Home › Learn › Lesson 16

Bull, bear and sideways — the three phases and how each one behaves

By TradingPrimer · Updated 19 Sep 2026 · Published 2 Sep 2026

Most explanations of market phases stop at the definitions, which are the easy part and which Lesson 15 already gave you. The useful question is the next one: what does each phase do — to your pullbacks, to your tools, to the number of times a year you are allowed to have an opinion? This lesson answers that, and along the way it turns the course's sharpest line on the subject — if the timeframes point in opposite directions the market will go sideways — into arithmetic. That arithmetic produces one result worth the whole lesson: the gap between a bull and a bear phase is not an event, it is a stretch of time you cannot avoid, its floor can be calculated before it happens, and its real length has now been measured on two years of Bitcoin.

Real BTCUSDT candles, one moment, three zoom levels. The weekly is still making higher highs and higher lows, the four-hour is already making lower highs and lower lows, and the daily cannot decide. No panel is wrong. Their disagreement is the sideways — and the swing labels are the 2-bar swings that had been confirmed by that moment, which is why the four-hour’s lowest wick has no label yet.

What are the three phases, in one paragraph each?

Two of them are shapes you can point at on a single chart. The third is not, and that difference is the whole lesson — but take the definitions first, because everything later is built on them. One note on convention: on this site the phase is read first from the RSI wave (Lesson 19) — a wave counts only once RSI has crossed both its EMA 9 and WMA 45 and the two averages have crossed the same way, and a frame whose wave has lost validity hands the call to the nearest larger frame that still has a live wave — then from the MA bundle (Lesson 18), and from structure last, with a 5-bar price-only pivot; the structural definitions below are that third, least clear read.

A bull phase is a sequence of higher highs and higher lows. Price makes a peak, pulls back, and the pullback stops above where the last pullback stopped; then it makes a new peak above the last one. Buyers are willing to pay more than they were willing to pay recently, and sellers are not managing to push price back to where they last managed to push it. The course this site is built from calls that a durable bullish structure, and adds a qualifier worth keeping: the higher the timeframe the structure forms on, the more durable it is.

A bear phase is the same machinery inverted: lower highs and lower lows. Each rally stops short of the last rally; each decline goes further than the last. But the course adds one behavioural note here that it does not add to the bull case, and it matters later: in a bear phase, every bounce tends to be short and to end early. That is not a symmetric statement. Hold on to it.

A sideways phase is where the tidy part ends. The conventional definition is "price moves within a range without a clear direction", which is true and nearly useless, because it describes what you see rather than what is happening. The course gives a mechanical definition instead: the market goes sideways when the timeframes point in opposite directions — and, put positively, for a trend to form at all it needs the timeframes to agree. The reasoning behind it is that each timeframe stands for a different group of participants, with different capital, different horizons and different exit points. When those groups want the same thing at the same time you get a trend. When they want different things you get chop.

Why is sideways not really a third phase?

Because bull and bear describe one chart, and sideways describes the relationship between several. They are not three items on the same list — the third one is a different category of thing, and treating it as a peer is what makes it feel so slippery.

Try to state the three definitions in parallel and the asymmetry shows up immediately:

• Bull: this chart is making higher highs and higher lows.
• Bear: this chart is making lower highs and lower lows.
• Sideways: these charts do not agree with each other.

The first two are properties of a single object. The third is a property of a set. You cannot look at one chart and know you are in a sideways market in the course's sense, any more than you can look at one person and know a committee is deadlocked. And this is not a quibble about wording — it changes what you do. Under the conventional definition, "is it sideways?" is answered by looking harder at the chart in front of you. Under this one, it is answered by opening the other charts.

It also explains something that otherwise looks like a contradiction. You can be in a sideways market while one of your charts looks strongly trending. The 4-hour can be making clean lower highs and lower lows — a textbook bear structure — while the weekly is still making higher lows. Nothing is wrong with either reading. They are both true, they are describing different groups of participants, and their disagreement is exactly the condition the course is naming. The 4-hour trader who says "this is clearly a downtrend" and the weekly investor who says "this is clearly still an uptrend" are both right, and both are trading a market that is, at the level of the whole, going nowhere.

One consequence to bank now: "the market is sideways" is an unfinished sentence. Sideways across which set of frames? Two frames that are one step apart disagreeing is an ordinary afternoon. Ten frames from daily to monthly disagreeing is a genuine regime. Same word, wildly different situations — which brings us to the arithmetic.

How much of the time should you expect to be in a trend?

