Confluence — Fibonacci with moving averages and candlestick patterns
Our course closes its Fibonacci chapter with three lines: never use Fibonacci on its own — Fibonacci + moving average — Fibonacci + candlestick pattern. Most guides read that as an invitation to stack: the more tools that agree, the better the trade. This lesson takes the sentence literally instead and asks what “agree” means when a Fibonacci line, an average and a zone are all drawn on the same chart. They never sit on the same price. They sit within a few dollars of each other, and those few dollars are not decoration — they are the distance between your entry and your stop. Priced out, that gives one rule most confluence guides do not have, and one case where the usual advice about waiting runs backwards.
The whole lesson in one picture: three tools that say where — the Fibonacci line, the moving average, the old zone — and one candle that says whether. They meet in a strip, not at a point, and the strip is your stop distance. Schematic only, deliberately drawn without a price scale; every measured claim below lives on the charts drawn from the numbers themselves.
KEY TAKEAWAYS
- Confluence is a band, not a point. In the worked example a Fibonacci 0.5 line at $2,270, an old zone at $2,240–2,252, a moving average at $2,246 and a hammer low at $2,230 span $40 — 0.91 of an average candle. That span is where your stop distance comes from.
- Counting a level means inheriting its stop. The stop goes under the lowest level you are using as evidence. Levels inside the band cost nothing; a level below it charges you the overhang. Here, counting a slower average at $2,192 turns 2.87R into 1.76R and lifts the break-even win rate from 25.9% to 36.3%.
- A moving average is the one level that walks. If the average is the lowest thing you counted and it is rising, each bar you wait brings the stop up while the entry stays put: 32.3% → 25.9% break-even over four bars in the example — the opposite sign to the cost of waiting priced in Lesson 23.
- Three “where” tools, one “whether” tool. Fibonacci, the average and the zone say where force might return. Only the candle says that it has. The course’s entry is the candle’s close, not the Fibonacci number.
- The course checks in an order that stops early. Bigger frame permits → the trading frame has the setup → the smaller frame confirms. Fail the first and the rest is not examined. Confluence is not a score.
- On the course’s own second chart the callout names Fibonacci and the shooting star — nothing else. A falling average runs through the same bar, a third “where” answer inside the band, and the course counts it for nothing: exactly what Section 3 says an inside-the-band level costs.
What does the course mean by “never use Fibonacci on its own”?
It means exactly two partners, and it shows a real chart for each. The chapter’s checklist slide is titled What to watch when using it, and beneath its subtitle are three lines and nothing else: Never use Fibonacci on its own · Fibonacci + moving average · Fibonacci + candlestick pattern. The two slides that follow are TradingView screenshots, one for each pairing. There is no third slide about stacking five indicators, and there is no scoring system.
Here is the first of those charts, reproduced from the course.
And the second, which is the more instructive of the two because of what the course chose to label — and what it left unlabelled.
Two charts, two pairings — and on the second one the course names just two of the tools even though a third runs through the same bar. That is the first thing worth taking from it: confluence was never described as “everything lines up”. It was described as Fibonacci plus one thing that answers a question Fibonacci cannot. The rest of this lesson is about what happens when you add the second thing, and the third, and why the arithmetic does not reward you for it the way the word “confluence” suggests.
Why do four tools that agree still name four different prices?
Because each tool is computed from something different, and nothing forces their outputs to coincide. A Fibonacci line is a fraction of one swing. A moving average is the mean of the last n closes. A support zone is a range where price turned before. A hammer’s low is wherever sellers gave up on one particular bar. Put them on the same pullback and they will land near each other — that is what “agree” means — but near is not equal, and the gap between them has a size you can read off the chart.
Take a worked example with illustrative, hypothetical numbers. An impulse runs from A = $2,100 to B = $2,440, so H = $340, and price pulls back. The average candle on the 4-hour chart is $44. The four tools nominate:
| Tool | What it nominates | Type |
|---|---|---|
| Fibonacci 0.5 of A→B | $2,270 (2,440 − 170) | where · fixed |
| Old resistance, now support (polarity) | $2,240 – $2,252 | where · fixed |
| Moving average on the entry frame | $2,246 at the moment of the test | where · moving |
| Hammer | low $2,230, close $2,268 | whether |
Four tools, four prices, and they span from $2,230 to $2,270. That $40 is the band, and it is 0.91 of one average candle wide. Nobody drew it. It is simply the distance between the highest and the lowest of the numbers your toolkit produced.
