Question: What factors are most predictive of horse racing outcomes?

Research into horse racing outcomes has highlighted numerous predictive factors that influence race results, with several machine learning and statistical models used to forecast these outcomes. Key factors identified across studies include:

Key Predictive Factors:

1. Jockey and Horse Characteristics:
   • The jockey’s experience and skills, as well as the specific traits of the horse (such as its temperament and physical attributes), have a significant impact on the outcome (Pudaruth et al., 2013).
2. Previous Performance:
   • A horse’s past performance is one of the most reliable predictors, including its finish times, consistency, and overall race history.
3. Race Conditions:
   • Factors such as race distance, track conditions, and number of competitors play a crucial role. Horses may perform differently under various conditions (Martin et al., 1996).
4. Fitness Parameters:
   • Physiological factors like maximum speed, aerobic threshold, blood lactate concentration, and hematocrit levels are associated with racing performance, as these reflect a horse’s overall fitness (Lo Feudo et al., 2023).
5. Purse Amount:
   • The amount of prize money available for the race can affect the level of competition, influencing how aggressively horses are raced (Martin et al., 1996).
6. Horse Age, Season, and Starting Position:
   • Age and seasonality can impact performance, as older horses may be less competitive, and races held during particular seasons may have different characteristics. The starting position (draw) also affects a horse’s chances, especially in shorter races (Wylie & Newton, 2018).

Prediction Models:

1. Machine Learning and Statistical Approaches:
   • Weighted probabilistic approaches (Pudaruth et al., 2013) and machine learning algorithms (Gupta & Singh, 2023) are commonly employed for predicting race outcomes based on these factors.
   • Regularized conditional logistic regression (Silverman & Suchard, 2013) and hierarchical Bayesian analysis (Silverman, 2013) are used to account for multiple variables in a structured way.
   • Artificial neural networks (Davoodi & Khanteymoori, 2010) can model complex, nonlinear relationships between predictors and outcomes.
2. Performance Measures:
   • Common metrics used in predicting and assessing horse racing outcomes include return to racing, number of starts, days to first start, and earnings (Wylie & Newton, 2018).

Overall, predicting horse racing outcomes involves analyzing a wide range of predictive variables like jockey expertise, horse characteristics, and physiological fitness, as well as factors like race conditions and track type. Machine learning and statistical models have become essential tools for synthesizing these variables to make predictions. With continuous improvements in data analysis and model sophistication, the accuracy of these predictions is likely to increase, potentially offering bettors valuable insights into race outcomes.

The amount of prize money available for the race can affect the level of competition, influencing how aggressively horses are raced

Summary of: Martin et al 1996

Anecdote

Have a story to share? Write to us at research@bettingresearch.org if you have a related, personal experience you would like to see placed here and share with the community.

Articles Cited

• “S. Pudaruth, N. Médard, Zaynah Bibi Dookhun (2013): Horse Racing Prediction at the Champ De Mars using a Weighted Probabilistic Approach, https://doi.org/10.5120/12493-9048
  • A computer program was developed to predict the winners of horse races at the Champ de Mars racetrack in Mauritius using a weighted probabilistic approach that analyzes various factors affecting the outcome of a race, and the system was able to outperform professional tipsters in predicting winners.”
• “Meenakshi Gupta, Latika Singh (2023): Predicting Outcomes of Horse Racing using Machine Learning, https://doi.org/10.17762/ijritcc.v11i9.8119
  • The paper by Meenakshi Gupta and Latika Singh (2023) aims to use machine learning algorithms to predict the outcomes of horse races, with a focus on addressing the challenge of imbalanced data using the Synthetic Minority Oversampling Technique (SMOTE).”
• “Noah Silverman, M. Suchard (2013): PREDICTING HORSE RACE WINNERS THROUGH A REGULARIZED CONDITIONAL LOGISTIC REGRESSION WITH FRAILTY, https://doi.org/10.5750/JPM.V7I1.595
  • The authors propose a novel conditional logistic regression model with regularization and frailty to predict horse race winners, and show that it outperforms other published methods when tested on a hold-out year of Hong Kong horse racing data.”
• “C. M. Lo Feudo, L. Stucchi, G. Stancari, B. Conturba, Chiara Bozzola, E. Zucca, F. Ferrucci (2023): Evaluation of fitness parameters in relation to racing results in 245 Standardbred trotter horses submitted for poor performance examination: A retrospective study, https://doi.org/10.1371/journal.pone.0293202
  • The paper investigates the relationship between fitness parameters measured during a treadmill test and racing outcomes, as well as the predictive value of these parameters and racing results for the lifetime racing career of 245 Standardbred trotter horses referred for poor performance examination.”
• “Noah Silverman (2013): A HIERARCHICAL BAYESIAN ANALYSIS OF HORSE RACING, https://doi.org/10.5750/JPM.V6I3.590
  • This study attempts to predict the running speed of horses in horse races using Bayesian linear models, comparing the predictive performance of models estimated using Gibbs sampling and Metropolis-Hastings methods.”
• “Claire E. Wylie, J. R. Newton (2018): A systematic literature search to identify performance measure outcomes used in clinical studies of racehorses, https://doi.org/10.1111/evj.12757
  • The paper systematically reviewed the veterinary literature to identify the various performance measures used in clinical studies of racehorses, with the goal of collating the most commonly used measures and identifying their advantages and disadvantages.”
• “Elnaz Davoodi, A. Khanteymoori (2010): Horse racing prediction using artificial neural networks, –
  • Artificial neural networks were applied to predict horse racing outcomes, and the performance of five different neural network learning algorithms were evaluated on real horse racing data.”
• “George S. Martin, E. Strand, M. Kearney (1996): Use of statistical models to evaluate racing performance in thoroughbreds., https://doi.org/10.2460/javma.1996.209.11.1900
  • The study developed a statistical model to evaluate the influence of factors such as distance raced, racetrack, and track surface conditions on racing performance in thoroughbred horses.”

Insufficient Detail?

At times it is difficult to answer the question as there are not enough relevant published journal articles to relate. It could be that the topic is niche, there’s a significant edge (and researchers prefer not to publish), there is no edge or simply no one has thought to investigate.