That depends entirely on a number you choose rather than one the market gives you: how many frames you insist must agree. And the relationship is steeper than almost anyone expects.

Set up the simplest possible yardstick. Suppose each frame you follow reads either up or down, and — deliberately assuming nothing about markets — suppose each reading is an independent coin flip. Then a trend, defined as unanimity, requires all n frames to land the same way:

P(all n frames agree) = 2 × (½)ⁿ = 2¹⁻ⁿ

The factor of two is because there are two ways to be unanimous — all up or all down. Run it:

Now read that table for what it is: a null. It says nothing about how markets behave — it is the amount of unanimity you would get from charts that carried no information at all. So we checked it against a real one. We scored BTCUSDT every day at 00:00 UTC for two years (19 Sep 2024 to 18 Sep 2026, Binance spot candles) on exactly the three frames the practice corner below asks you to use — 4-hour, daily, weekly — with a 2-bar swing setting and the two Lesson 15 questions, counting only swings that had already been confirmed by that moment. The reproduction details are at the foot of the page.

The frames agreed far less often than the coin flip — about a quarter as often at three frames — and the last row says why. A coin has two faces; a frame has three. Each frame spent roughly a third of its days reading “neither” (weekly 36%, daily 32%, 4-hour 41%), and one “neither” anywhere in the set breaks unanimity. Redo the null with three equally likely faces and you get 2 × (⅓)³ = 7.4%. Use each frame’s measured mix of up, down and neither and assume the frames are independent, and you get 6.7% — against 6.6% observed. On this pair, over these two years, the three frames agreed about as often as if they had nothing to do with each other.

That is the opposite of the reassuring sentence this page used to carry here — that real markets beat the null by a wide margin and the gap measures persistence. Persistence is real, but it does not live between frames; it lives inside each one. The weekly kept whatever label it had from one day to the next 98% of the time, with a median run of 42 days; the daily 88% and 6 days; the 4-hour 46% and a single day. Trends persist within a frame. Agreement across frames is a different and much rarer thing — and the faster your fastest frame, the rarer, because a 4-hour chart on a 2-bar setting changes its mind roughly every other day.

What both tables say, unambiguously, is that the strictness of your definition, not the market, sets how often you are allowed to act. Move from three frames to five and you have quartered your opportunities before a single candle has printed. That is worth knowing because the strictness usually arrives by accident: somebody adds a fourth chart to be safe, then a fifth, and then wonders why nothing ever qualifies. Nothing qualifies because they made it not qualify.

A note if you have read Lesson 15: 25% and 6.25% appear there too, and they mean something different. There they came from (¼)^k — the chance of k consecutive higher-high-and-higher-low pairs on one chart, over time. Here they come from 2¹⁻ⁿ — the chance of n charts agreeing at one moment. Same digits, different machinery, and they do not combine.

Why can a bull phase never flip straight into a bear phase?

Because the frames are nested inside each other, so they cannot confirm a new phase simultaneously — and if they cannot confirm it simultaneously, there is necessarily a stretch of time when some have and some have not. That stretch is the sideways phase, and it is compulsory.

The argument is short enough to check. A weekly candle is built from seven daily candles; a daily is built from six four-hour candles. The higher frame has no information the lower frames have not already delivered to it. So the sequence of confirmation is forced: a weekly swing needs weekly candles to close before it exists, and every one of those closes is also a daily close — the fast frame has always had the chance to register a new structure before the slow one can. A weekly structure cannot be confirmed before the daily candles that compose it have closed.

Which means a market that begins with every frame pointing up and ends with every frame pointing down must pass through states where they disagree. There is no path that skips them. Draw the three states and you cannot draw an arrow from the left box to the right box:

This is a statement about definitions, not about markets, and that is what makes it strong: the slow frame cannot confirm a new structure before its own candles close, and no data can contradict that. It also disposes of a fantasy that costs beginners a lot of money. There is no clean flip to be caught. Anyone offering to call the exact top of a bull phase is offering something the structure of the problem forbids — at the moment of the top, most of the frames still read bullish, because they have not had time to register anything. The top is only ever a top in retrospect, and the retrospect has a minimum length.