Lesson 33 established that a zone’s width is your stop, and showed a level getting wider — and more expensive — each time it was retested. The width here has a different source. It is not one level being worn down by repeated visits; it is several tools of different families disagreeing about where the level is in the first place. That distinction matters, because it changes what you can do about it. A worn level only gets wider. A band made of disagreeing tools has a structure — and the next section shows that only two of its four members are charging you anything.
Which of the four tools actually sets your stop?
The lowest one you are counting — and only that one. The rule underneath every confluence trade is simple enough to write in one line: if you count a level as evidence for the trade, your stop has to be below it. You cannot put “the moving average is support” in your reasons for buying and then set a stop above the moving average; if price trades through it, your reason has already failed. So counting a level means inheriting its stop, and the stop ends up below the lowest thing on your list, plus a buffer.
Run the example with a half-candle buffer of $22. The entry is the hammer’s close at $2,268 — the course’s entry, discussed in Section 5, not the Fibonacci line. The lowest counted level is the hammer’s low at $2,230, so the stop sits at $2,208. Risk is $60. The target is the old high at $2,440, so reward is $172, and the trade is 2.87R with a break-even win rate of 25.9%.
Now notice what the zone and the moving average contributed to that number: nothing. Both sit inside the band, above the hammer’s low. Counting them added two reasons to the list and did not move the stop by a dollar. That is the structural fact about a confluence band that the word hides: only the two outer edges charge you. The top edge is where the entry sits, on the candle’s close; the bottom edge is where you stop; everything in between is free.
The right-hand panel prices the opposite case. Suppose a slower average — a 200-period, say — runs through $2,192, below the hammer, and you count it: “price is also holding above the 200.” The stop now has to go under it, at $2,170. Risk is $98, reward is still $172, and the trade is 1.76R with a break-even of 36.3%. One extra reason cost 10.4 points of break-even win rate, and it cost that much for a reason that has nothing to do with whether the 200-period average is a good indicator. It cost that much because it sits $38 below the band.
So the honest version of “add confluence” is a two-column check, and it is not the one most guides teach:
| Where the new level sits | What counting it does to the stop | Worth adding? |
|---|---|---|
| Inside the band (between the lowest counted level and the entry) | Nothing | Yes, if it is a different family of tool — it is evidence at zero cost |
| Below the band | Moves the stop down by the overhang | Only if it lifts your win rate by more than the overhang costs — and you should say by how much before you add it |
| Above the entry | Nothing — but it is not support either | It is a target candidate, not confluence for the entry |
This is the same shape of question Lesson 23 asked about confirmation — does this extra condition earn more than it costs? — with one difference. Lesson 23’s cost arrived through time: each extra condition you demanded let the entry drift away from a fixed stop. The cost here arrives through space, all at once, the moment you count a level that sits lower than the others. And there is a second difference, which is the subject of the next section: one of these levels moves.
What changes when the lowest level you counted is a moving average?
It starts walking towards your entry, and every bar it walks refunds part of your risk. A Fibonacci line does not move once the swing is fixed. A zone does not move. A hammer’s low is printed and stays printed. A moving average is recomputed every bar, and in a rising market it rises. If the average is the lowest level you counted — the one setting your stop — then waiting one more bar does something no other tool in this lesson can do: it narrows the band from the bottom while the entry stays exactly where it was.
