BootUp Analytics
Overview
BootUp Analytics was my independent consultancy focused on advanced modelling, recruitment strategy, and decision support in professional women's football. I owned all work end to end, including data engineering, statistical modelling, machine learning, and client delivery, working primarily through written outputs and minimal meetings.
Modeling & prediction
I designed and maintained data pipelines across GCP and AWS to support match-, player-, and season-level analysis. I built probabilistic match outcome and expected points (xP) models using Zero-Inflated Poisson regression to estimate goals scored and conceded while accounting for the frequency of scoreless outcomes in football data.
Match prediction pipelines combined team and opposition VAEP aggregates with contextual features such as home/away and competition type, producing win, draw, and loss probabilities that fed into expected points frameworks.
Player valuation
I developed Expected Player Added above Replacement (xPAR) models to estimate individual player value relative to a statistically defined replacement baseline, typically anchored in the 15th–20th percentile. Replacement level was defined to remain stable across leagues, seasons, and positional groups, separating analytical benchmarks from recruitment intent.
Player value was calculated on a per-match basis rather than season totals, allowing assessment of form and contribution independent of minutes bias or team context. League normalization frameworks were implemented to contextualize performance across competitions of differing strength and to support translation between leagues.
Data quality & delivery
Historical data was weighted using exponential decay to balance recency and sample size, with additional season-weighting logic to control volatility and aging effects. I designed SQL data models and validation pipelines to detect issues such as duplicated matches, inflated minutes, and implausible player records, supported by extensive exploratory analysis.
I led player transition modelling projects, including NCAA to NWSL pathways, building longitudinal datasets, engineering development features, and training classification and regression models to estimate professional outcomes. Analytical outputs were delivered as recruitment shortlists, player valuations, and match preparation materials designed to be interpretable without technical explanation.