Do classic trading setups work on crypto? We graded ten, stage by stage

What a stage machine is

Most setups traders hunt are not one condition but a sequence. A Wyckoff spring is a dip under a range, then a quiet retest, then a breakout. A squeeze is compression, then a fire, then a run. We call that a stage machine: each stage is a screen that only lights once the stage before it has, so a market moves from stage 1 to stage 2 to stage 3 as the setup plays out. We built ten of them, one desk each, with every stage a plain rule that is either true or not.

How we graded them

Each desk promises a move: a flush, a breakout that keeps running, a bounce off a low, a trip across the value area. We asked one question of every stage: when it lit, how often did that move follow, compared with any random moment? A stage that lights before the move far more often than a random moment is a useful early warning. We graded every stage over 200 Hyperliquid crypto markets (120 for the order-book desk, whose data is heavier), on two separate windows of about 10 + 10 days ending 22 October 2026.

• Two chance checks on every stage: the same lit moments slid onto other markets at the same times, and onto other days at the same time of day. A stage only counts as real when it beats both — otherwise it is the whole market moving together, or the time of day.
• Direction measured separately: did trading the stage in its own direction make money after a 10 bp round trip, on both windows?
• 41 stages and variations graded in all, every one on both windows — the failures are reported beside the winners.

The ten desks, ranked as early warnings

1. 01

   Crowded → Trapped → Flush

   Strong

   Traders pile into one side (funding in the market’s top 10%, open interest up 5%), price breaks against them on a big sweep, then they are forced out.

   When it lit, the move followed about 11× as often as from a random moment.

   22–27%

   2.1–2.6%

   Stage read: Longs trapped (15-minute bars). The move: a 3% drop inside an hour with open interest down 2% (longs forced out), within the next 4 hours.

   The mirror works too: when Shorts trapped lit, a squeeze followed 5–6.7% of the time against 0.7–0.8%. Which way price went after the flush changed between the two windows.

   Open the Crowded → Trapped → Flush desk
2. 02

   Ignition → Follow-through → Holding

   Strong

   A huge buy sweep on a bar three times its usual size, moving on its own rather than with Bitcoin; then new positions follow; then the spike holds.

   When it lit, the move followed about 7× as often as from a random moment.

   14–16%

   1.7–2.4%

   Stage read: Spike ignition (15-minute bars). The move: a new 24-hour high that runs 3% further, within the next 3 hours.

   Requiring the spike to move apart from Bitcoin is what makes it: without that condition the same spike read 5×. Bought on the spike and held 15h, it made money after costs on one window and lost on the other, so it is not a buy signal.

   Open the Ignition → Follow-through → Holding desk
3. 03

   Absorption → Initiative

   Strong

   Sellers hit a low and price won’t go lower; the sellers run out while trapped longs exit; then buyers take over on a big buy sweep.

   When it lit, the move followed about 7× as often as from a random moment.

   5–8.8%

   0.8–1.1%

   Stage read: Buyers take over (15-minute bars). The move: a bounce off a 12-hour low that breaks out and runs 2% further, within the next 2 hours.

   Each stage sharpens the one before: stage 1 2.0×, stage 2 5×, stage 3 7×. On hourly bars the last stage read 5×.

   Open the Absorption → Initiative desk
4. 04

   Held Up → BTC Steadies → Leader Breaks

   Useful

   A market holds its price while Bitcoin dumps, Bitcoin steadies, then the market that held up breaks out first.

   When it lit, the move followed about 3× as often as from a random moment.

   14–32%

   6.8–7.5%

   Stage read: Leader breaks out (15-minute bars). The move: a breakout above the last 6 hours’ high that runs 2% further, within the next 3 hours.

   Holding up in a Bitcoin dump on its own read 1.2×. The last stage is itself a breakout, so part of its number is the move it is graded against; Bitcoin dumps are rare, so the desk is quiet most days.

   Open the Held Up → BTC Steadies → Leader Breaks desk
5. 05

   Squeeze → Fire → Run

   Useful

   Volatility compresses for six bars (Bollinger Bands inside the Keltner Channel), price breaks the 5-hour high, then the break holds.

   When it lit, the move followed about 3× as often as from a random moment.

   22–23%

   6.8–7.5%

   Stage read: Squeeze fired (15-minute bars). The move: a breakout above the last 6 hours’ high that runs 2% further, within the next 3 hours.

   The squeeze on its own read 1.2× and is on in 8.9–10% of all bars. Requiring buying and new open interest on the break did not help (3×).

   Open the Squeeze → Fire → Run desk
6. 06

   Sweep → Change of Character → Gap

   Useful

   The Smart Money Concepts sequence: stops under a swing low are taken, a big candle breaks the last swing high and leaves a gap, then price comes back into the gap.

   When it lit, the move followed about 2.7× as often as from a random moment.

   17–22%

   6.8–7.5%

   Stage read: Character changed (15-minute bars). The move: a breakout above the last 6 hours’ high that runs 2% further, within the next 3 hours.

   The change of character carries the read. The stop sweep on its own read 0.8×, and coming back into the gap 1.7×.

