Kickoff.ai uses machine learning to predict the results of football matches. Based on data about national teams from the past, we model outcomes of football matches in order to predict future confrontations. This page provides a little bit more information about what is happening behind the scenes.

How does Kickoff.ai predict results?

Kickoff.ai uses a statistical model of football matches. With respect to most other models out there, two key differences make Kickoff.ai better.

  1. We model team strength dynamically
    To predict future matches, a model needs to use data from the past. But how "far" in the past does it need to take data? Ideally, teams would have the same squad all the time and play frequently against each other. In practice, however, selected players change regularly and teams play only a few matches every year. To overcome this issue, our model allows the strength of a team to change over time. This enables us to take advantage of the many matches played over almost a century, while taking into account that recent confrontations should be more important to predict upcoming matches.
  2. We use Bayesian inference
    This is a fancy way of saying that we are able to understand how confident we are about a particular prediction. As an example, take Argentina against Iceland. It is likely that Argentina will win—in fact, almost no one (with the possible exception of Icelanders) would claim that Iceland has better chances. But just how much more likely is an Argentina win than an Icelandic win? 60%? 90% Perhaps 99%?1 On the one hand, Argentina is clearly the better team on paper, but on the other hand Iceland has never taken part in a World Cup final tournament, and might perform over its usual level. Bayesian inference takes this (and much more) into account.

1 Our model says 80%...

The football predictions that you see on this site are one particular application of the models and methods that we investigate in our research. In fact, we hope that some of the lessons we learned while building Kickoff.ai will be useful in improving some of the general machine learning techniques that we make use of.

Relevant publications

L. Maystre, V. Kristof, M. Grossglauser, Pairwise Comparisons with Flexible Time-Dynamics, Knowledge Discovery and Data Mining (KDD), 2019.

L. Maystre, V. Kristof, A. J. G. Ferrer, M. Grossglauser, The Player Kernel: Learning Team Strengths Based on Implicit Player Contributions, Machine Learning and Data Mining for Sports Analytics (MLSA), 2016.

L. Maystre, Efficient Learning From Comparisons, PhD thesis, EPFL, 2018.

What is the Kickscore?

The Kickscore is a rough measure of a team's performance over time. It encodes how our model "sees" a team based on the data it has. As it is dynamic, it is possible to interpret how a team's strength evolved over the past decades. We display the Kickscore of each team on their respective pages.

Our team

Kickoff.ai is brought to you by INDY Lab, the research group of Prof. Matthias Grossglauser and Prof. Patrick Thiran in the School of Computer and Communication Sciences at EPFL, Switzerland.


Lucas Maystre

Machine learning PhD student, likes to think of the world in terms of networks.

Visit Lucas' website

Victor Kristof

Machine learning PhD student, curious about the dynamics of our world.

Visit Victor's website

Frequently asked questions

We pride ourselves in providing a probabilistic prediction ("what is the probability that team A will win?"). In probabilistic terms, if team A has 78% chances to win and a match is played 100 times between these two teams, then we expect team A to win 78 times over team B (or to lose or tie 22 times). We believe that this is a much more interesting point of view, and it acknowledges that football is simply not always predictable.
Luckily for the interest of football, matches are not completely predictable. Sometimes, surprises happen (some surprises are less expected than others!). Focusing on a particular prediction misses the big picture, which is the overall accuracy when all matches are taken into account.
We had another model (and another website) for the Euro Cup 2016. You can find more information about the previous project at this address: http://euro2016.kickoff.ai.
All the information on this website is published in good faith and for general information purposes only. Any action you take based on the information on our website is strictly at your own risk. (In short, we are not responsible for any side effects of the luxurious lifestyle you will enjoy if you follow our predictions.)

Press coverage

Contact us

Still have questions? Don't hesitate to ask us - send us an email.


We are really grateful to Jason Reynolds for his amazing job giving a soul to Kickoff.ai with his impressive design skills.

We are also really grateful to Enable for their generous financial support in the development of this project.