08 — Emotional Drivers

Regret Aversion#

Heuristic1 min read
◆ The PatternFear of making a decision you'll regret.

Regret aversion — making decisions to minimize future regret rather than maximize expected value.

Two forms:
• Errors of commission: "I bought → it crashed → I regret acting"
• Errors of omission: "I didn't buy → it soared → I regret NOT acting"

Short-term: we regret actions more. Long-term: we regret inactions more.

In markets, regret aversion causes paralysis, herd-following ("at least I won't be the only one who's wrong"), and excessive conservatism.

Worked example

Three choices for a position you are up on, scored in a rally, a flat market and a drop. By expected profit, holding wins: 6.25, against 3.13 for selling half and 0 for selling now. Now score each by its worst regret — how far it falls short of the best choice in hindsight. Selling now can miss 30, holding can miss 20, selling half at most 15. The regret-minimiser sells half, giving up half the expected profit to never feel too foolish.

Where it misleads
  • Splitting a decision is not always regret-dodging; it can be sensible sizing when you are unsure. The question is whether the split is chosen for the odds or for the feelings.
  • People regret actions more in the short run and inactions more in the long run (Gilovich & Medvec 1995), so the same rule pushes in different directions over time.
  • Rules set in advance — when to sell, how much — remove most of the room regret has to work in.
In code
pay = pd.DataFrame({'sell now': [0, 0, 0], 'sell half': [15, 2.5, -10], 'hold': [30, 5, -20]},
                   index=['rally', 'flat', 'drop'])      # profit from here in each outcome
p = np.array([0.35, 0.35, 0.30])                         # your odds for the three outcomes
ev = pay.T @ p
regret = pay.max(axis=1).values[:, None] - pay.values    # best choice in hindsight minus yours
worst_regret = pd.Series(regret.max(axis=0), index=pay.columns)
Python, in your browser — edit the code and run it again

Assumes import numpy as np and import pandas as pd. Each snippet carries its own example numbers; the comments say which are made up.

Sources
  • G. Loomes & R. Sugden, “Regret Theory: An Alternative Theory of Rational Choice Under Uncertainty”, Economic Journal 92(368), 1982
  • D. E. Bell, “Regret in Decision Making under Uncertainty”, Operations Research 30(5), 1982
  • T. Gilovich & V. H. Medvec, “The Experience of Regret: What, When, and Why”, Psychological Review 102(2), 1995

Last reviewed October 2026 · How this site checks its content · Report a mistake

Pattern bridge: Avoiding regret over avoiding loss is a second-order bias. In ML, regularization trades some training error for less regret on unseen data.
Linked from 1
← Previous
Loss Aversion
Open in the full reader, with the topic sidebar