# Think in probabilities > Thinking in probabilities is the discipline of judging a decision by the quality of the process and the edge over a series, not by the outcome of any single instance — so a good decision that lost is not mistaken for a bad one, and a bad decision that won is not mistaken for skill. Category: Deals and decisions Also searched as: probabilistic thinking, expected value, process over outcome, edge Source: Mediator Solutions — https://mediatorsolutions.io/learn/#think-in-probabilities License: free to read, learn, cite, and apply, with attribution to Mediator Solutions. ## What it is A single outcome carries almost no information about whether the decision was sound, because any one result is drawn from a distribution. Thinking in probabilities separates the decision from the result: the process is evaluated on whether it holds a real edge across many repetitions, and individual wins and losses are treated as samples, not verdicts. This is what lets an operator stay disciplined through a losing streak that is statistically normal, and stay skeptical of a winning streak that is statistically lucky. ## Why it matters The expensive error is resulting — judging a decision by how it turned out, which rewards luck and punishes sound process the moment variance runs against it. The counter-discipline is to define the edge before acting and to accept that the edge only expresses itself over a series, so no single loss falsifies it and no single win confirms it. The hardest part is emotional, not analytical: holding the process steady while a normal run of losses tempts abandonment, and resisting the overconfidence a normal run of wins invites, because both reactions substitute the latest sample for the distribution. ## When to use it - Judging a decision or strategy after a single win or loss. - A losing streak is tempting abandonment of a process that still holds an edge. - A winning streak is being read as skill and inviting larger, less disciplined bets. ## Principles - Judge the process and the edge, not the single outcome. - A good decision can lose and a bad decision can win; the result is a sample, not a verdict. - Define the edge before acting, so no single instance can confirm or falsify it on its own. - Hold the process through a normal losing run; tighten, do not loosen, through a winning one. ## Practice 1. State the edge and the expected variance before you act, not after you see the result. 2. After each outcome, ask whether the process executed as designed — separately from whether it won. 3. Size exposure to survive a normal losing streak, so variance cannot force abandonment. 4. Review decisions over a series, not one at a time, to see the distribution rather than the last draw. ## Where it fails - **Resulting** — A decision is judged by its outcome, so a lucky win looks like skill and an unlucky loss looks like error. - **Streak panic** — A statistically normal run of losses is read as proof the process is broken, so a sound edge is abandoned at the worst time. - **Streak overconfidence** — A normal run of wins is read as mastery, so discipline loosens and exposure grows just as variance is due to turn. ## In practice An operator runs a process with a real but modest edge and hits five losses in a row. Resulting says the process is broken and should be scrapped. Probabilistic discipline says five losses is an expected sample from the distribution if the edge is small, checks that the process executed as designed, and holds. Over the next fifty instances the edge reasserts. The decision to continue was correct not because it then won, but because it was sound before the next result was known. ## Verification The decision record shows the edge and expected variance defined before the outcome, and reviews cluster decisions over a series — so a reviewer can confirm the process was judged on its distribution rather than on the most recent result. ## Reference ### Separate at every review - The decision quality - The outcome - Whether the process executed as designed - Where the result sits in the expected distribution ### Two errors to resist - Scrapping a sound edge during a normal losing run - Loosening discipline during a normal winning run - Reading one win as skill - Reading one loss as error --- Previous: https://mediatorsolutions.io/learn/#objection-taxonomy Next: https://mediatorsolutions.io/learn/#deterministic-state