Trader's View: Monte Carlo models - The hidden engine behind betting's next revolution
Genius for Sportsbooks
We help sportsbooks scale-up, cut costs and maximise profits with a world-class player experience.
It’s fair to say the rise of in-play betting, micro-betting and same-game parlay (SGP) products has had a mega impact on the trading floor. In this edition of Trader’s View, Thomas Holland , VP - Product at Genius Sports, explains how the Monte Carlo method - a simulation-based theory of sports modelling – is increasing pricing accuracy, market uptime and the capacity for in-play innovation.
The butterfly effect—born from chaos theory—suggests that the flap of a butterfly’s wings could result in a tornado weeks later. It’s a compelling metaphor that shows how even the smallest chain of events can impact complex outcomes.
In the world of sport, a seemingly minor slip, mistimed jump or marginal offload can drastically swing match momentum. Having spent time on the trading floor, I know too well how small moments can alter the probability of an entire game. Today, sportsbooks price-up thousands of granular markets at a time. The growth of in-play betting, and in particular micro and prop betting, means the way of modelling that worked a decade ago is no longer suitable for the speed and scale of today.
Over the last year, the team at Genius Sports has invested in building state-of-the-art models that apply the Monte Carlo theory. Unlike models which simply use past events to predict future ones, Monte Carlo-based models:
If a game is played one million times, for example, how often does the Home Team score the next goal? This approach to trading delivers greater pricing accuracy and increased margins for our partners.
Probability ranges beat fixed values
What was once a totally manual art, soon became reliant on static models.
When in-play betting first emerged, traders priced markets using their specialist knowledge and historical data. Making judgement calls based on the live match, traders set each probability, suspended markets, and accepted and rejected live bets.
As operators added more fixtures and market-types, algorithmic trading models became mainstream. They allowed sportsbooks to price more markets and events every single day, while increasing margins and sportsbook turnover.
The problem? These models are heavily influenced by fixed rules and historical data. They are no longer ideal for today’s betting landscape, in which micro-betting and prop markets are growing in popularity, with the volume of American Football wagers bet on player props as high as 30%, according to Eilers & Krejcik Gaming.
As market-types get richer, more granular data is required - and Monte Carlo models are designed for this environment.
Say Saka scores next for Arsenal, but moments later Rice is dismissed for a second yellow card. Using advanced models, we’re able to not only consider historical and live datasets (as static models currently do), but importantly, also assess how these complex scenarios affect real-time pricing; across singles and betbuilders based on the match context, liabilities and customer betting activity.
Simulations mean more confidence and uptime
One of the big benefits of Monte Carlo modelling? The simulated approach allows us to finely tune specific parts of each event, giving us greater confidence in the outcomes because the models leverage rich datasets and apply them to a large sample size. This reduces the need to suspend markets and reject bets. For the end user, the in-play betting experience becomes faster and more seamless.
Traditionally, markets like “Next Drive Outcome” or “Team to Score in the Next 10 Minutes” have proved tricky to price due to their specificity and volatility. Continuous model simulations in response to the live match state allows traders to price live prop and micro-markets with increased confidence. Our model developments mean that traders need to adjust parameters less, thanks to increased accuracy and automation.
With Monte Carlo trading models, we’re setting the bar even higher. We currently deliver 99.99% uptime on elite soccer leagues, but now with the ability to model isolated parts of the game at the most granular level, we’re unlocking the ability to hit 100% across all sports.
Recommended by LinkedIn
Built for the modern sportsbook
On Super Bowl LIX, Caesars Sportsbook reported the number of SGP (same-game parlay) bets exceeded the number of single wagers. The shift towards SGPs and betbuilder in recent years has raised challenges from a pricing and risk management standpoint.
Monte Carlo models have transformed our betbuilder capabilities, laying the foundation for further innovation. For example, our MultiBet solution leverages this automation for sharp multi-leg pricing at scale, including when market-types are correlated, but at the same time ensuring that pricing is consistent with singles.