How long a minimum? The identity from Lesson 17 gives the only hard number: a swing point marked with a setting of n bars cannot be confirmed until n further bars have closed, because you have to see that nothing exceeded it. On a weekly chart with a 2-bar setting that is two more weekly candles — 14 days — before the weekly can register any new swing at all; and a new structure needs both a new high and a new low, so in practice the wait is longer. Everything beyond the 14 days is a measurement rather than a theorem, and here is ours. On BTCUSDT between September 2024 and September 2026 the three-frame score went from all-up to all-down or back five times. The disagreement window in between lasted 70, 32, 58, 47 and 135 days — never under a month, median 58. And in all five, the order in which the frames confirmed the new phase was the same: 4-hour first, daily second, weekly last. One honest wrinkle, because leaving a phase and confirming the next are different events: the weekly dropped its old label — to “neither”, not to the opposite — before the daily did in four of the five, because a lower weekly high can print while the daily’s own last two swings are still rising. What the nesting makes overwhelmingly unlikely is the slow frame confirming a new phase first, and in five flips out of five it did not.

The 2-bar setting is a choice made for this measurement, and a stricter one lengthens the floor proportionally — the site’s own convention is a 5-bar price-only pivot, so its weekly floor is 35 days. What does not change is that the floor grows with the slowest frame you consult — and the measured windows show how far above the floor the typical wait sits.

The practical version: if you follow a weekly frame, then for a month or more after a major top your frames will be in conflict — on BTC it was never less than 32 days — and no amount of staring will resolve it early. That period is not a failure of your analysis. It is the analysis working correctly and telling you it does not know yet.

Does a bear phase behave like a bull phase upside down?

No — but the asymmetry is not quite the one the textbook draws, so this section does two things: the arithmetic first, then what two years of BTC did to it. The course's description of a bear phase includes a detail it never applies to the bull case: bounces are short and end early. Price that in R and the two phases stop looking like mirror images.

Take the identity from Lesson 33 and trade with the phase: in a bull phase buy the pullback of an up-leg, in a bear phase sell the bounce of a down-leg. Either way, enter at depth f of the last leg, put the stop beyond the origin of that leg, target the leg’s extreme, and the trade pays:

R:R = f ÷ (1 − f)     break-even = 1 − f

The size of the swing cancels out entirely, which is what makes this comparable across phases. Now put the two typical depths side by side:

Same rule, same discipline — and the deep retracement pays 2.618 times the shallow one, with 23.6 percentage points less win rate required to break even. That is the arithmetic, and it is exact. The open question is whether bear phases really hand you shallower retracements than bull phases. So we measured it, on the same BTCUSDT data and the same 2-bar swings, classing each leg by its own frame’s label at the moment the leg’s end was confirmed:

Two things in that table matter more than the classical numbers. First, on this pair the bear-phase bounces were not shallower. If anything they retraced a little more of the leg than bull-phase pullbacks did — 0.87 against 0.76 on the 4-hour. The course’s line that bounces are short and end early was not visible here at this swing setting; it may be a statement about time rather than depth, or about other markets, and it goes into “when this lesson is wrong” below rather than into the arithmetic. Second, the asymmetry that did show up is the one that actually empties accounts. Trade against the frame’s label and the origin of the leg — where your stop sits — was taken out about 60% of the time; trade with it and the figure was 38 to 41% on the 4-hour and 19 to 36% on the daily. That is what “carrying bull-phase dip-buying into a bear phase” costs, and it is a larger effect than the R:R shift, because the identity assumes the target is reachable — and in a bear frame the previous high is, by the definition of a lower high, precisely the level price is failing to reach. You will experience that as “losing my edge”. It is not. It is the same action being priced by a different phase.

The classical depths are not constants of nature, and the R:R table is a conditional calculation: if a retracement stops at 0.382 rather than 0.618, then the arithmetic is as above. Note too that the measured medians (0.76 to 0.88) are deeper than either classical number — a deep retracement pays more when it holds, and the stop-hit column is the same measurement telling you how often it did not. What does not depend on any of these numbers is the shape of the conclusion: shallower retracements pay worse, and they pay worse faster than proportionally, because the depth shrinks the reward and grows the risk at the same time.

There is a second asymmetry, and this site has already documented it elsewhere so it only needs a pointer: losses and gains are not symmetric in account terms either. A 50% decline requires a 100% advance to undo, which is why a bear phase costs more than the mirror-image arithmetic suggests. The drawdown guide has the full table.

Which of your tools breaks in which phase?

Every one of them breaks somewhere, and the failures are predictable enough to tabulate. This is the practical payoff of knowing the phase: not that it tells you what to buy, but that it tells you which instrument on your dashboard is currently lying.

