Greyhound Racing Age And Speed Deep Study

Executive Study Summary

This interactive dashboard outlines the research framework for analysing the correlation between age (months) and speed in UK greyhounds. It synthesizes expected outcomes based on biological performance curves, providing a blueprint for predictive modelling and operational decision-making.

Hypothesis H1

30-34 Months

Projected Peak Speed Age

95% CI: [29.5, 34.2] months

Operational Rule

-1.5% / Year

Expected Speed Decline

After breakpoint (42 months)

Data Requirements

N ≥ 200

Min. Races Per Month

For stable percentile estimates

Model Target

AUC > 0.65

Win Probability Model

Using XGBoost with SHAP analysis

Projected Age-Performance Curve (Hypothesis H1)

Visualisation of LOESS smoothed curve with 95% Confidence Intervals.

Age vs. Performance Deep Dive

Detailed breakdown of speed metrics across different distances and age buckets. This section addresses Hypothesis H2 (Distance-dependent decline) and Hypothesis H5 (Injury impact).

Analysis Controls

Distance Category

Show Confounders

Adjust for Injury Censoring

Simulates inverse probability weighting results.

Key Insight (H2)

"Sprinters show a sharper decline post-peak (approx -2.1% per year) compared to stayers (-1.2% per year), likely due to reliance on explosive fast-twitch muscle fibers."

Table A Sample (Metrics)

Age (Mo)	Speed (m/s)	Win %

Detailed Performance Trajectory

Age × Distance Speed Heatmap

Normalised speed (z-score) across age buckets and distance ranges.

Age: 18mo Age: 72mo+

Retirement Survival Analysis

Kaplan-Meier estimates for time-to-retirement. This analyzes survivorship bias: healthier, faster dogs stay in the dataset longer, confounding simple age comparisons.

Kaplan-Meier Survival Curve

Interpretation: The probability of a dog remaining active in the racing pool at a given age. Note the accelerated dropout rate after 48 months.

Feature Importance (Withdrawal Prediction)

Survivorship Bias Note

Older dogs (60mo+) in the dataset represent an elite subset. Raw speed averages for this group are likely overestimated compared to the general population. Recommendation: Use Mixed-Effects Models (Table E) to separate within-dog aging from between-dog heterogeneity.

Operational Rule-of-Thumb Calculator

Translate statistical models into practical decision support. Estimate performance decline based on current age and distance.

Current Dog Age (Months)

18mo 30mo 80mo

Race Distance (Meters)

Caution: This tool assumes average health. Injury history acts as an accelerant not modelled here.

Projected Status

Peak Phase

Exp. Speed vs Peak

0.0%

loss

Time Loss (Standard)

+0.00s

Recommendation

Monitor performance. Optimal physiological window.

Methodology & Data Schema

Technical specifications for reproducibility, data cleaning, and statistical modelling.

Minimal Schema Requirements

Core Identifiers & Metrics

race_id (String): Unique key.
race_datetime (ISO8601): For seasonality.
runner_id (String): Dog identifier.
age_months (Int): Key Predictor.
finish_time_ms (Int): Target Variable.

Critical Confounders

race_class: Controls for grading quality.
trap_number: Bias correction.
weight_kg: Body mass changes.
post_race_injury_flag (Bool): Censoring indicator.

Recommended Analysis Pipeline

1. Data Cleaning

Exclude voided races. Calculate speed (m/s). Z-score normalization by Course/Distance.

2. Changepoint Detection

Apply segmented regression to identify the inflection point (Peak Age) where slope becomes negative.

3. Mixed-Effects Modelling

Random intercepts for `runner_id` and `trainer_id` to account for biological diversity and training quality.

4. Robustness Checks

Stratify by distance buckets. Bootstrap Confidence Intervals for all operational recommendations.

Greyhound Analytics

Menu