classification
The odds of default
It is the era when banks first automate the lending decision, and a regulator will read the model line by line. Predict who will default from a handful of financial features — but with a model whose output is a genuine probability and whose coefficients carry meaning: each the change in log-odds per unit of a predictor, defensible in a hearing. A pure black box may rank borrowers a hair better yet cannot answer "why was I denied," which the law increasingly demands. Fit the model, interpret the weights, and be honest about which are causal and which are mere proxies. Get it wrong and you fail the audit, or worse, launder a forbidden variable through a correlated one.
Who this problem belongs to
The two figures whose methods fit it best, out of 71 in contention.
This problem is Cox's own territory almost by name. His 1958 paper on the analysis of binary sequences and his broader development of regression for categorical and survival outcomes gave statistics the logistic link function whose coefficients are exactly the change in log-odds per unit of a predictor this problem demands, defensible to a regulator because the model is a genuine probability, not a black-box score. His proportional hazards model, from 1972, shares the same discriminating instinct: keep the outcome model interpretable and let covariates enter linearly on a transformed scale. He spent a career in applied regulated-adjacent settings, medicine and industry, where 'why was this predicted' was always the second question after 'what was predicted.' Almost nothing else in this roster maps this closely.
Hastie's Elements of Statistical Learning gives the standard modern treatment of logistic regression, its maximum-likelihood fitting, its log-odds interpretation, and its regularized extensions, making him one of the most fluent expositors of exactly the tool this problem demands. His generalized additive models work, developed with Tibshirani, extends logistic regression's interpretable-coefficient philosophy to smooth, still-explainable nonlinear terms, directly useful for a regulator who wants more flexibility without losing an audit trail. He did not originate logistic regression or the log-odds interpretation, so his score sits below Cox and Fisher, but he is deeply fluent in the exact regularization tradeoffs, honest coefficients versus predictive power, that this credit-scoring problem is built around.
Fought here
In the mind map
The same ideas, as concepts rather than history — in John's ML knowledge map.
71 figures are scored on this problem. Draw it in a battle to see where you land.