicae et Informaticae, 61, 202-214. https://doi.org/10.33039/ami.2025.10.016
About the author
The paper's first author is the founder and developer of LabOfMetrics. The professional background, the talks and the contact details are on the About page.