Comment by radarsat1

Are you just demonstrating overfitting when estimating using too little data? Or is there something deeper going on in your example? What does the bank have to do with anything?