Turtle Soup is the original failed-breakout fade. Linda Raschke and Laurence Connors published it in Street Smarts (1995) and aimed the name squarely at the Turtles — the famous trend-following experiment whose system bought every 20-day breakout. Their observation: most of those breakouts fail, so take the other side. The classic recipe runs on daily bars: price makes a new 20-day low while the previous 20-day low sits several sessions back, the break fails back through the old level, and you buy the failure with a tight stop beyond the extreme.
Thirty years later the same idea trades under a different name — the liquidity sweep — and mostly intraday. So we tested an honest intraday adaptation: on 15-minute BTC and ETH candles during New York hours (09:30–16:00), fade any wick through a deeper, older confirmed swing (fractal lookback 10, versus the lookback-3 micro-swings of our generic sweep test) that closes back inside, with a structure stop, a 2R target, and flat by the close. Twelve months of real Binance data, 0.05% commission per side, 1% risk per trade — the same deterministic engine that powers Secuora’s AI backtester. Results below.
Verified Result
No edge: net negative after costs across 2 markets.
| Market | TF | Trades | Win | PF | Max DD | Net |
|---|---|---|---|---|---|---|
| BTC | 15m | 777 | 21.5% | 0.22 | 99.8% | -99.8% |
| ETH | 15m | 823 | 20.5% | 0.23 | 99.9% | -99.9% |
How the SVS 20 breaks down ▾
12 months of real 1-minute data, fees on (0.05%/side), $10k start, 1% risk. How the score works →
The exact rules we tested
- Trade only between 09:30 and 16:00 New York time (DST-correct), both directions, on 15-minute candles.
- Swings are confirmed fractals with lookback 10 — a swing exists only after 10 later candles close, no look-ahead. The deeper lookback targets older, more obvious levels than a lookback-3 micro-swing.
- Sweep: the current candle’s wick trades through the most recent confirmed lookback-10 swing, and the candle closes back inside.
- Enter at that candle’s close, against the breakout (fade the failed move).
- Stop beyond the sweep extreme via the last opposing lookback-10 swing; target 2R; any open trade closes at 16:00 NY.
- Risk 1% of equity per trade; 0.05% commission per side; 10× max notional leverage.
Results
Binance spot 1-minute klines (data-api.binance.vision), aggregated per strategy timeframe · starting balance $10,000 · risk 1%/trade · Commission 0.05% per side; no spread/slippage modeled (BTC/ETH spot spreads are sub-basis-point); position size capped at 10× notional leverage. Generated 2026-06-12 by the Secuora ai-strategy deterministic runner (same engine as the in-app AI backtester).
| Month | Trades | Win rate | Net P&L |
|---|---|---|---|
| 2025-06 | 66 | 23% | −$4,244 |
| 2025-07 | 68 | 25% | −$2,457 |
| 2025-08 | 67 | 12% | −$1,683 |
| 2025-09 | 71 | 20% | −$763 |
| 2025-10 | 71 | 21% | −$381 |
| 2025-11 | 57 | 28% | −$144 |
| 2025-12 | 83 | 19% | −$172 |
| 2026-01 | 70 | 24% | −$58 |
| 2026-02 | 38 | 21% | −$28 |
| 2026-03 | 78 | 23% | −$32 |
| 2026-04 | 64 | 22% | −$16 |
| 2026-05 | 44 | 20% | −$7 |
| Month | Trades | Win rate | Net P&L |
|---|---|---|---|
| 2025-06 | 56 | 25% | −$3,283 |
| 2025-07 | 75 | 21% | −$3,247 |
| 2025-08 | 58 | 16% | −$1,469 |
| 2025-09 | 92 | 11% | −$1,351 |
| 2025-10 | 65 | 15% | −$307 |
| 2025-11 | 49 | 24% | −$107 |
| 2025-12 | 85 | 16% | −$132 |
| 2026-01 | 74 | 23% | −$45 |
| 2026-02 | 55 | 25% | −$17 |
| 2026-03 | 80 | 21% | −$20 |
| 2026-04 | 63 | 32% | −$7 |
| 2026-05 | 71 | 23% | −$6 |
Assumptions (how loose terms were pinned down)
- Intraday adaptation of Connors/Raschke Turtle Soup: the original fades a 20-DAY breakout; here a 15m wick through a deeper (lookback-10) confirmed swing that closes back inside, during NY hours
- Stop beyond the sweep extreme via the last opposing lookback-10 swing; 2R target; flat at 16:00 NY
From 20-day extremes to a 15-minute sweep: what the adaptation changes
The 1995 original is slow by construction. On daily bars, a new 20-day low whose predecessor sits several sessions back happens a handful of times a year on any one market; Connors and Raschke traded the snap-back with a stop just beyond the extreme and were typically out within days. Small sample, long holds, commissions nearly irrelevant to the math. That version cannot be compared one-to-one with anything intraday — so we say plainly that ours is an adaptation, not a replication.
