Why yellow pad math still beats the spreadsheet in builder deals

[In the third of seven excerpts from ‘Less Wrong, More Right,’ Scott Cox argues that models can hide what matters by weighting every input equally. Simple back-of-the-envelope comparisons can identify what must be true for a homebuilder deal to work and where opportunity may be underpriced.] One of the most impressive people I got to observe early in my career was a vice chairman of the bank, widely considered the smartest person in the place. What made him remarkable had nothing to do with depth in any particular market. He would be handed a credit write-up or a new project package – often a deal in a market he did not know, on a product type he had not financed, for a borrower he had never met — and he would flip pages for two or three minutes. Then he would tell you what the deal’s success or failure was actually depending on. I never saw him build anything. He did a few simple calculations on the relationships between the numbers, and the relationships told him where to look. He was not recognizing the borrower and he was not recognizing the market. He was recognizing a pattern that lacked internal consistency – or one that could only hold if something were true that nobody in ...