Does a Structural Model Add Anything to the Closing Price? Calibrated forecasting, incremental information, and match leverage in the Italian Serie A

Battle Count: 54

The paper is primarily about sports forecasting, but several methodological elements are directly transferable to quantitative trading: (1) the logarithmic opinion pool as a diagnostic for incremental information over a market price is analogous to testing alpha signals against consensus pricing; (2) the separation of calibration from discrimination mirrors the distinction between well-calibrated risk models and predictive edge; (3) the Shin margin-removal model addresses the same favourite-longshot bias relevant to prediction markets; (4) the walk-forward evaluation protocol and paired bootstrap are standard in backtesting. However, the domain (football match outcomes) and the specific models (Poisson goal models) are not directly applicable to financial time series.

The Dixon-Coles model is well-documented and relatively straightforward to implement (Poisson likelihood with low-score correction, exponential decay weighting, L-BFGS-B optimisation). The Shin margin-removal solver requires bisection on a monotone function. The logarithmic opinion pool is simple. The main complexity lies in the walk-forward harness, proper identifiability constraints, Monte Carlo season simulation, and the post-hoc conditioning for leverage estimation. The author provides a tested Python package with 126 tests, reducing practical implementation burden.

Full code, data pipeline, and figures available on GitHub (https://github.com/pitcany/seriea-leverage). 126 unit tests. Four components subjected to independent adversarial audit. Positive control for look-ahead bias included. Bit-for-bit reproducibility on same machine; three-decimal agreement across machines due to floating-point reduction order. All data from public source (football-data.co.uk).