Liquidity sweep and confirmed rejection

Research status: untested hypothesis. No backtest, paper or live results exist for this playbook. Every number below is an untested candidate parameter to be studied, not a recommendation and not a proven edge. Hypothetical examples are not Josh’s trades.

The idea, and what it does not claim

An obvious prior high or low is a place where many traders may have stops or breakout orders resting. Price sometimes pushes just past that level and then snaps back. This playbook asks one narrow question: when a completed candle closes decisively back inside after running the level, does price tend to travel back toward the middle of the prior range far enough to justify a predefined risk?

The “liquidity” here is a hypothesis inferred from visible chart structure. A sweep candle does not prove that stops were actually there, that a bank or “smart money” engineered the move, or that anyone intended anything. Sometimes a wick past a level is just the first leg of a continuation. That is why the rule waits for a completed close and has a hard stop.

Setup: fixed before the day starts

All values below are untested candidate parameters.

Item	Definition
Instrument	One BTC/USD spot pair on one named venue, chosen before collecting data. Use that venue’s prices, not a blended chart.
Bars	5-minute candles, aligned to UTC.
Reference day	The previous completed UTC day (00:00–24:00 UTC). Record its high H, low D, range W = H − D and midpoint M = (H + D) / 2. Skip if W ≤ 0.
Sweep threshold	X = max(3 ticks, 0.05 × W)
Stop buffer	b = max(2 ticks, 0.02 × W)
Level corroboration	A confirmed 15-minute swing high within 0.10 × W of H (or swing low within 0.10 × W of D). A swing counts only after the two bars that follow it have closed; never retroactively.

Why UTC? Crypto trades 24 hours a day, 7 days a week, so there’s no exchange “session” in the stock-market sense. This playbook uses the UTC calendar day as a consistent label for the reference range; it’s our convention, not an official market session. In November, 00:00 UTC is 4:00 PM Pacific (PST) on the previous calendar day. US stock sessions (9:30 AM–4:00 PM ET) and session VWAP are not used here. If you adapt the idea to stocks or futures, the session definition, gaps and VWAP behavior change and need their own test.

IF–THEN rules (long branch)

For a completed 5-minute candle with open O, high h, low l, close C and range r = h − l (skip if r ≤ 0):

1. IF l ≤ D − X (price ran the low by at least X)
2. AND C ≥ D + X/2 (it closed at least X/2 back inside)
3. AND C ≤ H, C > O (a bullish candle that stayed inside the range)
4. AND the lower wick, min(O, C) − l, is at least 0.5 × r
5. AND C is in the top third of the candle (C ≥ l + 2r/3)
6. AND every no-trade check below passes
7. THEN the simulated entry is the first executable price of the next bar. No buying while the sweep candle is still open, no limit orders waiting for a better price.

• Stop: s = l − b (below the sweep wick plus buffer), frozen at entry. Never widened.
• Target: the frozen midpoint M. One target. No partial exits and no runners.
• Time stop: if neither stop nor target is hit within 12 bars (one hour), exit at the next bar’s first executable price. Always exit before the UTC day ends.
• Reward/risk gate: after costs, the target must offer at least 2R, or there’s no trade (see the shared risk protocol).

Short branch. The high-sweep version mirrors these rules (h ≥ H + X, close back below H − X/2, bearish candle, upper wick ≥ 0.5r, close in the bottom third, stop above the wick). It stays excluded: you can’t short unowned spot, and a margin or perpetual version would need its own study of borrowing, funding, liquidation and fees.

Entry is on a close; exit is not

Entry waits for a completed candle. The stop does not. If price trades through the stop, the exit happens at the first executable price; you don’t wait for the candle to close. Your planned loss is 1R, but the filled loss can be larger because of gaps, fast markets and slippage. Waiting for “confirmation” on the way out is how a planned −1R turns into −2R.

When NOT to trade

Skip the setup when any of these is true. These filters come from the source playbook and aren’t loosened here.

