Tool
Fibonacci Cluster / Confluence
calc.fibonacci_cluster
Inputs
The first Fibonacci source — a bundle of levels from one retracement, extension, or projection.
The second Fibonacci source, pooled with the others when source_count ≥ 2.
The third Fibonacci source, used when source_count ≥ 3.
The fourth Fibonacci source, used when source_count ≥ 4.
The fifth Fibonacci source, used when source_count ≥ 5.
The sixth Fibonacci source, used when source_count is 6 (the maximum).
An ATR line used to scale the clustering band when tolerance_mode is atr; ignored by the percent and points modes.
Outputs
A bundle of cluster zones — each a price band where enough source levels stacked up, with its member count signalling confluence strength.
Fibonacci Cluster finds price zones where Fibonacci levels drawn from several different swings stack up close together. When levels from many sources land near the same price, that price is a high-confluence area traders watch for reactions.
How it works
You wire in up to six Fibonacci bundles (each one a set of levels from a separate retracement or projection) and tell the node how many of those slots are live with source_count. It pools every level from every source, then groups levels that fall within a tolerance band of each other into a cluster. tolerance_mode picks how that band is measured — scaled by ATR, as a percent of price, or in fixed points — and tolerance_atr_mult / tolerance_percent / tolerance_points set its width. A cluster is only emitted if at least min_members separate levels fall inside it.
Modes
Source count options
Tolerance mode options
The clustering band is scaled by current ATR, so it widens in volatile conditions and tightens in quiet ones.
When to use: The adaptive default — keeps confluence detection sensible across symbols and changing volatility.
The band is a percent of price, so the same setting behaves consistently whether price is 1.10 or 1900.
When to use: When you want one portable tolerance that scales with price level across instruments.
The band is a fixed number of points, ignoring volatility and price level entirely.
When to use: Single instrument where you know the exact price width that counts as the same level.
Configuration
How many of the six source slots are live, from 1 to 6. Only that many bundles are pooled when searching for clusters; the rest are ignored even if wired. More sources means more potential confluence but also more noise.
How the clustering band is measured: 'atr' scales it by current ATR, 'percent' makes it a percent of price, 'points' uses a fixed price width.
The ATR multiple defining the band width when tolerance_mode is atr; e.g. 0.25 groups levels within a quarter of an ATR of each other.
The band width as a percent of price when tolerance_mode is percent; e.g. 0.1 means levels within 0.1% of each other join the same cluster.
The band width in fixed points when tolerance_mode is points; the absolute price distance under which levels merge.
The fewest separate levels a group must contain to be emitted as a cluster. 2 reports any pair of agreeing levels; raise it to demand stronger confluence.
Connects to
Examples
- Wire three retracements from three timeframes, set source_count to 3 and min_members to 3, and trade reactions only where all three agree on a price zone.
- Use tolerance_mode atr with tolerance_atr_mult 0.5 so the confluence band breathes with volatility, then enter pullbacks that tag a cluster zone.
Common mistakes
- Setting source_count higher than the number of bundles you actually wired — empty slots can't contribute levels, so clusters never form.
- Leaving the tolerance band too wide, which merges unrelated levels into one giant 'cluster' that no longer marks real confluence.