Epsilon-Optimal Decision Making

mental-model

An epsilon-optimal decision accepts any option close enough to the best that the remaining difference does not justify further search. “Good enough” is a deliberately high standard that preserves attention, time, and momentum.

The rational choice may be an option you already know is not the absolute best. If it falls within epsilon—a deliberately small distance from the optimum—choosing it still counts as a high-quality decision.

E1

Price the gap, not the ranking

Rankings hide magnitude: first and second place may be separated by almost nothing. Epsilon makes that difference explicit. Once an option lies inside the acceptable band around the best, further comparison is chasing an improvement defined as too small to matter.

This changes the objective from finding the winner at any cost to capturing nearly all available value while conserving the resources required to search. The saved attention and time can then move to decisions where the gap is consequential.

E1

Epsilon cannot excuse a careless threshold

The model is only as sound as the tolerance you set. When tiny differences compound, carry severe downside, or are difficult to reverse, the meaningful epsilon may be extremely small. Calling a large gap “close enough” is not epsilon-optimality; it is merely lowering the standard.

Set the stopping band before comparing

For tomorrow’s bounded decision, define the smallest improvement that would actually change the outcome—for example, a material difference in cost, time, or quality. Stop searching as soon as a candidate falls within that distance of the best available option, and spend the saved effort elsewhere.

Episodes that teach this