Imagine betbuilder pricing driven by millions of simulations, where every possible bet combination is considered, including the relationship between legs, and generating these probabilities becomes as easy as counting the number of times it happened in our simulation.
This also solves a risk management problem. Many sportsbooks lack visibility over the risk exposure that’s spread across multi-leg combinations. For example, a sportsbook’s liabilities might include a series of seven-fold bets with “Mbappe to Score First”, and at the same time, various three and four-fold bets with “Mbappe to Score Anytime”. Traditional risk management software would separate these outcomes, making it impossible to know the true liability.
Last year, we launched Edge, an automated pricing tool that increases margins by leveraging your real-time bet and liability data. Ideal for improving betbuilder risk management, Edge is strengthened further by Monte Carlo modelling.
Our pricing, whether it’s pre-match or in-play, is automatically adjusted based on your liabilities that are calculated using the correlated probabilities and customer betting activity, so you can truly maximise margins at scale.
It’s a totally proactive, forward-thinking approach to trading – and it’s now entirely possible. Forget reactive odds changes. Increase profit margins, increase uptime – and set more accurate odds across every kind of bet in real-time.
Enabling future in-play innovation
The value of Monte Carlo models becomes even greater when you consider the pace of change in live sports data and market-type offerings.
Take “Shot Speed” as a hypothetical outcome – our models are designed to understand new logic, which is derived from historical data, and then use live match state to simulate how often the outcome will occur across a large sample size.
Our Monte Carlo models are built to ingest new forms of data into pricing logic, like player-level data or third-party feeds, and use this data to drive market updates with a richer understanding of the match state.
By investing in our modelling technology, we’re uniquely set up to integrate the next-generation of sports data into our pricing. This is ideal for operators looking to break new ground with their in-play products. Static models aren’t built with this in mind.
Final word
When I say Monte Carlo models; accuracy, scale and innovation are the three words that spring to mind. We’re excited about the change simulated-based modelling is driving across sports like American Football, basketball and soccer.
A single moment can change the course of a game. It can impact thousands of market-types in a matter of seconds. Monte Carlo models simulate millions of iterations, so you don’t leave anything down to chance.
Useful tips
This is a good article that sheds light on the evolution in sports betting pricing. Indeed, the Monte Carlo method seems like a powerful tool for achieving accuracy and scale. I'm interested in adjusting odds based on real-time bets and liabilities. As we know, there's a correlation between the odds values and bettors' behavior, which is often more behavioral than purely mathematical—we've seen all sorts of examples of 'strange' or irrational behavior. My question is: To what extent can Monte Carlo modeling, with its simulations, be fine-tuned to encompass these nuances of human betting behavior? Can these unpredictable behavioral spikes be predicted and modeled to react adequately and minimize the risk for operators, beyond the purely statistical probabilities of the sporting event? #SportsBetting #OddsMaking #RiskManagement #MonteCarloSimulation #TradingInnovation #DataScience
Super insightful look into advances in sports modeling and addressing correlation!
Fantastic read, Thomas! Your explanation of how Monte Carlo models are reshaping the sportsbook industry is both clear and insightful. Monte Carlo simulations are indeed a powerful tool, bridging the gap between prototyping and fast implementation. However, they also introduce new backend challenges for sportsbooks — particularly in terms of computational load. The shift from static to simulated models means embracing both vertical and horizontal scalability to maintain performance. From my experience in telecommunications research, Monte Carlo simulations were invaluable not only for accuracy but also for gaining insights that led to simplified, faster models. In many cases, I transitioned from multi-second simulations to millisecond results by replacing sections of the model with numerical or even closed-form solutions, all while maintaining the necessary confidence levels. Additionally, a well-optimized Finite Events Queue can significantly reduce computational overhead, making Monte Carlo methods even more efficient without sacrificing their benefits. Interesting to see this approach being applied in the sportsbook space.