• No corroborated swing level near H or D.
• The candle misses any of the quantitative rejection tests.
• This side already had its one attempt today (max 1 attempt per side, 2 per UTC day).
• Chop: the level was crossed in both directions on 3 or more of the last 12 completed 5-minute candles.
• Strong trend against you: for a low sweep, the last three 15-minute closes are all below their 20-bar average and that average fell by more than 0.05 × W over three bars.
• Both H and D were swept in the same candle.
• Any shared no-trade filter fails (spread, liquidity, data quality, scheduled CPI/FOMC window, daily loss limits).

Worked example (hypothetical numbers, not a real quote)

Reference day: H = 100, D = 90, so W = 10, M = 95. With a tick of 0.01: X = 0.50, b = 0.20.

Step	Value	Check
Sweep candle	O 90.10, h 90.50, l 89.40, C 90.40	Low 89.40 ≤ 89.50 (D − X) ✓
Close back inside	90.40 ≥ 90.25 (D + X/2) ✓	Bullish, inside range ✓
Wick and close	Range 1.10; lower wick 0.70 ≥ 0.55; close ≥ 90.13 ✓	Qualifies
Entry (next bar)	90.41
Stop	89.40 − 0.20 = 89.20	Frozen
Assumed costs	0.10 per unit each way (illustrative only; use your venue’s real fees and spreads)
Risk per unit L	(90.41 − 89.20) + 0.10 = 1.31
Net reward per unit G	(95 − 90.41) − 0.10 = 4.49	G / L ≈ 3.4R, passes the 2R gate
Size	Hypothetical $10,000 equity × 0.25% = $25 budget; 25 / 1.31 ≈ 19.08 units (about $1,725 notional)	Inside the 25% notional cap

Two hypothetical outcomes from the same entry:

• Target path: price trades up to 95 within the 12 bars and the limit fills. Net result ≈ +4.49 × 19.08 ≈ +$86, about +3.4R before any queue or partial-fill issues.
• Failure path: price breaks lower and trades through 89.20 in a fast move; the first executable exit is 89.05. Realized loss per unit ≈ (90.41 − 89.05) + 0.10 = 1.46, about −1.11R (≈ −$28), larger than the planned −1R. That’s the gap between a planned loss and a filled loss.

These outcomes are illustrations of the arithmetic. They say nothing about how often either path happens.

Ratigan Capital rules in this playbook

• Patience as a Position: no trade until the sweep candle has closed and passed every test. Watching price near the level isn’t a signal.
• Early Is Wrong: buying the first tick below D, then pointing to a later reclaim, is still a rule violation, even if it makes money.
• Never Average Down: no adds if the trade moves toward the stop. No second attempt on the same side today.
• Price Is the Only Truth: react to the completed candle and the stop, not to a story about who “engineered” the move. News, costs and liquidity still matter, which is why the filters exist.

Full rules and the psychology checklist: Core rules for every playbook.

Why it can fail

• A wick past a level can be a pause before continuation, not a reversal. Several contrarian attempts can stop out in a row.
• Spreads widen and liquidations happen around these moves; venue-specific spikes can print misleading candles.
• The 0.05 × W and 0.10 × W thresholds, the 12-bar time stop and the midpoint target are all choices. They may be poor ones. Testing may reject the whole idea.

Pre-trade checklist

1. Reference H, D, W, M, X and b written down before the UTC day starts?
2. Corroborating swing level confirmed?
3. Sweep candle completed and every quantitative test passed?
4. No-trade filters clear, attempt count available?
5. Stop, target, time stop, costs and size written; net reward ≥ 2R?
6. Can I accept the full loss, and a filled loss larger than planned?

Sources

These explain general concepts. The thresholds and combinations above are Ratigan Capital’s own candidate research definitions; no author endorsed them.

• StockCharts ChartSchool: Support and resistance
• StockCharts ChartSchool: The Wyckoff method, a tutorial