What Swing Definition means
Swing Definition is a self-contained, testable way to describe a “swing” move in price action. In this context, a swing is not a guaranteed pattern. It is a classification you apply to historical or observed price behavior using rules you can repeat.
A useful way to think about “rules” is: you first define what inputs you use (price series and optional filters), then you define what qualifies as a swing (entry/turning behavior and minimum movement), and then you define what makes a swing invalid (failure or overlap rules). If you change any of those elements, you have changed the Swing Definition.
Because markets vary, your rules should separate:
- Stable mechanics: how you measure and classify swings.
- Variable conditions: volatility regimes, liquidity, costs, and how your data is processed.
The “rules” below are written to be independently verifiable using only historical price data and your chosen assumptions.
Direct answer: a testable Swing Definition rule set
Below is a rule set you can explain to someone else and apply consistently. It is intentionally generic; you must choose specific thresholds before you can test it.
1) Choose a time horizon and sampling rule
- Pick a timeframe to analyze (for example, hourly, 4-hour, or daily bars).
- Fix how you sample data: use the close of each bar, or use highs/lows, but do not mix methods within the same test.
Rule requirement: the timeframe and sampling method are part of the definition.
2) Define a swing turning point
A “turning point” is the pivot where direction changes. One testable approach is:
- In an up-swing, the price makes a local low followed by movement that breaks above a prior reference.
- In a down-swing, the price makes a local high followed by movement that breaks below a prior reference.
To make this measurable, define the reference using one consistent method, such as:
- Lookback window rule: a pivot is a bar whose low (for up-turns) or high (for down-turns) is the extreme over a fixed number of preceding and following bars.
3) Set minimum swing size (movement threshold)
To avoid labeling small noise as swings, define a minimum size:
- Range rule: the swing’s peak-to-trough distance must be at least X (in price units) or Y (in percent).
Assumption note: X or Y must be explicitly stated before testing.
4) Define how you identify the swing leg boundaries
A swing leg can be measured from:
- pivot A (turning point) to pivot B (next turning point), or
- pivot A to the most extreme point until the next opposite qualifying pivot.
Choose one boundary rule and keep it consistent.
5) Define invalidation: when your swing definition stops applying
Even if a turning point exists, it may not represent a clean swing leg. Add invalidation rules such as:
- Overlap rule: if price later re-enters and violates the original leg’s structure beyond a threshold, you stop treating the earlier leg as a valid swing.
- Time limit rule: if a swing does not progress within N bars, you classify it as not meeting the swing criteria.
This is a material limitation: without invalidation, two people can label different “swings” from the same chart.
6) Decide how you handle costs and spreads (optional but necessary for comparisons)
If you use Swing Definition to evaluate realized outcomes later, measurement depends on costs. For a purely definitional test, you may ignore costs, but any comparison across providers or execution methods becomes sensitive.
A testable statement you can use:
- “Classification uses price bars only; profitability or returns are not inferred.”
Keeping classification separate from performance avoids turning the definition into a predictive claim.
How Swing Definition works in practice (a concrete example)
This example is written as a classification exercise, not as a trade plan.
Example assumptions
- Timeframe: 4-hour bars.
- Pivot rule: a pivot low (for up-turn identification) is the lowest low over the prior 3 bars and the next 3 bars; pivot high is the highest high over the same windows.
- Minimum swing size: the difference between pivot high and pivot low must be at least 1%.
- Leg boundary rule: measure from pivot low to next pivot high for an up-leg.
- Invalidation: if after the pivot low the price first makes a new lower low than the pivot low by more than a small tolerance (for example, 0.1%), then the earlier leg label is rejected.
Procedure (classification steps)
- Scan the series for pivot lows and pivot highs per the pivot rule.
- Identify an up-leg by finding pivot low A followed by pivot high B.
- Check minimum swing size: compute (B - A) / A. If it is below 1%, label “not a swing” for that leg.
- Apply invalidation: if price revisits beyond the tolerance below pivot low A after A appears, reject that up-leg label.
- For completed up-legs, record the leg size and duration (number of bars from A to B).
What you can verify
- Another person using the same exact assumptions should label the same swings, except for edge cases where bar indexing and the “next 3 bars” overlap decisions matter.
This is why assumptions must be explicit: the definition becomes verifiable only when the rules are fully specified.
Evidence and comparison logic you can run without predictions
Because no single “right” Swing Definition exists, you verify a rule set by comparing it against itself under controlled changes.
Compare both candidate definitions by one criterion
Use a simple compare step: pick a test dataset and run two definitions that differ in only one rule.
Example comparison criteria:
- Swing count: how many swings are labeled per month?
- Average swing size: what is the distribution of measured leg ranges?
- Failure/invalidation frequency: how often does the invalidation rule reject legs?
- Stability across time: do results change sharply when volatility changes?
Compare both options per criterion
For instance, you can compare:
- Pivot rule window: (3 bars on each side) vs (5 bars on each side).
- Minimum size threshold: 1% vs 0.5%.
You then document differences:
- If swing counts rise sharply with a lower threshold, your definition is likely capturing more noise.
- If longer lookback windows reduce swing labels, your definition may lag turning points.