Definition first: what “swing high” and “swing low” mean
A swing high (or swing low) is a local turning point in a price series, identified using a rule-based pattern of surrounding bars. In practice, you need a definition that specifies what “surrounding” means and how many bars must confirm the turning point. Without that rule, two people (or two tools) will often label different highs and lows from the same chart.
Direct answer: data you need to assess swing highs lows
To assess swing highs lows reliably, collect four categories of inputs.
1) Price data inputs (the raw series)
You need a complete time-ordered price dataset for the instrument you are studying. At minimum, that usually means open, high, low, and close values (OHLC) for each bar, or equivalent tick-to-bar data if you later build bars.
Materiality note: a swing high/low is defined by the bar extremes (high for a swing high; low for a swing low), but the context bars are required too—so gaps or missing candles can distort the turning-point detection.
2) The swing-detection rules (the algorithm choice)
You need to state the detection rules you are applying, including:
- The timeframe (e.g., 1H bars vs 1D bars).
- The “left” and “right” look window: how many prior and subsequent bars must be lower/higher.
- The tie-handling rule: what happens if highs are equal or nearly equal.
- Whether you require strict inequality or allow “within tolerance” comparisons.
These rules are not market facts; they are methodological assumptions. Two rule sets can produce different swing labels even with identical price data.
3) Context and provenance (where the data comes from)
You should record:
- The data provider or source feed (for example, a charting platform, an exchange feed, or a data vendor).
- Any transformations applied (e.g., resampling ticks into bars, timezone alignment, corporate-event adjustments if applicable).
Because swing labels depend on how bars are constructed, provenance is essential for independently verifying claims.
4) Timeliness and data quality checks
You need checks that the dataset is consistent and usable:
- Confirm no missing bars in the assessment window.
- Check for timezone or daylight-saving shifts that can shift bar boundaries.
- Ensure the bar construction method is consistent across the period.
- If using historical data, clarify the snapshot version: later revisions by providers can change stored history.
How it works: turning inputs into swing labels
Assessment becomes a repeatable process once you combine the four categories:
- Choose the timeframe and build/obtain the OHLC series.
- Apply your swing-detection rules to each candidate bar.
- Mark a bar as a swing high if its high is higher than surrounding highs per your left/right window rules (and similarly for swing low using lows).
- Store the rules and provenance alongside the resulting swing points so others can reproduce the same labeling.
Evidence/example (rule sensitivity)
Suppose you use a 3-bar left and 3-bar right window on a 1H chart. A bar that appears to be a swing high under a strict “must be higher than all surrounding highs” rule may no longer qualify if you switch to a larger window (e.g., 5 left and 5 right) or if you allow a tolerance for “nearly equal” highs. The swing structure is therefore not only a property of price; it is also a property of the rule set and window.
Limitations and risks (material failure modes)
At least one major limitation should be considered any time you assess swing highs lows:
- Rule disagreement: Different, equally reasonable swing definitions produce different swing points. This can change any downstream analysis.
- Missing or inconsistent data: Gaps, incorrect bar boundaries, or provider revisions can create or erase turning points.
- Timeframe dependence: Swing points on a 1H chart are not the same objects as those on a 1D chart, because the bar aggregation changes which highs/lows are locally extreme.
- No guarantee of predictive meaning: Historical turning points do not establish that future prices will behave similarly.
Verification and next questions
To verify what you are seeing, compare the labeling process against at least one independent method:
- Re-apply the same stated rules on the same timeframe using a different chart/data source if available.