Ralph Nelson Elliott's observation was that markets move in a repeating shape: five waves in the direction of the larger trend, then three waves against it. The five-wave move is called an impulse, the three-wave move a correction.

The part that makes it more than pattern-spotting is that the shape is fractal. Each of the five waves in an impulse is itself made of five smaller waves; each of those, five smaller again. A single count says very little on its own — the claim the theory actually makes is about waves inside waves.

Turn it on from Indicators → Overlays → Elliott Wave on any chart. It is off by default: a count is an interpretation, not a measurement, and it should be asked for.

This distinction is the whole theory, and collapsing the two is how automated counts start lying.

• Rules are requirements. A structure that breaks one — even slightly — is not that pattern. The correct response is to try a different pattern, never to shade the answer.
• Guidelines are reliabilities. They hold often enough (commonly cited at 80–90%) to carry real predictive weight, and fail often enough that treating one as a rule throws away correct counts. They rank legal counts. They never reject one.

So the overlay applies rules as pass/fail and uses guidelines only to choose between structures that already passed. When it rejects something it can tell you which rule broke.

Eleven, not the three or four most tools stop at. The difference matters because a real chart spends much of its time in the shapes that get left out.

Diagonals — where the overlap rule inverts

A diagonal is the one motive pattern where wave 4 must overlap wave 1, inside converging boundaries. A leading diagonal appears in wave 1 or wave A; an ending diagonal in wave 5 or wave C, where it often marks exhaustion.

Corrections — zigzags, flats and combinations

A zigzag is the sharp kind: wave B retraces at least 30% of wave A but never passes where A began, and wave C travels beyond A's end. A flat is the sideways kind: wave B retraces more than 70% of A. In an expanded flat, B carries past where A began; in a running flat, C fails to reach A's end — a sign the trend underneath is unusually strong.

When one correction is not enough, two or three join through connecting X waves into a double three (W-X-Y) or triple three (W-X-Y-X-Z). Those labels are what you will see on a chart that has been chopping sideways.

Triangles

Five overlapping waves labelled A-B-C-D-E inside converging boundaries, each shorter than the one two before it. Triangles appear in wave 4 and wave B positions and resolve in the direction of the larger trend — which is what makes them worth recognising rather than waiting out.

Every wave is decomposed into the next degree down, and a wave may only contain patterns its position permits. Wave 3 of an impulse can never be a zigzag. Wave 2 can never be an impulse. Wave C of a flat is motive; every wave of a triangle is corrective.

Because a sub-wave ends exactly where its parent wave ends, one bar can carry a label at several degrees at once. Those stacks are not duplicates — they are the chart telling you a turn mattered at more than one scale.

When nothing legal fits inside a wave, it is marked unresolved and the count says how many. That is information, not failure: it usually means the bars simply do not resolve that finely.

Once a structure completes, the guidelines say where the next move commonly goes. The overlay draws those levels to the right of the last bar, one set per degree — the smallest structure gives the nearest target, the largest the furthest.

• After a completed five, expect a correction of it. The shaded band is the 50%–61.8% retracement — the most commonly cited zone, and where wave 4 of one lesser degree usually sits.
• After a completed correction, expect the trend to resume. The shaded band is 0.618×–1.0× of the correction, projected from its end: under 0.618 is noise, and 1.0 gives the correction back in full.
• The invalidation is drawn in red as void $X. Past that price the count is not doubtful, it is dead — a retracement cannot exceed the structure it retraces.

The engine builds one nested count and reads it at three magnifications. Major anchors the tree on five years of weekly bars. Intermediate re-counts the last Major wave on daily bars, constrained to the patterns legal in that wave's position. Minor re-counts the last Intermediate wave on hourly bars, subdividing the newest action finest — the trading count reaches the newest bar. A child count cannot contradict its parent, by construction: when Minor points down inside a Major pointing up, that is a pullback within an advance, stated as one coherent sentence rather than two unrelated stories about the same bars.

The degree tabs in the Elliott zones panel are navigation: clicking Major, Intermediate or Minor also moves the chart to that degree's natural view — Weekly · 5Y, Daily · 1Y, 60m · 3M — because fine labels are unreadable on coarse bars and coarse structure is invisible on fine ones. When a refinement cannot be built (hourly history not reaching the wave's start, no legal sub-pattern fitting), that horizon falls back to an independent count and says so — never silently.

• Live-wave projections. The count projects the wave you are in, not only what follows a completed structure — a running wave 3 carries its own target band, updated as it develops. Rows built from a live wave carry a LIVE tag so a forecast-in-progress is never mistaken for a completed one.
• The alternate count. Serious Elliott work always keeps a second count. The engine re-runs the decomposition with the primary's root pattern excluded, so the alternate is structurally different by construction — drawn as a hollow dashed band labelled ALT. Between the primary's invalidation and the alternate's target, both directions are bracketed.
• The channel and the time ticks. The classic Elliott channel (through waves 2 and 4, parallel through 3) draws for the deepest live structure, and Fibonacci time ticks — 0.382, 0.618 and 1.0 of the completed structure's duration — mark when the next turn is commonly looked for, beside the price zones that say where.
• Invalidation in plain sight. Every count states its kill price in the summary strip — invalid past $X — not buried in a tooltip. Past that price the count is not doubtful, it is dead.

The Elliott zones slot on the Charts page reads the same tree as strike zones, tabbed Minor / Intermediate / Major. Clicking a tab moves the chart to that degree's view and draws that level of the count — labels, channel and time ticks follow, with Minor leading by default. The ALT switch beside the tabs swaps everything — the panel's zones and the chart's whole drawn tree — to the alternate count, one count on screen at a time. Panel and chart can never show two different analyses of the same stock.

The count is not a signal and does not tell you to trade. What it does give you is structure around a strike you are already considering — and two levels that are meaningful whether or not you believe the count.

Cash-secured puts

• The invalidation is the level that matters. It is a price the structure must break to be wrong. A put strike below it sits under a level the chart would have to violate before assignment became likely.
• A strike inside the projection zone is a strike in the way. If the structure projects a move down into $84–90 and your put is at $88, the count and the position disagree about the same prices.
• Check the degree against your expiry. A Primary- degree zone describes months. It says little about a weekly put and a great deal about a 45-day one.

Covered calls

• A call strike above the zone is one the structure does not expect price to reach — the case where you keep the premium and the shares.
• A strike inside the zone is where the structure expects price to arrive, so assignment is the likelier outcome. On a name you are happy to sell, that is the plan working, not a problem.
• An ending diagonal or a completed five in the direction you hold is the classic exhaustion reading — the moment the guidelines say a correction is next.

Elliott counts are subjective. The same price data legitimately supports more than one count, which is why firms that do this professionally publish a primary count and alternates. Anything that draws a confident 1-2-3-4-5 on every chart is inventing certainty it does not have.

• Sometimes nothing is drawn. When no structure satisfies the rules for any of the eleven patterns, the chart says so and draws nothing. That is a correct answer and a common one.
• A count can change. New bars can invalidate a structure — the moment price crosses the invalidation, the reading was wrong. Counts are revised; that is how a count can be falsified, which is the only way it can also be useful.
• Unresolved waves are normal. The count states how many rather than hiding them, because a count that always resolves everything is a count that is guessing somewhere.
• It is one interpretation. The summary above the chart names the pattern, the degree and how much of the chart was counted, so a thin reading is not mistaken for a strong one.

Educational only — not financial advice. Nothing on this page is a recommendation to buy or sell anything.