4 / 6 · 5 min
Liquidity sweep, knife, density bounce, impulse and range breakout
Five kinds about overshoot and burst: where a move deceives, where it exhausts, and where it is only being born. It is among these that the statistics most often finds a skew towards the bounce.
The kinds of the previous lesson looked WITH the move. These five are different: the liquidity sweep and the knife catch deception and overshoot, the density bounce — price meeting big money in the order book, the volume impulse and the range breakout — the birth of a move. In the statistics under the feed the first two have the «Bounce» column underlined: their nature is that the move does NOT continue. Understand that before the thresholds: the kind decides which way to look at all.
Thresholds, verbatim from the detector
- Liquidity sweep: the wick went beyond an untouched level by more than the tolerance, and the CLOSE came back. The stops behind the swing were filled, the move did not continue; the liquidity line disappears from the chart at that moment. The card's direction is AWAY from the poke.
- Knife: over 1–5 bars an impulse of three or more typical bars — followed immediately by a return of at least half. A blade on the chart; the direction is where price rolled back to.
- Density bounce (5m only): price came right up to a live order-book wall of $100K+ and closed away from it. The card shows the wall size and the exchange.
- Volume impulse: bar body ≥ 2 NATR, volume ≥ 4 medians. A move with money behind it — but note: it is the most numerous kind in the feed, and its statistics are modest.
- Range breakout: at least 12 bars inside a box no wider than two NATR — then an exit with a body of half NATR or more and volume of one and a half medians. The squeeze ended, the spring released.
A liquidity sweep in numbers — and why stop hunters love this kind
A coin at 0.5000, NATR 2% (typical bar 0.010, tolerance 0.0025). An untouched swing high at 0.5120: above it sit shorts' stops and stop-orders for the breakout. A bar wicks to 0.5155 (0.0035 beyond the level — more than the tolerance) and closes at 0.5080 — BELOW the level. What actually happened: the stops above the swing got filled — and no buyers showed up for the continuation. The card: «Liquidity sweep · down». Price now has no fuel above: whoever wanted to buy the breakout has already been filled at a worse price. Hence this kind's statistical skew towards the bounce.
Knife vs impulse: equally scary, opposite in meaning
Both kinds start with a violent move, but the impulse ENDS with it while the knife does not: a knife requires an immediate roll-back of at least half the move. The impulse says «money came in this direction», the knife — «this direction was just deceived». The practical consequence: chasing a knife in the blade's direction is the worst trade in the feed (you are buying the top of a panic), while the return after a knife is statistically one of the steadiest shapes. The detector tells them apart for you — do not mix them back together.
Trusting an order-book wall like concrete
The density bounce is the only kind tied not to candles but to the live book: a $300K wall at 0.4950 is real orders right now. The mistake is to treat «right now» as forever: a wall can be pulled a second before the touch, and the feed accounts for that after the fact — your order does not. The rule: a wall is an argument IN FAVOUR of the level next to it, not a support of its own. A stop behind a wall «because they won't break it» is a stop placed on someone else's promise.
In the feed, select «Liquidity sweep» and «Knife» on 15m and 1h and browse the decided cards (the «With outcome only» button). Count over ten cards of each kind: how often price went away from the poke after a sweep, how often the knife returned to the blade's origin. Compare your impression with the «Statistics» table — same numbers there, but over all findings of the week.
Feed: sweeps and knives with outcomesA wick went beyond an untouched high by more than the tolerance, and the bar CLOSED above the level too. What will the detector record?
A level breakout, not a sweep: a sweep requires the close to RETURN to the original side. The sweep and the breakout are a fork of the same moment, and only the bar's close decides it. That is exactly why both kinds are computed strictly at close: before it there is no event yet — only a wick that could still become anything.
Why does the volume impulse — the most «powerful»-looking event — have a follow-through share around one half?
Because an impulse is visible to everyone at once: by the time the impulse bar closes, the move has already covered two NATR, and buying its continuation means entering at a price where early participants start taking profit. The money was in the bar — but part of it has already left. That is the main lesson of the feed's statistics: the spectacle of an event and its predictive power are different things, and it is the second one you must measure.