4 ideas
Rubin's method starts from the premise that nothing can be proven certain, so every decision is a bet across a range of possible outcomes. You identify the plausible outcomes, estimate their probabilities and the size of the payoff or loss for each, and act on the expected value. You keep those probabilities in your head explicitly rather than collapsing them into a single confident forecast.
A good decision can produce a bad result and a bad decision can get lucky, so grading choices by their outcome teaches the wrong lessons. Evaluation should ask whether the probabilities were weighed soundly, whether the information available was used well, and whether the reasoning was careful. Organizations that punish bad outcomes regardless of process push people toward excessive caution and toward hiding reasonable risks.
In January 1995 Mexico was close to defaulting on its dollar-linked tesobono debt, and contagion threatened other emerging markets. Mexico stabilized and repaid the loans early with interest, and Rubin presents the episode as a bet taken because the expected cost of doing nothing was far worse, not because success was assured.
As a Goldman risk arbitrageur, Rubin bet on whether announced mergers would close, pricing the probability of a deal breaking against the spread he could earn. He carried that habit into Treasury, treating policy choices as positions with an asymmetric downside that has to be sized and priced. Seen this way, the question is not whether an action will work but whether its risk-reward profile beats every alternative, including doing nothing.