Direct answer: test Swing Highs Lows with a measurable hypothesis
To test Swing Highs Lows, convert the chart idea into a written hypothesis, then evaluate whether a predefined property holds under repeatable rules. A practical testing workflow has five parts: (1) define the pattern precisely, (2) pick a baseline definition and measurement method, (3) collect data and split it into separate time periods or segments, (4) measure results using outcomes net of assumed trading frictions (even if you only use a paper framework), and (5) run robustness checks that stress assumptions.
This is not about claiming prediction accuracy. Historical patterns can fail, and results can change with market conditions, execution quality, costs, and how you define swings.
Mechanism or definition: what you are actually testing
Swing Highs Lows are chart-based turning points. Before you test anything, specify the exact rule for identifying them, because “swing” is often used loosely.
A clear definition usually includes:
- A swing high/low rule: for example, a local maximum/minimum relative to neighboring candles, or a requirement that a point is higher/lower than a number of subsequent/preceding bars.
- A window or lookback length: how many bars on each side must confirm the turning point.
- A tie-breaking rule: what happens if highs/lows are equal or very close.
- An identification timing rule: whether the swing point is labeled only after it is confirmed, which affects any “entry/exit” timing in backtests.
Baseline mechanics
Your baseline should be one concrete set of identification parameters. For example, choose a single lookback window and a single confirmation rule, then keep them fixed during the initial test. This baseline becomes your “control definition.”
Then state your hypothesis in property terms. Examples of hypothesis styles (no prediction promise):
- Occurrence property: Swing highs/lows cluster in time or align with certain market states.
- Reaction property: After a labeled swing high, the next movement has a measurable distribution compared with other points (e.g., median move, tail behavior).
- Structural property: Market structure transitions (e.g., from higher-highs to lower-highs) follow systematically from the swing labeling.
The key is that the hypothesis must map to a measurable variable.
Evidence or example: a testing design you can reproduce
Below is one testing framework that supports self-verification without assuming live data. Use paper or historical candles and treat every assumption explicitly.
Step 1: Write explicit assumptions (kostensoorten + aannames)
Even if your test is conceptual, document costs and frictions because they change whether a strategy-like hypothesis survives.
Common cost categories to model as assumptions:
- Transaction cost: a per-trade spread/commission equivalent.
- Slippage: a small deviation between expected and realized execution price.
- Data quality: candle resolution and whether your source timestamps are consistent.
Assumptions example format (not a recommendation):
- “Commission model: constant per round-trip.”
- “Slippage model: fixed number of ticks (or percentage of price).”
- “Candle resolution: 1H bars.”
Clearly label these as assumptions, because different assumptions can produce different conclusions.
Step 2: Data split to reduce leakage
Use separate time segments. A typical approach is:
- In-sample (training/definition): choose identification parameters and measurement windows.
- Out-of-sample (testing): keep everything fixed and evaluate.
Even if you only test one baseline, use at least two non-overlapping periods. Market regimes vary, so a single period can mislead.
Step 3: Define outcomes and measurement windows
To avoid vague claims, decide what you will measure after each swing label. Examples of measurable outcomes (again, not guarantees):
- Forward return distribution over a fixed horizon (e.g., next N bars).
- Maximum adverse excursion: how far price moves against the hypothesis before a horizon ends.
- Hit rates for structural thresholds: whether a predefined price distance is reached within a time limit.
If you include costs, compute a net metric such as:
- Net movement = measured move − transaction cost equivalent − slippage equivalent.
Step 4: Compute robustness across variable factors
Swing labels depend on variable factors. So test sensitivity by changing one element at a time while keeping others fixed:
- Lookback length for identifying swings.
- Confirmation strictness (how many bars must validate a turning point).
- Horizon length for the outcome measurement.
- Cost levels within a plausible range as assumptions.
This addresses the question: does your conclusion survive small changes, or does it collapse when definitions shift?
Step 5: Compare against a baseline that is not the same hypothesis
A strong test includes a comparison. Examples of comparisons:
- Compare swing-labeled events to randomly selected non-swing points under the same timing rule.
- Compare different types of structural labels (e.g., swing highs vs swing lows) using the same measurement procedure.
The purpose is to detect whether the measured effect is just an artifact of your selection method.
Limitations and risks: what can go wrong
At least one material failure mode should be tested for directly.
1) Definition drift and researcher bias
If you adjust swing detection parameters to make results look good, you risk overfitting to noise. This can create a false sense of effectiveness that disappears on new segments.
2) Timing and confirmation ambiguity
Swing points are often only knowable after later bars confirm them. If you accidentally “look ahead” when measuring, your results become unrealistic.
3) Costs can erase small effects
Even if a reaction after swings exists, net outcomes can change when you model transaction cost and slippage assumptions. If your measured effect is close to zero, results can flip sign.
4) Regime dependence
Historical relationships do not establish future results. Trend, volatility, and liquidity conditions can change, and the swing behavior can become less meaningful or more erratic.
5) Resolution effects
Using different candle sizes (or different data sources) can change which highs/lows are labeled and therefore change your measured outcomes.
Verification or next question: what to check before trusting results
To independently verify the facts behind any claim about Swing Highs Lows, check these items:
- Your swing labeling rules are written precisely and applied consistently.
- Your hypothesis maps to a measurable outcome with fixed windows.
- You used out-of-sample segments or a second period for evaluation.
- You modeled costs and execution frictions as explicit assumptions.
- You tested sensitivity to variable factors (lookback, confirmation, horizon, cost assumptions).
A useful next question is: How sensitive are the results to the swing definition parameters? If small changes in lookback or confirmation radically change outcomes, the “signal” is likely to be definition-dependent rather than robust.