What is goto_conversion?
The Sports Forecasting Model that powered $47,000 of prize money, 10+ Gold Medals and 100+ Medals on Kaggle. Kaggle is basically an Olympics for Data Science
Its tournament pages compare separately labeled rating sources through stage-of-elimination probabilities, match predictions, and prediction-accuracy charts.
The project currently covers football, tennis, and rugby. Each sport keeps its own match model and tournament structure rather than forcing every competition into one generic format.
How are forecasts produced?
Forecasts are built and maintained by the goto_conversion project on GitHub. Every tournament page links its specific comparison sources, which currently include The Analyst, World Football Elo Ratings, Nate Silver PELE, Tennis Abstract, Tennis-Data.co.uk, official ATP and WTA rankings, World Rugby, and Rugby4Cast.
This site is an independent goto_conversion analysis. Source names and materials belong to their respective owners. goto_conversion is not affiliated with, sponsored by, or endorsed by any listed source. It is for research and informational use only, not betting, financial, or professional advice.
How are predictions evaluated?
The site reports Avg Probability Assigned, a log-loss-derived score. It is the geometric mean of the probabilities that a source assigned to the outcomes that actually occurred. Higher values are better.
Avg Probability Assigned = exp(-mean log loss)
For a decisive match, the observed outcome contributes the negative logarithm of the probability assigned to that result. Rugby draws are evaluated as the average negative log probability of the two decisive outcomes, matching the site's draw-normalized forecast treatment.
What do the tournament tables mean?
The Stage of Elimination Table shows each participant's probability of exiting at each stage, including the chance of winning. Match Predictions show source-specific probabilities for the next match day. Prediction Accuracy compares sources only after outcomes are known.
Stage probabilities and match probabilities are generated from the same eligible rating snapshot and the same known results for that date.