   Open the Sweep → Change of Character → Gap desk
7. 07

   Holy Grail Pullback

   Weak

   Linda Raschke’s Holy Grail: a fresh 24-hour high, the first pullback to the 20 EMA, then a break of the pullback’s high after traders took profit in the pullback.

   When it lit, the move followed about 1.9× as often as from a random moment.

   13–14%

   6.8–7.5%

   Stage read: Trend resumes (15-minute bars). The move: a breakout above the last 6 hours’ high that runs 2% further, within the next 3 hours.

   The pullback adds nothing: the fresh 24-hour high on its own read 2.9×, the textbook pullback to the 20 EMA 0.6× and the break of its high 1.9× — with or without falling open interest (1.7× without).

   Open the Holy Grail Pullback desk
8. 08

   Wyckoff Spring → Test → Strength

   Weak

   A dip under a trading range that snaps back (the spring), a quiet retest of that low, then a breakout on heavy volume.

   When it lit, the move followed about 2.6× as often as from a random moment.

   16–20%

   6.8–7.5%

   Stage read: SOS breakout (15-minute bars). The move: a breakout above the last 6 hours’ high that runs 2% further, within the next 3 hours.

   The spring read 0.5× and the test 0.7×: both came before rallies less often than from a random moment. Only the breakout scores, and it is partly the move itself.

   Open the Wyckoff Spring → Test → Strength desk
9. 09

   Fake Wall → Pulled → Lift

   Weak

   A big sell wall held by one or two orders pins price, the wall disappears without being bought, then price lifts through.

   When it lit, the move followed about 1.5× as often as from a random moment.

   4.1–7.2%

   3.5–4.2%

   Stage read: Lifts through (5-minute bars). The move: a breakout above the last 2 hours’ high that runs 2% further within 80 minutes, inside the next hour.

   The wall read 0.7× and its disappearance 0.6×: on 5-minute bars a wall that was cancelled looks the same as one that was traded into, so the setup can’t be told apart from noise yet.

   Open the Fake Wall → Pulled → Lift desk
10. 10

    80% Rule

    Fails

    The Market Profile rule: price falls under the last 24 hours’ value area, comes back in and stays two bars, and is said to travel to the top 80% of the time.

    When it lit, the move followed less often than from a random moment.

    21–23%

    27–32%

    Stage read: Accepted in value (15-minute bars). The move: price reaches the top of the last day’s value area, within the next 4 hours.

    Given a whole 12 hours, it reached the top 46–54% of the time against 53–61% from a random moment — still a little less often.

    Open the 80% Rule desk

The strongest early warnings

Crowded → Trapped → Flush, Ignition → Follow-through → Holding, Absorption → Initiative are the three to watch. Crowded longs that get trapped are the clearest read in the study: the setup is rare, but when it lit, the flush followed 22–27% of the time against 2.1–2.6%. A spike ignition — a buy sweep in the market’s top 3% on a bar three times its usual range, moving on its own — came before a fresh 24-hour high that ran 3% further 14–16% of the time, against 1.7–2.4%. And Absorption → Initiative is the one desk where every stage sharpens the last, on 15-minute bars and on hourly bars alike.

What didn’t work

Some of the most taught setups did not survive the grade on crypto.

• The 80% rule. After price came back into the value area and stayed, it reached the top within 4 hours 21–23% of the time — against 27–32% from a random moment. Given 12 hours, 46–54% against 53–61%.
• The Wyckoff spring and its test. They came before a rally less often than from a random moment (0.5× for the spring, 0.5–0.8× for the test).
• The textbook Holy Grail pullback. 0.6× — its pullbacks came before rallies less often than from a random moment. The rally came back only once price broke the pullback’s high, and even then less often than straight after the fresh high itself.
• A stop sweep on its own. Taking the stops under a swing low read 0.8×; the change of character after it is what carries the Smart Money Concepts desk.
• Order-book walls on 5-minute bars. A thin-held wall read 0.7× and its disappearance 0.6× — our book data can’t yet tell a pulled wall from a filled one.

What we’re building next

Building these ten desks showed us exactly where the tools fall short for setups like these. What we’re building, so the next round of this study is sharper and so you can build your own:

• Grade a setup on the whole market, and keep a desk to the markets it was graded on — so any rare setup gets a real verdict, and a crypto desk shows only crypto.
• Grade a setup as a trade — enter on the last stage, stop at a level an earlier stage set, a target and a time limit, after costs, against a random market at the same moment.
• See which stage every market has reached — rank a desk by the furthest stage reached and how recently, colour a stage without calling it a buy or a sell, and alert only on the last stage.
• Build stage machines faster — grade a whole chain in one step before saving it, write both sides of a setup at once, and use windows like “since stage 1” instead of a fixed number of bars.
• A session clock — time of day and session in formulas, for opening-range breakouts, prior-day levels and the 80% rule as it was written.
• Tell a pulled order from a filled one in the order book, so spoofing and pulled-wall setups can be tested at all.
• Fewer sharp edges — a warning when two screens share a name, and lists that show only the markets actually firing.