Find another term. Follow an explanation link; use Back to return here. Behavioural terms describe tendencies, not diagnoses or instructions to trade.
Anchoring
Giving an initial reference excessive influence over new judgements. Waiting for purchase-price recovery regardless of a holding’s current purpose or spending deadline is an investment example.
Backtest
A historical simulation of stated investment rules, distinct from live investing. Data availability, costs, assumptions and choices made after inspecting results determine what it can establish.
The search behind a simulation
Benchmark
A defined performance reference, such as an index/mix. Match exposure, dates, currency, income and costs. Outperforming it still differs from having enough for your goal.
Confirmation bias
Overweighting supportive information while dismissing contrary evidence. Reviewing a favoured holding can become collecting justifications; identify evidence that could change the rationale.
What could change the decision?
Data mining
Searching data for patterns. Many holdings/dates/rules can produce impressive chance results; hiding that search overstates evidence. A discovered pattern needs a credible explanation and further tests.
How many alternatives were tried?
Expected return
A probability-weighted estimate of possible returns, dependent on assumed outcomes/probabilities. It need not occur and differs from historical averages or promises. More risk compels no estimated reward.
FOMO
Fear of missing out: pressure to join because others seem to gain or an opportunity seems fleeting. It can rush decisions; another investor’s gain establishes no fit with your finances.
Hindsight bias
Treating an event as more predictable after knowing its outcome. Retrospective explanations can hide once-plausible alternatives. Assess anticipation using information actually available before the decision.
In-sample and out-of-sample
In-sample data develop/select an approach; out-of-sample data test it independently. Repeated inspection and redesign can compromise that separation. The label guarantees neither independence nor future performance.
Look-ahead bias
Using information in a simulation before investors could have known it, such as revised statistics or later index membership. Correct arithmetic cannot make such historical decisions feasible.
Loss aversion
Giving losses greater weight than comparable gains relative to a reference point. It can encourage risk-taking or avoidance, differs from financial capacity and implies no universal loss-to-gain ratio or behaviour.
Model risk
Risk of misleading decisions from a model’s design, data or assumptions, including omitted/understated losses. More simulations using unchanged assumptions cannot repair their omissions.
Outcome bias
Overweighting the eventual result when judging a decision. Gains can follow poor reasoning and losses sound reasoning. Results remain evidence alongside the original purpose, information, alternatives and risks.
Overconfidence
Used here for treating a judgement or forecast as more certain than its evidence warrants. As a decision check, ask whether a few successes justify the confidence behind a larger commitment, and what being wrong about that amount would mean.
Overfitting
Adapting an approach to data-specific features that generalise poorly. Selecting simple static portfolios can overfit too. One poor later period alone does not prove it.
Regret aversion
Used here as a decision check: how much of the choice is driven by avoiding future regret? Consider both acting and waiting against the plan, rather than assuming either response will prevent regret. This is an editorial prompt, not a claim that regret always makes investors more cautious.
Reasons for action or inaction
Rolling window
A fixed-length period moved through a dataset. It adds starting dates, but neighbouring windows can share almost all observations; window count is not a count of independent investment experiences.
Selection bias
Distortion from how observations are chosen. Successful investors/holdings or favourable dates can misrepresent the opportunity set despite correct displayed arithmetic. Examine selection rules and exclusions.
Survivorship bias
Selection bias from including only surviving/available investments. Closed or merged funds belong to the original opportunity set; closure need not mean total loss. Establish their returns and proceeds treatment.
Track record
Results attributed to an investment/manager. Identify live, simulated or mixed history and mandate changes; assess period, risk, costs and completeness. A long chart need not represent long live management.
Uncertainty
Incomplete knowledge of outcomes or probabilities. Estimated success depends on its model/plan; this handbook keeps those limits visible without assuming reliable numerical probabilities exist for every important risk.
Related materials
- Bonds and interest-rate risk: investing glossary
- Currencies and investment infrastructure: investing glossary
- Luck, skill and a convincing story
- Fear, FOMO and other people's returns
- Count the investments that disappeared
- When a backtest has learned the past too well
- See how the chosen dates change the story
- Uncertainty and rare events