Hippos AI scores harness racing days and optimizes the Suomen Hippos race calendar
Artificial intelligence helps Suomen Hippos optimize the race calendar and schedule trotting race events nationwide. By leveraging vast amounts of data, the tool reduces logistical carbon emissions while also supporting the well-being of horses, owners, and the entire harness racing industry.
Suomen Hippos manages and supervises trotting races in Finland, covering everything from amateur to professional levels. The organization manages and utilizes extensive data related to horses and competitions, some of which dates back to the previous century.
Now, this vast amount of data has been harnessed for race planning: HiQ has developed an AI-driven race calendar for Hippos, utilizing information such as horse statistics, stable locations, audience numbers, and betting turnover from races.
“Emotions, perceptions, and traditions play a role in trotting races, but events must also be organized in an increasingly sustainable and cost-effective manner. The AI tool strengthens the fact-based foundation for decision-making and the national placement of race days. This supports not only the well-being of the horses but also that of their owners and the entire industry,” explains Arto Hytönen, Director of Harness Racing at Hippos.
The AI tool strengthens the fact-based foundation for decision-making, supports the well-being of the horses and increases cost efficiency for owners and trainers.
Arto Hytönen
Director of Harness Racing (Raviurheilujohtaja), Hippos
Goal: An automated race calendar to reduce carbon footprint, improve race fill rate, and support horse welfare
Previously, Hippos’ calendar was created manually in collaboration with local racetracks, negotiating event locations, dates, and numbers. While functional, this approach was labor-intensive. The new goal was to automate the planning process using data-driven methods.
“The number of racing horses is declining, so we need to ensure that the number of races matches regional demand. Additionally, we want to shorten travel distances to races, reducing the carbon footprint, improving horse welfare, and increasing cost efficiency for owners and trainers,” Hytönen outlines Hippos’ goals.
The task wasn’t straightforward. The race calendar affects the entire trotting race ecosystem, and planning requires consideration of numerous factors and extensive race data.
“The key was determining whether the data could define the laws governing calendar planning or whether expert judgment was always necessary. To find it out, we analyzed data quality, need for enrichment, and potential statistical phenomena and causality,” says Jonas Pomoell, Lead AI Consultant at HiQ.
Solution: Unlocking a goldmine of trotting race data! Decades worth of data gleaned and analyzed
Hippos’ database is a goldmine of harness racing data: it contains information on 700,000 horses and complete race records dating back to 1984, covering millions of starts. The AI project began with expert interviews and an in-depth evaluation of the data repository.
“Often, more data is available than expected, and it’s worth extracting all of it. However, data is not always uniform, and forcing it into a structured format can require significant work. For instance, analyzing anomalies like the impact of COVID-19 years was essential. Additionally, the data contained previously unknown but statistically significant categories, which had to be identified and processed separately to ensure a high-quality solution”, Pomoell says.
Based on this analysis, a preliminary algorithm was created, and the final AI solution was built using an XGBoost regression model. The multi-phase algorithm constructs the best possible race calendar by scoring race days and favoring local horses that have not competed too recently.
Solution: What makes a ‘good’ race calendar? The right data aligned with industry needs
To ensure the final outcome met Hippos’ needs as closely as possible, the project focused on fine-tuning the algorithm’s parameters.
“A mathematically perfect calendar is not necessarily the best one for people in practice. Our goal was to develop the most functional solution for Hippos, one that also meets the quality criteria of manually designed calendars,” Pomoell explains.
We weren’t given technology for technology’s sake, but rather a tool that genuinely helps in our daily operations and addresses real problems instead of assumed challenges.
Jukka Niskanen
CIO, Hippos
So what makes a harness race ‘good,’ and how is its quality measured? Pomoell states that after a thorough analysis and discussions, the most important factors among hundreds of potential data fields were race attendance, audience numbers, and betting turnover. These are influenced by various factors, including race location, the location of thousands of horses, their age, prize earnings, and rest and recovery periods.
“Defining the right parameters required a partner who could listen to the client, understand the operating environment, and recognize its specific needs. HiQ succeeded in this. We weren’t given technology for technology’s sake, but rather a tool that genuinely helps in our daily operations and addresses real problems instead of assumed challenges,” says Jukka Niskanen, CIO at Hippos.
Results: Besides optimizing race days, AI enhances decision-making and the vitality of the trotting industry
AI enriches Hippos’ decision-making with data and compiles the best possible race calendar within clearly defined parameters. However, the final decision on the race calendar is still made by Hippos’ board and racing committee.
“We gain valuable factual insights from the data to support decisions, such as the total number of race days. The algorithm distributes race days evenly and flexibly across Finland while also considering the preferences and criteria of racetracks, professionals, and hobbyists,” Niskanen explains.
“In addition to reducing manual work, the new calendar allows us to justify decisions based on analyzed data from races and horses instead of relying on opinions and intuition. AI-assisted scheduling makes data-driven decision-making a daily practice.”
The new calendar made data-driven decision-making into a daily practice.
Jukka Niskanen
CIO, Suomen Hippos
The AI-powered calendar has also brought significant improvements to the internal allocation of race days at tracks. The track-specific structure now better reflects the actual distribution and location of horses in Finland.
Additionally, race day optimization improves race attendance rates, which directly impacts betting turnover and maintains interest in trotting races.
“As a result, we can offer larger prizes that support horse owners and increase interest in horse ownership and breeding. Ultimately, the race calendar supports the vitality of the entire industry,” concludes Arto Hytönen.
Results: New location data and ongoing collaboration help regional development of trotting races
A valuable byproduct of the project was new, precise location data on the home stables of racehorses. Hippos previously lacked this information, and according to Niskanen, the new data has already proven useful for the regional development of harness racing.
Hippos and HiQ have collaborated since 2006, and the next steps are already planned: an online platform where users can modify the AI-generated base calendar and simulate various scenarios. The calendar will also receive an objective scoring system to minimize the risk of suboptimal race days. This initiative provides a solid foundation for Hippos’ future data-driven solutions in harness racing development.
Defining the right parameters required a partner who could listen to the client, understand the operating environment, and recognize its specific needs. HiQ succeeded in this.
Jukka Niskanen
CIO, Suomen Hippos ry
Collaboration in a nutshell:
- Jatkuva, avoin kommunikaatio ja keskustelu läpi projektin oikeiden parametrien varmistamiseksi
- Final Product: AI-powered race calendar for optimizing race days and nationwide scheduling
- Analysis of Hippos’ database (containing data on 700,000 horses and millions of races since 1984)
- Data evaluation, enrichment, and identification of statistical phenomena
- AI model based on XGBoost regression with a strong algorithmic framework
- New location data and additional insights for data-driven decision-making
- Fine-tuning of parameters to ensure the most functional race calendar
- Continuous, open communication throughout the project to validate correct parameters
