Cillnabradden Evo and Oliver Townend have 10% chance of Badminton win, says new EquiRatings prediction tool

Cillnabradden Evo, Oliver Townend’s first ride, has a 10% chance of winning the Mitsubishi Motors Badminton Horse Trials this week (1-5 May), according to a new predictive tool from equestrian data company EquiRatings.

The Eventing Prediction Centre, an innovation from EquiRatings and their partners SAP, will be launched to the public next week ahead of Chatsworth, but is available to media at Badminton this week.

Using the latest SAP predictive analytics technology, the EquiRatings team has designed a statistical model that calculates the likely outcomes of major events before and during every competition. After each phase, it simulates every combination’s score at that event 10,000 times, exploring every plausible outcome.

According to the model, Cillnabradden Evo has a 10% chance of winning Badminton, a 15% chance of being in the top three and a 22% chance of being in the top 10.

The next two horses on the ranking, Mr Bass (Laura Collett) and Vanir Kamira (Piggy French), have an 8% chance of winning, but both have a higher chance of finishing in the top three or 10 than Cillnabradden Evo.

After each phase, every score is updated and the event simulated 10,000 times, allowing fans to see the movers after each phase and what effect that phase has had on the likely winner, podium or top 10 chances. The tool will also allow the public to input their predicted top three and assess how other fans are voting.

The inputs the Eventing Prediction Centre uses are:

• a horse’s 6RA (average of past six dressage scores, all levels)
• a horse’s average cross-country jumping clear rate at the given competition level
• a rider’s cross-country jumping clear rate at the venue
• a horse’s OBP (opposition beaten percentage)
• a venue’s historical average showjumping clear rate
• a horse’s historical average showjumping faults
• a horse’s most recent showjumping faults

EquiRatings has been testing the model at all four- and five-star (previously three- and four-star) competitions over the past two years.

They report that there is a high level of accuracy when looking at top 10 and top three results. A 20% chance of placing in the top 10 really does occur roughly 20% of the time, and a similar trend holds for top-three rates.

Testing calibration at the top end of the scale (a top three chance of above 30%, or a win probability of above 15%) is difficult.

This is due to the very few cases where a combination is scored this high, and in the case of top threes and wins, relatively few cases of a successful result to measure.

Nevertheless, they report, for the vast majority of combinations the model is pretty accurate in predicting their win and place probabilities.

Article continues below…


You might also be interested in:


The EquiRatings team report there are areas they can improve on in the model and that they are continually making updates with the aim of improving accuracy.

Check back later for a report from the trot-up at Badminton and pick up this week’s magazine (dated 2 May) for the full H&H Badminton form guide.