Put numbers on it. Keep the same hammer and the same entry at $2,268, but move the average to $2,208 at the moment of the hammer — $22 below the hammer’s low — and give it a slope of +$6 per bar, which is an assumption for illustration. You count it, so the stop goes $22 under it, at $2,186. Then watch what waiting does:
| Bars waited | Moving average | Stop (average − $22) | Risk | R:R | Break-even |
|---|---|---|---|---|---|
| 0 | $2,208 | $2,186 | $82 | 2.10R | 32.3% |
| 1 | $2,214 | $2,192 | $76 | 2.26R | 30.6% |
| 2 | $2,220 | $2,198 | $70 | 2.46R | 28.9% |
| 3 | $2,226 | $2,204 | $64 | 2.69R | 27.1% |
| 4 | $2,232 — now above the hammer low | $2,208 (hammer low − $22) | $60 | 2.87R | 25.9% |
After four bars the average has climbed above the hammer’s low, so the hammer becomes the lowest counted level again and the average stops charging anything. The stop has climbed $22. The entry has moved $0. The break-even win rate fell from 32.3% to 25.9% — 6.4 points refunded, about 1.6 per bar — and you did not give up the moving average as evidence to get it. The number of bars is not a guess either: it is the overhang divided by the slope, $22 ÷ $6, rounded up to 4.
This is worth setting against Lesson 23 directly, because on the surface the two lessons disagree. Lesson 23 priced waiting for a three-way agreement at +33.3 points of break-even, and it was right: in that setup the stop was fixed at a swing low and each extra condition you demanded let the entry drift $2,000 further from it. Here the roles are reversed. The entry is fixed by the candle that already closed; the thing that moves is the stop, and it moves the right way. Both lessons are describing the same quantity — the distance from entry to stop — and the difference is only which end of it is moving. The running tally this site keeps of what confirmation costs, lesson by lesson, has run from +7.9 to +33.3 points, with +11.7 for a wedge and +33.3 for the three-way stack. This is the first entry with a minus sign in front of it.
The sign of the slope is the whole story. Rising average, rising floor, refund. Flat average, nothing happens, and you are simply waiting. Falling average: the floor drops under you and the same arithmetic runs backwards — at −$6 per bar the 2.10R trade is 1.95R one bar later. So “wait for the average to catch up” is not a general rule. It is a rule for one configuration — a rising average that is the lowest level you counted — and it stops applying the bar the average crosses above the next fixed level.
It also comes with two honest limits that the table cannot show. The first is in its own assumption: every row prices a fill at $2,268, so it assumes price is still sitting there to be bought after k bars. Every dollar the fill ends up above $2,268 comes straight out of the refund — at a bar-four close of $2,272, the trade pays 2.63R and breaks even at 27.6%, not 2.87R and 25.9%. The second is that every one of those four bars is a bar in which price can leave. The refund only exists while price waits above the band for the average to arrive, and price does not owe you that. If it rallies first, you have a cheaper trade you are not in. If it breaks the band first, you were right not to be in it. The table prices what waiting is worth; it says nothing about how often you get to collect.
Why can three answers to “where” not replace one answer to “whether”?
Because they are answers to different questions, and the trade needs both. Look again at the four tools in the table in Section 2. Three of them — the Fibonacci line, the moving average, the zone — are computed before price arrives. They tell you where buying might come back. Not one of them can tell you that it has, because they contain no information about the current bar at all: the Fibonacci level was fixed the moment the swing ended, the zone was fixed weeks ago, and the average is a summary of closes that have already happened. Price can trade straight through all three, and it frequently does.
The candle is different in kind. A hammer that closes at $2,268 after printing a low at $2,230 is a statement about this bar: sellers pushed price $38 below where the bar would close, and buyers took all of it back. It carries a timestamp. It is the only tool of the four that does. That is why our course pairs Fibonacci with a candle rather than with a second ruler, and why its note on the tool says that the number is not the entry — the support area is, and the entry is the moment the smaller timeframe turns back up. The Fibonacci line tells you where to look. The candle tells you that what you were looking for has happened.
Three “where” answers do not add up to one “whether” answer. If a Fibonacci level, an average and a zone all coincide and no reversal candle has closed there, you have a location and no event. Adding a fourth “where” tool does not change that. Our course’s own sequence puts it as three filters, checked in order — and the last filter, the one most often skipped, is precisely has the buying actually started, which is a different question from has the selling stopped.
This is also why the entry in the worked example is $2,268 and not $2,270. The Fibonacci line at $2,270 is where you were watching. The close of the hammer is where the evidence arrived, and you can only act on evidence after it exists. Two dollars is nothing on this chart; the principle is not. An order resting on the Fibonacci line is filled in every case where price slices through — Lesson 33 priced exactly that — and an entry on the candle’s close is filled only in the cases where the candle happened. They are different trades with different win rates, and only one of them used the fourth tool.
