What the current curriculum actually supports
The 2026 CMT Program Guide places the Fibonacci sequence and golden ratio in Level I foundations, including how retracements and extensions may help an analyst. Its Level II objectives mention Fibonacci relationships within Elliott Wave analysis. It does not list named harmonic patterns such as Gartley or Bat in those public objectives. Harmonic pattern vocabulary can be useful for broader technical-analysis study, but do not treat the old version of this article as evidence that named harmonic patterns are a required 2026 Level II topic. Verify any such requirement in your assigned official curriculum.
This legacy workshop is about drawing a retracement correctly and deciding what evidence would make a projected level useful. The companion Fibonacci Analysis for the CMT Exam provides a broader conceptual reference.
Construct the ratios
The sequence begins 0, 1, 1, 2, 3, 5, 8, 13, and so on; each new term is the sum of the two before it. Ratios of adjacent large terms approach about 0.618, and the reciprocal is about 1.618. The square of 0.618 is roughly 0.382. A 50% retracement is commonly displayed, but 0.5 is not a Fibonacci ratio. A charting platform may offer 23.6%, 38.2%, 50%, 61.8%, 78.6%, and extension levels. Their presence in a menu does not make each equally useful or uniquely predictive.
For an advance from low L to high H, the arithmetic retracement at fraction r is H − r × (H − L). For a decline from H to L, a rebound retracement is L + r × (H − L). The direction matters. A frequent mistake is to subtract from the low after an advance, creating levels on the wrong side of the swing.
Worked swing: the anchors matter
Suppose price advances from 80 to 120, a 40-point rise. The retracement levels below 120 are:
| Fraction | Calculation | Level |
|---|---|---|
| 38.2% | 120 − 0.382 × 40 | 104.72 |
| 50% | 120 − 0.5 × 40 | 100.00 |
| 61.8% | 120 − 0.618 × 40 | 95.28 |
If price pulls back to 104.5 and then closes at 106, the 38.2% zone has coincided with a rebound. It has not caused the rebound. Now suppose another analyst selects an earlier swing low of 70 and the same high of 120. Their 38.2% level is 120 − 0.382 × 50 = 100.90. That 3.82-point difference is larger than the apparent precision of the decimal labels. The anchor selection should therefore be documented before looking for confluence.
A workable anchor rule might use the last swing low that led to a clearly defined closing high, with pivots identified by a consistent bar rule. Another analyst might use a larger primary swing. Neither choice is guaranteed. Compare both only if each corresponds to a defensible horizon, and do not shift anchors after seeing where the next turn occurred.
Turn a level into a decision
Imagine your 104.72 level lies near prior resistance at 105 that may now act as support. Price approaches on falling volume and then forms a close above the prior day's high. A conditional plan could be: consider a rebound setup only after that close; use a stop or thesis invalidation below the local pullback low of 103; reassess if the broader trend has also weakened. If entry is 106 and invalidation is 103, the initial price risk is 3 per unit. A possible retest of 120 offers 14 points, so the simple reward-to-risk ratio is 14/3 ≈ 4.67 before costs and probability. A favorable ratio does not imply a positive expected value without evidence about outcomes and execution.
Extensions project beyond the original swing. One common arithmetic convention is L + 1.618 × (H − L), giving 80 + 1.618 × 40 = 144.72. Another convention extends from a later corrective pivot. Always state the endpoints and convention; otherwise two “161.8% targets” may differ while both are mathematically legitimate. Treat the level as a scenario for exits or risk planning, not a promise of price.
What counts as useful confluence?
Confluence should come from sufficiently different evidence. A retracement near a prior horizontal level and a change in price structure may be more informative than stacking several ratios from nearly identical anchors. Three overlapping Fibonacci grids can create an illusion of precision because many levels will cluster by chance. Record the number of candidate levels you considered, not only the one that seemed to work.
A test can compare outcomes after predeclared Fibonacci-zone contacts with outcomes after randomly chosen zones or ordinary horizontal levels, using the same market and execution rules. Include costs, failed contacts, and later periods. This is research design, not a claim that one method outperforms another.
Harmonic patterns and Elliott relationships
Harmonic patterns combine ratios across multiple swing legs. Their definitions vary across authors and software, so a named shape without exact anchor and tolerance rules is not reproducible. Elliott Wave work in the public Level II objectives explicitly includes computing Fibonacci relationships. For example, if a hypothetical first wave travels 20 points, a 1.618 multiple is 32.36 points. That arithmetic helps compare candidate wave labels, but the label still needs valid structure and can change as new data arrive. Avoid presenting a ratio match as proof of a wave count.
Common mistakes and self-check
Common errors are reversed anchors, mistaking 50% for a Fibonacci-derived ratio, reporting too many decimal places, and moving swing points to fit a later turn. Also keep arithmetic and log scaling consistent across wide price ranges. Levels are areas for investigation, not precise barriers.
Self-check: A decline runs from 150 to 90. Where is a 61.8% rebound retracement? Answer: The decline spans 60. A rebound restores 0.618 × 60 = 37.08 points; 90 + 37.08 = 127.08. Using 150 − 37.08 would answer a different question.
Next, select one chart and mark the two anchor prices in writing before drawing any grid. Calculate two retracement levels by hand, then define the price behavior that would invalidate your interpretation.