Curriculum boundary
The 2026 CMT Program Guide places the Fibonacci sequence and golden ratio in Level I theory objectives and asks how retracements and extensions might assist an analyst. Level II public objectives include Fibonacci relationships as applied to Elliott Wave analysis. The public objectives do not explicitly name Gartley, Bat, Butterfly, or Crab harmonic patterns. Those names may be interesting beyond the public scope, but do not infer a required 2026 exam topic from an old third-party article. The official registered curriculum is the final study authority.
This guide explains the mathematics and its role in comparing technical scenarios. For a chart-by-chart anchor-selection exercise, see Fibonacci Analysis Workshop.
Sequence and ratios
The Fibonacci sequence is built by adding the previous two terms: 0, 1, 1, 2, 3, 5, 8, 13, 21, and so on. Ratios of successive larger terms approach about 0.618; the reciprocal approaches 1.618. The square of 0.618 is roughly 0.382. Other charting percentages are often derived by square roots or complementary relationships. A 50% retracement is useful as a midpoint reference but is not a Fibonacci-derived ratio.
The precision of a calculator does not imply precision in price. A level of 96.18 drawn from approximate chart pivots should be treated as a zone for analysis, not a barrier that price must respect to the cent. The selected swing endpoints are often a larger source of variation than the decimal rounding.
Retracement arithmetic
For a prior rise from low L to high H, a downward retracement fraction r is H − r × (H − L). Consider a rise from 40 to 90, a span of 50:
| Fraction | Calculation | Level |
|---|---|---|
| 23.6% | 90 − 0.236 × 50 | 78.20 |
| 38.2% | 90 − 0.382 × 50 | 70.90 |
| 50% | 90 − 0.5 × 50 | 65.00 |
| 61.8% | 90 − 0.618 × 50 | 59.10 |
These levels sit below the high because the question is how far the advance may be retraced. For a decline from 90 to 40, a 38.2% upward rebound is 40 + 0.382 × 50 = 59.10. The same arithmetic levels can arise in reverse order; the direction and hypothesis are different.
Suppose price falls from 90 to 70.5 and then returns to 74. The pullback landed near the 38.2% level of 70.9. The match is an observation, not an explanation of why buying occurred. An analyst might require a close above a short swing high at 75 before treating the bounce as a continuation setup. If price then closes below the pullback low of 70.5, the setup's immediate premise fails. The specific triggers are illustrative personal rules, not official CMT signals.
Extensions and two conventions
A simple extension of the original 40-to-90 advance to 161.8% of its length is 40 + 1.618 × 50 = 120.90. Another analyst might take a 50-point first leg and add 1.618 times that leg to the end of a later correction at 70, producing 70 + 80.90 = 150.90. Both may be called “161.8% projections” in software. The first projects from the original low; the second projects a new leg from the correction. Without specified anchors, the label is ambiguous.
A projected objective is useful when it makes an exit plan explicit. It is not a forecast, and its large reward distance does not establish favorable odds. If entry at 75, invalidation at 70, and an illustrative objective at 90, the reward-to-risk ratio is (90 − 75)/(75 − 70) = 15/5 = 3. The scenario still requires an assessment of failure rate, costs, and position size.
Elliott Wave context
Elliott analysis assigns wave labels to observed swings. Fibonacci relationships can compare candidate counts. Suppose a first wave rises 12 points. A possible wave-three length at 1.618 times that first move is 12 × 1.618 = 19.416 points. If that wave starts from a corrective low at 103, the candidate objective is about 122.42. The calculation does not validate the count: the market can follow another path, and alternate labels may fit the same recent bars. Write down invalidation for a wave interpretation rather than rearranging labels to preserve it indefinitely.
Time projections sometimes use similar ratios, but the public 2026 objectives specifically establish the price-sequence foundations and Elliott relationships noted above. Do not equate a Fibonacci time count with a verified requirement or a precise turn date unless your assigned official chapter supports it.
Confluence and testing
A Fibonacci level near a previous support area may focus attention. Confluence is stronger when the pieces arise from distinct evidence, such as price structure and a separately calculated volatility measure. Several grids drawn from nearly identical pivots are highly correlated evidence. Record every level considered; otherwise a chart full of lines will almost always place one near a later turn.
To study whether a rule adds value, fix anchor criteria and the trigger before examining a holdout period. Compare outcomes with a plain support-level rule or another benchmark, including false signals. If the result changes sharply when anchors move by one bar, the method is fragile. This is a research procedure, not a claim that Fibonacci levels outperform.
Common mistakes and self-check
Avoid reversing the direction of a retracement, treating 50% as a Fibonacci ratio, displaying false precision, and reporting a target without its anchor convention. Also avoid presenting named harmonic templates as publicly verified 2026 curriculum content. For deeper math testing, see Statistics for CMT Candidates.
Self-check: Price rises from 60 to 100. Calculate the 61.8% retracement and a 161.8% extension from the original low. Answer: Span = 40. Retracement = 100 − 0.618×40 = 75.28. Extension = 60 + 1.618×40 = 124.72. State which one you would use only after defining the chart context and decision trigger.
Next, draw one grid from a written pair of anchors, calculate its 38.2% and 61.8% levels by hand, and list both confirming and invalidating price evidence.
Charts load as you approach this section.