Equidia relies on AWS Cloud AI/ML solutions to win the digital race

This is THE race that racegoers wouldn't miss in a million years. The one with the highest stakes for horse racing enthusiasts. Among the hundred or so races that take place every day, the Quinté+, for which punters have to find the leading five horses, is by far the most watched. For Equidia, a French television channel that broadcasts 11 hours of live coverage per day, i.e., a race every 15 minutes, this race is a must-see daily event that should be promoted as much as possible, not only on the channel, but also digitally.

For the past two years, Equidia has been restructuring its digital services, with the creation of an entity dedicated to digital and separate from the broadcast teams in charge of distribution. The channel does not take bets but offers broadcasts of horse races and events with detailed information on the horses or jockeys, and of course, the predictions of its experts. Today, all this content has to go to new platforms. "Punters go to betting shops or racecourses to watch the races and play," explains David Deschamps, CTO of Equidia. "But we need to renew the audience and reach a new audience. A person who discovers Equidia without being familiar with the horse world is quickly lost. There's a lot of information on the screen and it can seem quite complex to the casual viewer. We need to address that person too." Facebook, Instagram, Twitch or even TikTok, Equidia has now set up on multiple channels to capture this new audience and offer more diversified and adapted formats. But the Quinté+ remains at the heart of the strategy!

The channel's flagship programme on Quinté+ is "Ticket Gagnant » (Winning Ticket). This daily 20-minute programme is shot in the afternoon, broadcast in the evening, and then scheduled on a loop until the next morning. Two journalists present all the horses, give their predictions and bring in a live punter. For the digital team, the challenge is to extract the most important sequences and information from the programme and use them on the channel's various digital channels. The programme's web page now allows access to a chaptered version of the programme, but also to see at a glance the experts' opinions on each horse and their tickets thanks to various modules. Equidia uses Amazon Rekognition's image identification solution to achieve this without costly and manual production work.

The machine learning algorithm will automatically identify the speakers, create chapters to go directly to the horse they want, extract the ratings given to each one or even isolate the predictions, to continuously improve and update the programme page. The aim is to offer a personalised experience that allows each audience to obtain the information they are looking for at a glance. "To speed up the operation and reduce costs, the programme images are first sent to an Amazon ECS container before being transferred to Amazon Rekognition. On the container, an application will crop the images so that 50 of them are combined into a single mosaic. Since there are only a few images that are of interest to us out of the twenty minutes of transmission and that the service is charged per image, this allows us to avoid a frame-by-frame detection that would be longer and more expensive. It takes only 5 minutes from the moment the ECS container is launched to the moment we receive the result. And the whole process is fully automated.”
 

The number of visitors to the programme page has doubled since the introduction of this improved content. Equidia is now deploying this approach much more widely on the channel's live feed and analyses 45,000 images per day to detect important sequences. An average of 150 key sequences are extracted each day. Each element thus isolated can then be exploited freely and in an automated manner on different web platforms and social networks. This automation is essential for producing content at a lower cost.

"The Equidia channel is part of an economic system where the results are brought by all the contributions of the Institution (France Galop - Le Trot - PMU - GTHP and Equidia). Equidia contributes with the broadcasting of horse races organised by LeTrot and France Galop, which creates bets for the PMU (betting shops). The stakes generated by the PMU are paid back to the Institution's companies," explains David Deschamps. "Without the bets, we have no direct ROI to show for it. Consequently, new projects must contribute to improving the image of racing, broadening the audience and reducing costs to enable us to continue innovating."
 

All in all, Equidia currently uses more than 30 AWS services in production and has no shortage of ideas for continuing to innovate and be ever more exhaustive in its broadcasting. "The more races we broadcast, the more bets we generate. But we need to find a balance so that the cost of broadcasting is not greater than the winnings from the stakes. Sometimes we broadcast 300 races in one day on 28 channels simultaneously. So we pay for 28 channels and 28 encoders for only a few days of use in a year. We are testing with AWS Elemental Link devices to deploy 5G-connected encoders at racecourses and be able to have pay-per-use pricing." Another major focus is the analysis of horse racing data to generate predictions. Equidia is conducting experiments in this regard in partnership with a school. "We bring our expertise in AWS environments and our knowledge of horse racing, and they provide us with their grey matter to help us find the proper results," says David Deschamps. All bets are off!