Read down the "fails when" column and the pattern is stark: trend tools manufacture signals in a range, and range tools sell you the top of a trend. Neither malfunction announces itself. Both produce perfectly ordinary-looking outputs; the moving averages really did cross, RSI really did print 78. The tool is not broken. It is being asked a question its assumptions do not cover.

Which is why the phase question is not preliminary work you do before the real analysis. It is the analysis. Lesson 20 puts a number on the cost of getting it wrong — running the range playbook in a trend and vice versa — and the number is large enough that it dominates almost everything else you might improve.

How do you tell a phase has actually changed?

By checking the frames rather than the candles, and by accepting a delay you cannot negotiate away. There is a workable sequence, and the honest version of it includes an outcome most methods leave out: "not yet".

1. Fix your set of frames and write it down. Three is a reasonable working number — a fast frame you enter on, a middle frame you plan on, and a slow frame that grants or withholds permission. Spacing them about four to seven times apart keeps them from telling you the same thing twice. Write the three down before you look at price, because a set chosen after the fact will always agree with whatever you already believe.

2. Label each frame independently. On each one, ask only the two questions from Lesson 15: is the last high above the previous high, and is the last low above the previous low? Both yes is up, both no is down, anything else is neither. Do not let your reading of one frame contaminate the next; that is the entire value of using several.

3. Read the score, not the chart. Three ups is a bull phase. Three downs is a bear phase. Anything else is the sideways condition, and the correct output is "no phase yet" rather than a guess about which way it will resolve. This is the step people skip, because "I don't know" feels like a failure of analysis rather than a result of it.

4. Expect the change to start at the fast end and take a known minimum. When the fast frame flips first you have not seen a reversal, you have seen the first frame of one — and the earliest the slowest frame can confirm anything is the floor from the table above. If your slow frame is the weekly, nothing new can confirm on it for 14 days, and on BTC the full flip never took less than 32 — so put a mark a month out and stop asking until then.

5. Check volume at the boundary. The course is specific here: consolidation and accumulation run on low volume, and a break out of one that comes without a clear rise in traded volume is very likely a false break. Lesson 14 shows how to measure "a clear rise" rather than eyeball it.

When is this advice wrong?

In four places, and the first one undermines the headline number badly enough that it has to come first.

The coin-flip null is a two-state model of a three-state reading, so 2¹⁻ⁿ is neither a floor nor a forecast. A frame can read up, down or neither, and the coin has no “neither”. An earlier version of this page argued that nesting must make frames agree more often than independent coins; measured on BTC they agreed less often (6.6% against 25%), and almost exactly as often as independent frames with the measured mix of three states would (6.7%). Use the coin-flip table for the one thing it is good for — your choice of n drives how often you may act, geometrically — and the measured table for everything else. And treat the measured table as what it is: one pair, one swing setting, two years. Change the setting and every number moves.

The phase labels are lagging by construction, so acting only on a confirmed phase means acting late. Everything above is built on structure, and structure needs swings, and swings need candles to close. If you insist on unanimity across slow frames you will be right about the phase and late to every move within it. That is a legitimate trade-off and some approaches accept it deliberately; it is not a free lunch, and anybody presenting it as one is not counting the cost.

"Bear phases have shallow bounces" is what the course teaches, and our own data did not confirm it. On BTC at a 2-bar setting, bounces in a bear-labelled frame retraced a median 0.87 of the leg against 0.76 for pullbacks in a bull-labelled one. The 0.382-versus-0.618 comparison is therefore a conditional calculation dressed in classical numbers: if the depths differ that way, the arithmetic follows. Two markets in the same nominal phase can pay very differently, and the same market pays differently at a different swing setting. Measure your own retracements over a few dozen swings before you let either table set your expectations.

On a long enough horizon, phase is a story about your holding period rather than about the market. A position held for months in a bull phase passes through many bear phases on the 4-hour chart, and none of them matter to it. Whenever you say "we are in a bear market", the sentence only carries information once you have named the frame — and if the frame is much faster or much slower than your holding period, the label is true and irrelevant at the same time. See Lesson 4 for how the holding period ends up choosing your frames for you.

Common mistakes

Frequently asked questions

What is the difference between a bear market and a correction? Depth and structure, and the structural test is the more useful of the two. The popular definition is a threshold — a decline of more than 20% from the high is called a bear market, less than that a correction — but a threshold tells you nothing until it has already been crossed. The structural test asks instead whether the sequence has changed: in a correction inside a bull phase, price falls but the low still holds above the previous low, so the higher-high-and-higher-low sequence survives. In a bear phase it does not. That test can be applied while the move is happening rather than after, and it works identically on a 4-hour chart and a monthly one, which the percentage rule does not.

