I saw some banks offer these positions. Are they the same as model validation(MV)? I know what a MV anayst does, but really have no clue about modeling/scoring analyst. What could be the differences, if any? I wonder if some one only the forum has done or is doing related jobs.Any input will be appreciated. Thanks

I am not sure but it might be a position in risk management on the retail banking side, where you need to work on a score model. This model would assign different scores to credit card applicants in order to make a decision to issue a credit card or not and what credit limit to assign. If it is what I think it is, then it's a pretty boring job from quant percpective and is not comparable with model val. A friend of mine was doing score modeling - mostly writing SQL and SAS scripts.

Does it really have to be so boring (modulo the fact that probably all IB jobs are boring)? In a simple case, scoring is done by multivariate logistic regression, but it often requires more sophisticated modelling and optimisation of algorithms. It probably depends on the role details and your experience, but it can be fun.

QuoteOriginally posted by: katastrofaDoes it really have to be so boring (modulo the fact that probably all IB jobs are boring)? In a simple case, scoring is done by multivariate logistic regression, but it often requires more sophisticated modelling and optimisation of algorithms. It probably depends on the role details and your experience, but it can be fun.Before the crisis I knew a guy working on this kind of stuff for a major BB (i.e. scoring systems for retail and small business loans). He had a PhD and was interested in applying Bayesian techniques. However they would never shift away from what they were using (backpropagation neural networks, which amounts to a method of fitting the kind of models that katastrofa cites). And, as he readily admitted, they had a point. Turns out that given enough data (which the bank had) such models are very efficient at making these discriminations (basically a supervised learning problem about the liklihood of default).

QuoteOriginally posted by: neuroguyQuoteOriginally posted by: katastrofaDoes it really have to be so boring (modulo the fact that probably all IB jobs are boring)? In a simple case, scoring is done by multivariate logistic regression, but it often requires more sophisticated modelling and optimisation of algorithms. It probably depends on the role details and your experience, but it can be fun.Before the crisis I knew a guy working on this kind of stuff for a major BB (i.e. scoring systems for retail and small business loans). He had a PhD and was interested in applying Bayesian techniques. However they would never shift away from what they were using (backpropagation neural networks, which amounts to a method of fitting the kind of models that katastrofa cites). And, as he readily admitted, they had a point. Turns out that given enough data (which the bank had) such models are very efficient at making these discriminations (basically a supervised learning problem about the liklihood of default).I also know a guy who works for a consultancy on similar scoring type problems (in this case, the aim being detection of credit card fraud). It seems they are also almost exclusively using neural networks for this.