Go back to the course’s two charts with that in mind. On the BTC chart, C is the bar that touched 7,662.90 and the moving average at once, and turned. On the ETH chart, C is a shooting star on 237.97 with a falling average running through the bar. In both, the thing circled is a candle. Neither slide circles a number.
In what order does the course check all this?
Bigger frame first, and it stops the moment a step fails. The course’s method for an entry is not a list of tools to tick; it is three questions asked in sequence. Does the larger timeframe permit this direction? If not, stop — nothing on the smaller frames is examined, however pretty the confluence looks. Does the timeframe you are trading have the setup? That is where the Fibonacci level, the zone and the average live. Has the smaller timeframe confirmed? Not merely stopped falling — the course separates the selling having stopped from the buying having started — but printed the candle that says buyers are in. It is the one most people skip because the first two already felt like enough.
The order is doing real work. A conjunction checked in sequence with an early stop is not the same object as a score, even when the items are identical. A score lets a strong third item compensate for a weak first one — a beautiful hammer at a perfect level on a chart whose daily trend is down. The sequence does not allow that trade to exist, because it never reaches the hammer. Lesson 21 covered why the frames have to agree before a trend exists; this is the same idea applied to an entry.
The course pairs this with a piece of paper: two columns, factors supporting a rise and factors supporting a fall, written out by hand before the trade. The instruction that goes with it is the opposite of a score. Five items on one side and seven on the other counts as even, not as a win for seven. And when the columns are even, staying out is the third position — a decision in its own right, and usually the correct one. Most losing trades, on the course’s account, are not the wrong choice between long and short; they are a long or a short taken when the right entry was neither.
Put that next to Sections 3 and 4 and the shape of the lesson is clear. Confluence is not a quantity you increase. It is a band whose edges you should know, a set of tools most of which cost nothing to count and one or two of which cost a great deal, one moving floor that can refund you if it is rising, and one candle without which none of the rest is a trade.
When is everything above wrong?
When the market is not trending. Every tool in this lesson assumes an impulse to measure. In a range there is no swing worth a Fibonacci grid — Lesson 20 is the test — the moving average goes flat, so its slope is zero and the refund in Section 4 is zero with it, and old zones get retested until they are used up. Confluence inside a box is three tools describing noise in agreement.
When the average is falling. Section 4 only refunds you for a rising floor. A falling average under your entry widens the stop every bar you wait, and the honest response is either to stop counting it or to stop waiting.
When the tools are the same family. A 20-period and a 50-period average on the same frame are one opinion expressed twice; so are two Fibonacci grids drawn from neighbouring swings. Lesson 23 measured how much a stack of related tools overstates its own evidence, and the answer was a lot. The free evidence in Section 3 is only free if it is actually evidence.
When the band is narrower than the market’s noise. A $40 band on a chart with $44 candles is fine. A $12 band on the same chart puts the stop inside the range of an ordinary bar, and Lesson 33 already priced what a stop inside one candle’s noise is worth: a paper number you cannot hold. Confluence does not shrink the market’s noise; it only tells you where your levels are relative to it.
When the chart is thin. On an illiquid pair the levels are drawn from a handful of prints. Lesson 33 called confluence between two lines on such a chart confluence between two opinions, and nothing here improves on that.
When price does not wait. The four-bar refund is worth 6.4 points only if price is still there to be bought after four bars. Nothing in this lesson measures how often it is, and you should be suspicious of anyone who quotes you a number.
What are the most common confluence mistakes?
| Mistake | What it actually does |
|---|---|
| Treating “more tools agree” as “better trade” | Each tool below the band moves the stop down. Count the cost before the reason: 2.87R became 1.76R for one extra line |
| Entering on the Fibonacci number | Fills you in every case where price slices through. The course’s entry is the close of the reversal candle on the smaller frame |
| Counting a level as evidence and stopping above it | Your reason has already failed by the time the stop is hit. Counting means inheriting the stop |
| Waiting for a falling average to “catch up” | It is moving the wrong way; each bar widens the stop. The refund is for rising floors only |
| Stacking two averages, or two Fibonacci grids, and calling it confluence | Same family, one opinion. Free-looking evidence that is not evidence |
| Checking the candle before the higher timeframe | Reverses the course’s order and lets a perfect hammer justify a counter-trend trade the sequence would have stopped at step one |
| Nudging the Fibonacci anchors until the line sits on the average | Removes the measurement while keeping its appearance. Lesson 34 covers why you can never untangle it afterwards |
What else do people ask about confluence?
