The pattern that was described
The claim is structural. Moves that go with a larger trend tend to unfold in five legs, labelled one to five, and moves that go against it tend to unfold in three, labelled A, B and C. Legs one, three and five are motive, meaning they subdivide into five of their own at a smaller degree. Legs two and four are corrective and subdivide into three. The whole thing is self similar, so a five leg sequence on an hourly gold chart is meant to be one leg of a five leg sequence on a higher timeframe.
That nesting is the source of both the appeal and the trouble. Appeal, because it gives you a single vocabulary for every timeframe and a reason to expect alternation between trending and corrective behaviour. Trouble, because any sequence of moves can be assigned to some degree of some larger pattern. A framework that can accommodate every outcome has not told you anything about the next one. The useful part is narrower than the vocabulary suggests, and it lives in the handful of statements that can actually be violated.
The three statements that are genuinely rules
Almost everything written about wave counting is guidance. Three statements are hard, and they are the only parts that can be falsified by price. First, leg two never retraces the entire length of leg one. If price trades through the origin of leg one, that count is dead. Second, leg three is never the shortest of legs one, three and five. Third, leg four does not trade into the price territory covered by leg one, with a documented exception for diagonal formations where overlap is permitted.
Everything else, including alternation between sharp and sideways corrections, channelling, and the various Fibonacci proportions that legs often respect, is a tendency. Tendencies are not worthless, but they cannot invalidate a count, so they cannot protect you. The reason to learn the three rules properly is that each one hands you a specific price. That price is the only thing in the whole method you can put a stop behind and still claim you are trading the idea rather than a feeling about the idea.
Corrections are where counts go to die
Corrective structure is where the vocabulary expands beyond usefulness. A correction can be a zigzag, a sharp three leg move. It can be a flat, which comes in regular, expanded and running variants. It can be a triangle, contracting or expanding, with barrier versions. It can be a combination of two or three of those, joined by connecting legs. Each form has its own internal subdivisions and its own permitted proportions.
Count the degrees of freedom there. Nearly any sideways or choppy sequence that gold produces between directional moves can be assigned to one of those labels afterwards. This is the honest weak point of the method and it deserves to be stated plainly rather than buried. When a count spends weeks in correction and the label keeps changing from flat to combination to triangle, nothing is being learned. The label is being fitted to price. If you find yourself relabelling a correction more than once, the correct conclusion is not that the structure is complex. It is that you do not currently have a count worth trading.
Two legal counts on one gold swing
The ambiguity is not hypothetical and it is not rare. Consider a five leg advance that has completed on an hourly chart. One reading says the sequence is finished at its own degree, so the next significant move should be corrective and should hold above the origin of the advance. A second reading says the same five legs are only leg one of a larger five, in which case the next move is a leg two, which can be deep, and the move after that should extend well beyond the current high.
Both readings obey all three rules. Both are drawn on identical candles. They imply opposite things about what to do with a long position at the high. In real time you cannot settle the question by looking harder, because the information that settles it has not printed yet. What you can do is notice that the two readings share an invalidation level, the origin of the advance, and that they differ in what a deep pullback means. Build the trade around the shared level and you have used the framework without pretending to know the degree.
Why the hindsight chart looks so clean
Three things happen when you label a completed move. You already know where it ended, so you choose the pivots that make the rules work. You already know the final extreme, so you know the degree. And you discard the alternates that were live at the time, because they are no longer interesting. The result is a chart with no ambiguity in it, which is a chart that never existed while the move was happening.
This is why reviewing your own counts is more informative than studying other people's. Save a screenshot with the count you held at the time, including the alternate you were carrying, and come back when the move has resolved. You will usually find that the count was reasonable and the confidence was not. The same discipline applies to any pattern based claim, which is why a backtest that lets you choose pivots after seeing the outcome tells you about your drawing rather than about the market.
The part that survives: an invalidation price
Strip out the forecasting and something useful remains. Each rule hands you a price at which a specific count stops being possible. If you believe gold is in a leg two, the origin of leg one is a price you can trade against. If you believe a leg four is forming, the territory of leg one gives you a boundary. A count with no invalidation price attached is a story, and stories do not size positions.
Practically, that means writing the count and the kill level together before entry, placing the stop with reference to that level and the current volatility rather than to a round number, and accepting that being stopped means the count was wrong rather than that the market was unfair. The question of where the stop sits matters more here than in most methods, because the invalidation prices a count produces can be a long way from the entry, and a count that needs a stop you cannot carry is a count you cannot trade.
Gold specifics: releases, gaps and degree confusion
Two features of this market interact badly with counting. The first is scheduled repricing. Gold can cover the distance of an entire leg in a handful of candles around a data release, and inside that burst the subdivisions you were relying on are unreadable. A leg that should contain five smaller legs contains one long candle instead. Dropping a timeframe sometimes recovers the structure and sometimes recovers only noise.
The second is the weekly break. Channelling and proportion both assume a continuous price series, and a Sunday open away from the Friday close puts a discontinuity into the middle of a leg. Counts that depend on a channel touch are unreliable across that boundary. Neither problem is fatal if you treat the count as a frame for where you are in a larger move, and let the actual entry come from something that happens at a price you can see on the live chart. The failure mode is the reverse, deciding the count is correct and then waiting for price to comply.
FAQ
Can two analysts have different counts on the same gold chart?
Routinely, and often both are legal. The rules constrain far less than the vocabulary implies, especially once corrective forms and nested degrees are allowed. That is not a sign that somebody has made a mistake. It is a property of the framework, and the sensible response is to trade only what the competing counts agree on, which is usually a single invalidation price.
What is the single most common mistake with wave counting?
Relabelling. A count that changes every time price makes a new extreme has become a description rather than a forecast. The fix is mechanical: record the count and its kill price before you act, and treat a relabel as a loss of the idea rather than an improvement of it. If you cannot write a kill price, you do not have a count yet.
Do the Fibonacci proportions make a count more reliable?
They make it more specific, which is not the same thing. Proportions are guidance, so they cannot invalidate a count and therefore cannot protect a position. They are useful for choosing between two legal counts and for setting a reference level you will watch. Treat them as a tiebreaker rather than as evidence that the count is right.
Which timeframe should I count on for gold?
Pick the timeframe where the legs are large enough that a single news candle cannot rewrite them, and accept that your degree labels are provisional. Counting on fast charts during a volatile stretch produces constant revision. Counting on higher timeframes gives you fewer decisions and a more durable invalidation price, at the cost of patience.
Is wave counting worth learning at all?
It is worth learning for two reasons that have nothing to do with prediction. It trains you to think in terms of position within a larger move rather than in terms of the last candle, and it forces an explicit invalidation price. If you take those two habits and leave the certainty behind, the time spent is not wasted.
ⓘ See these ideas on real price: open the free XAUUSD live chart.