How long does a sideways market last? Longer than most people expect, and there is a hard floor plus a measured range. The floor: the slowest frame in your set cannot confirm any new swing until enough of its own candles have closed — with a 2-bar setting that is 8 hours on the 4-hour chart, 2 days on the daily and 14 days on the weekly. The measurement: on BTCUSDT between September 2024 and September 2026, the five complete flips between all-frames-up and all-frames-down on the 4-hour, daily and weekly had 32, 47, 58, 70 and 135 days of disagreement in between. The upper end is genuinely unpredictable, and any specific figure you see quoted depends on a definition and a frame set that the person quoting it has usually not stated.

Can different timeframes be in different phases at the same time? Yes, and it is the normal state of affairs rather than an anomaly. The 4-hour chart can be in a textbook bear phase while the weekly is still making higher lows, and neither reading is wrong — they describe different groups of participants with different horizons. In the course's framing, that disagreement is not a puzzle to resolve; it is the sideways condition. The practical consequence is that "what phase are we in?" is not answerable until you say which frames you mean, and the useful answer is a score across your chosen set rather than a single word.

Should I trade at all during a sideways phase? That is a question about which tools you own, not about whether the market is tradeable. A range is perfectly tradeable with range tools — the boundaries are defined, so the risk is defined — and it is hostile to trend tools, which will generate crossings and breakouts that go nowhere. The mistake is not trading a sideways market; it is trading it with the instruments you were using the week before. If your method needs a trend and the score says no phase yet, the honest options are to stand aside or to switch to a method built for boundaries, and the second one requires having practised it beforehand.

How do I know a new bull phase has started rather than a bounce? By the same two questions applied in order, and by refusing to answer early. A bounce inside a bear phase produces a higher high or a higher low, but not both, and not repeatedly. A new bull phase produces both, and then does it again. The arithmetic in Lesson 15 is worth revisiting here: one higher-high-and-higher-low pair is a 25% event on a neutral yardstick, so a single pair is the base rate rather than evidence. Combine that with the frame score — a bounce usually flips only the fastest frame, while a phase change eventually flips all of them — and the useful test becomes: has the score changed, or has one chart changed?

How these numbers were produced. Three pieces of arithmetic and one measurement. First, P(all n frames agree) = 2¹⁻ⁿ, from treating each frame’s direction as an independent fair coin — a deliberately neutral null, not a model of markets. Second, the hard floor on the disagreement window follows from the swing-confirmation identity in Lesson 17: a swing point at setting n cannot be confirmed for n further bars, so 2 bars on a weekly chart is 14 days; the earlier version of this page multiplied that into “about five candles, 35 days”, which was an estimate presented as a floor, and has been corrected. Third, retracement pricing uses R:R = f/(1−f) and break-even = 1−f, the identity derived in Lesson 33, applied to the classical 0.382 and 0.618 depths as a conditional comparison. The measurement: Binance spot BTCUSDT klines at 4-hour, daily and weekly resolution, closed candles only, snapshot taken 19 Sep 2026; swing points are 2-bar (a high above the two highs either side, likewise for lows), confirmed at the close of the second bar after them; each frame is labelled U, D or N by the two Lesson 15 questions on its last two confirmed highs and lows; the score is sampled daily at 00:00 UTC from 19 Sep 2024 to 18 Sep 2026 (730 days); a “flip” is a run of UUU followed, after mixed days, by a run of DDD or the reverse; retracement f is measured between alternating confirmed swings and classed by the same frame’s label when the leg’s end was confirmed. Script and frozen data: do-so-bai-16.py. All R figures are before fees, funding and slippage. The definitions of the three phases, the observation that structure on a higher timeframe is more durable, the description of bear-phase bounces as short and early-ending, the statement that a trend requires the timeframes to agree, and the note that consolidation runs on low volume, all follow the slide course this site learned them from; where our measurement disagreed with the course (bounce depth) the page says so. Every figure here is for one pair, one swing setting and one two-year window, and will differ elsewhere. Sources: our own arithmetic and measurement, stated inline. Published 2 Sep 2026. Updated 19 Sep 2026: cover and all three figures redrawn by code (the cover from real candles, the bar chart to scale, the pullback figure to the price scale, the flip diagram with the corrected floor); two-year BTC measurement added for frame agreement, flip windows and retracement depth; the “35-day floor” and the “markets beat the null” claims corrected; end-of-lesson quiz moved back to the end of the lesson.