What is confluence in trading?
Confluence is when two or more independent tools point at the same area of a chart at the same time — in our course, a Fibonacci retracement level together with either a moving average or a candlestick pattern, and in practice often an old support or resistance zone as well. The word suggests one price, but the tools never produce one price. They produce a band a few dollars wide, and the band is what you trade: the entry sits at or just under its top edge, on the close of the reversal candle, and the stop is below its bottom edge. Lesson 33 called real confluence a level the market built earlier and a ruler that happens to agree with it; this lesson is about what the agreement is worth once a third and fourth tool join in. A worked example with a Fibonacci 0.5 line at $2,270, a zone at $2,240–2,252, a moving average at $2,246 and a hammer low at $2,230 gives a $40 band, which on a chart with $44 candles is about one candle wide.
Does adding more indicators to a confluence setup improve it?
Only if the extra indicator sits inside the band you already have. The stop on a confluence trade goes below the lowest level you are counting as evidence, so a new level inside the band adds a reason at no cost, while a new level below the band moves the stop down and charges you the overhang. In the example, counting a slower average at $2,192 below a hammer low at $2,230 turned a 2.87R trade into a 1.76R trade and raised the break-even win rate from 25.9% to 36.3%. Nothing about the indicator’s quality caused that; its position did. And two tools of the same family — two averages, two Fibonacci grids — are one opinion written twice, which Lesson 23 measured.
Should I wait for the moving average to reach the Fibonacci level before entering?
If the average is below your other levels and rising, waiting refunds risk rather than costing it, which is unusual on this site. The entry is fixed by the candle that already closed; the average climbs towards it each bar, so the stop under the average climbs too. In the example the break-even win rate fell from 32.3% to 25.9% over four bars, the number of bars being the $22 overhang divided by the $6 slope. Two conditions have to hold: the average must be rising, and price must still be there when it arrives. A falling average reverses the arithmetic, and price is under no obligation to wait for you. Once the average passes the next fixed level, waiting stops paying.
Is a Fibonacci level with no candlestick pattern still a valid entry?
Not in our course’s method. The Fibonacci level, the moving average and the zone all say where force might return; they are computed before price arrives and contain nothing about the current bar. The reversal candle is the only tool of the four that says buying has actually begun, and the course’s note on Fibonacci is explicit that the number is not the entry — the entry is the moment the smaller timeframe turns back. Three levels coinciding without a candle is a location without an event. The course’s own ETH chart makes the opposite point too: its callout names a Fibonacci level plus a shooting star and nothing else — a complete setup in the course’s own labelling — even though an average happens to run through the same bar. The candle is not optional; the third ruler is.
PRACTICE CORNER
Do this once on a chart you actually watch, because it turns Section 3 into a habit. Find a pullback in a trend, draw the retracement, and write down every level you would be tempted to count: the Fibonacci line, the average on that frame, any old zone, and the low of the last reversal candle if one has closed. Sort them. The close of the reversal candle is where you would enter — at or just under the highest level — the lowest is where your stop has to go, and the difference plus a half-candle buffer is your risk — compute the R against the prior high before you look at anything else. Then, for each level in between the two edges, write “free” next to it, and for anything below the lowest, write down in dollars what counting it would cost. If the lowest level is a rising average, divide its distance below the next level by its slope per bar and write that number too: it is how many bars of waiting would refund the overhang.
You need a chart with a Fibonacci tool, a moving average you can read a value off, and a candle you can inspect bar by bar. These are the three exchanges this site uses for its own worked examples; all three have the tools, and the exercise costs nothing.
We may earn a commission if you open an account through these links, at no cost to you. It does not change what is written above.
