Overfitting
When a model clings so tightly to its training examples that it captures noise instead of the real pattern, and then fails on new cases.
At a glance
Key signals
14% cross fields · reaches 11 more
- Machine Learning
- Statistics
- Data Science
- Explanation
- Examples
- Misconception 2
- Sourced relations
- Attribution
Dependencies
What this concept builds on and what it makes possible — derived from the atlas’s dependency, causal and structural relations, not from every related edge.
Foundations · builds on
Overfittingdepends onTraining dataEstablished
Overfitting is clinging too tightly to the training data.
Mechanism: Overfitting happens when a model memorises the particular training examples instead of the pattern behind them.
System context
Overfittingis part ofBias-variance tradeoffEstablished
Overfitting is a part of Bias-variance tradeoff.
Overfittingis aTrade-offStrongly supported
Overfitting is one side of a fit-versus-generalise trade-off.
Mechanism: Fitting the training data and generalising to new data pull in opposite directions; overfitting is choosing fit too far.
Structural role & consequence
Interpreted from the current atlas graph — what the connections mean, not just how many there are.
Currently dark in the atlas: no key date stored · 7 of 7 of its relations lack claim-level evidence.
atlas representation · Describes the current Thinking OS representation, not the state of the world.
Builds on 1 foundation (requires / depends-on / derived-from / emerges-from).
structural · Structural graph analysis — not a claim of importance, causation or history.
Structural neighbourhood: 7 → 24 → 71 concepts reachable within 3 hops.
structural · Structural reach — being reachable is not the same as being understood.
cross-field
6 within-field, 1 cross-field
Strengths & constraints
Constraints
- Evidence coverage currently thin in the atlas — few of its relationships carry claim-level evidence. atlas representation
- No dated history stored — the atlas records no key date for this concept. atlas representation
Conditions
- Its dependency reading rests on 1 foundation relation. structural
- Read structurally — most of its relationships carry no external evidence yet, so claims here are graph-derived. structural
Dependency radial
What this concept builds on (left) and what it makes possible (right) — derived from dependency and causal relations.
Seen through each discipline
How this concept sits in each of its fields — derived from its real connections in the graph, not asserted.
Through this lens it connects to Generalization and Training data.
Through this lens it connects to Generalization and Training data.
Through this lens it connects to Bias-variance tradeoff, Regularization, Cross-validation and Dimensionality reduction.
What it looks like
Concrete cases that make this idea recognisable in the real world.
Memorising vs understanding
A student who memorises last year's exam answers aces a re-run but fails a fresh paper. An overfit model is that student: high marks on the questions it saw, poor marks on the ones that matter.
Check yourself
A quick check against a common misconception. Nothing is scored — picking the tempting-but-wrong answer just flags an idea worth revisiting.
Which statement is correct?
Which statement is correct?
A more powerful, more complex model is always better.
More complexity can memorise noise. The right amount balances fit against generalisation.
Look for: Learner always reaches for the bigger or more complex model, assuming more capacity means better results.
A model that fits the training data perfectly is the best model.
Perfect training accuracy often means the model memorised noise and quirks of that data, and it will do worse on new data. The goal is generalisation, measured on held-out data — not a flawless fit to what it already saw.
Where you'll meet it
Journeys that walk you through this idea. You may recognise it from more than one.
Related ideas to explore
Concepts that look related but are not yet connected here — candidates for a connection to reason about, not established links.
This idea also appears in…
The same structure shows up in other disciplines. These are real recurrences drawn from the graph — a starting point for asking “what carries over, and what changes?”
Trade-offs41 disciplines · 33 concepts
Concepts
shares a mental model · crosses a discipline boundary
shares a mental model
shares a mental model
shares a mental model
shares a mental model
shares a mental model
Imagine memorising every past exam word-for-word instead of understanding the topic. You would ace those exact questions and fail new ones. Overfitting is a machine doing the same. Fitting the data and generalising pull against each other — it is a trade-off.
Mental models at work here
- Connects 7 other ideas across 3 disciplines.
- A cross-disciplinary bridge — its connections reach into 11 other fields.
- Most of its connections are of the “Cause & effect” kind.
- It exercises 1 reusable thinking pattern.
- It features in 1 learning journey.
Derived from the graph’s real structure — observations, not a score.
Sources
- Overfitting and Underfitting (2026) verified
- Overoptimism and Overfitting (2021) verified