In  and , the authors give formulas for the expected true error of various resampling protocols (CV, bootstrap) associated with a score function Q. I'm not sure to understand exactly what is this notion. In particular, does is it have any relationship with the bias of CV (or boostrap) as an estimator of Q? Can this be used to detect under/overfitting (with a measure of variance)?
 "Cross-Validation, the Jacknife, and the Bootstrap: Excess Error Estimation in Forward Logistic Regression", Gong G., Journal of the American Statistical Association, 1986
 "Improvements on Cross-Validation: The .632+ Bootstrap Method", Efron B. and Tibshirani R., Journal of the American Statistical Association, 1997