Epsom Derby Supercomputer Prediction Epsom Downs will be full of the sound of thundering hooves and excited spectators as some of the best horses will contend for the Epsom Derby this Saturday. City Of Troy rode to victory last year and we at irishracing.com consulted our supercomputer to see which horse stands the best chance of winning this year. The Algorithm Powering The Supercomputer We collected comprehensive data on the past winners and contenders of the Epsom Derby since 1995. The next step was to determine the importance of the horses’ characteristics, in terms of what is advantageous for the race in question. This was done by consulting domain experts, and the characteristics were further classified as ‘numerical’ and ‘categorical’ features. These features are: Horse form and course form Country of origin (for the horse) Betting position Trainer and trainer country Sire and Sire country of origin Dam and Dam country of origin In consultation with the domain experts, the features were assigned different weightages. In the case of missing values for any of these, we used the median from historical data to represent the values. Categorical features were encoded using LabelEncoder for modelling purposes. Numerical features were scaled using StandardScaler to ensure comparable scales Determining The ‘Typical Winner’ Using the algorithm mentioned above, we extracted the characteristics of all winning horses at the Epsom Derby, to determine the characteristics of a ‘typical winning horse’ at the race. Once they were extracted, we used the median values as a representation of the characteristics of a typical winner. From our analysis, these are the characteristics of a winning horse at the event: Betting Position1 Initial Odds7.25 Country of origin (Trainer)Great BritainCountry of origin (Horse)IrelandCountry of origin (Sire)IrelandCountry of origin (Dam)Ireland Tabulating Similarity Scores With comprehensive numerical information on the characteristics of a winning horse ready, the next step was to find which horses from the competing field in 2025 come closest to a typical winner. We extracted the relevant data for all horses competing in the 2025 Epsom Derby and then assigned similarity scores. The closer the similarity score to 1, the higher the winning probability of the particular horse. And The Predicted Winner Is... !function(){"use strict";window.addEventListener("message",(function(a){if(void 0!==a.data["datawrapper-height"]){var e=document.querySelectorAll("iframe");for(var t in a.data["datawrapper-height"])for(var r,i=0;r=e[i];i++)if(r.contentWindow===a.source){var d=a.data["datawrapper-height"][t]+"px";r.style.height=d}}}))}(); The supercomputer’s pick to win the 2025 Epsom Derby is Ruling Court, an American origin horse trained by Charlie Appleby. The horse has a similarity score of 0.9857 which translates to a 95% chance of contending for the top spot. That prediction means the supercomputer predicts heartbreak for trainer Aidan O’Brien who’s aiming to add another hat-trick of Epsom Derby wins, having done that previously in 2014. He trained Auguste Rodin and City Of Troy to victories in 2023 and 2024 but as per our supercomputer, his best finish this year would be second, through Delacroix. Other O’Brien trained horses are predicted to finish as follows - The Lion In Winter (4th), Lambourn (5th). It is important to note that we capped the maximum chance of contenting to 95 percent so as to account for a level of uncertainty. Similarly, the minimum threshold was put to five percent. Methodology Data on past winners was collected via irishracing.com and then analysed to determine the characteristics of a typical winning horse Data is from 1995 to 2024 That data for the characteristics of a typical winning horse was then compared to the characteristics of the provisional field entered in 2025 We assigned similarity scores to the field entered in 2025 - the higher the similarity score, the higher the chance of contending Maximum chance to 95 percent so as to account for a level of uncertainty and the minimum threshold was put to five percent