Translated to 15-minute candles, every dial moves: a lookback-10 confirmed swing plays the role of the 20-day extreme (deeper and rarer than the lookback-3 swings our generic liquidity-sweep test fades), the structure stop replaces "pennies below the low", and flat-at-16:00 replaces the multi-day exit. The biggest change is the cost profile — many small trades instead of a few big ones means fees compound into the outcome. Read the table above accordingly: it is the unfiltered intraday baseline for the idea, the floor a filtered, discretionary version has to beat — not a verdict on the daily original.
How to backtest Turtle Soup on Secuora
Two ways: replay it candle by candle for screen time, or hand the rules to the AI backtester — the liquidity-sweep detector behind this research is a built-in primitive.
- Open the free replay demo at /backtest/demo (no sign-up) and load a crypto pair on the 15-minute timeframe.
- Step the replay to the 09:30 New York open. (Sign up free for the replay terminal’s one-click, DST-correct session-open skips — New York, London, Tokyo.)
- Mark the older swing highs and lows with the drawing tools, then step forward bar by bar and watch for a wick through one that closes back inside.
- To fade it with a simulated order — stop beyond the sweep extreme, take-profit at 2R — sign up free and run the session in the replay terminal; from the session summary you can push the trade to your journal in one click, with a screenshot and your reasoning.
- For the statistical answer, describe the exact rules in plain English at /backtest/ai — sweep lookback, session, stop, target — and the same engine that produced this page re-runs it on your parameters.
Methodology, in one paragraph
Data: Binance spot 1-minute klines (data-api.binance.vision), aggregated per strategy timeframe, June 1, 2025 – June 1, 2026 (12 months). Execution: Secuora’s deterministic strategy runner (the same engine behind the in-app AI backtester) — single position at a time, entries at the close of the signal candle, commission 0.05% per side; no spread/slippage modeled (btc/eth spot spreads are sub-basis-point); position size capped at 10× notional leverage, starting balance $10,000, 1% risk per trade. Swings are confirmed fractals with no look-ahead. These are mechanical results: no discretion, every signal taken. Past performance does not predict future results; this is research, not financial advice.
Frequently asked questions
What is the Turtle Soup trading strategy?
A failed-breakout fade from Linda Raschke and Laurence Connors’ 1995 book Street Smarts, named as a jab at the Turtles, the famous trend-followers who bought 20-day breakouts. When price makes a new 20-day low (or high) and immediately fails back through the old level, Turtle Soup takes the other side with a tight stop beyond the extreme.
Is Turtle Soup the same as a liquidity sweep?
Same family, different vintage. "Liquidity sweep" is the generic modern term — a wick through a swing that closes back inside; Turtle Soup is the classic named version that specifically fades a deeper, older breakout level (originally the 20-day extreme). Our test reflects that distinction: it sweeps a lookback-10 swing on 15-minute candles, versus the lookback-3 swing on our liquidity-sweep-reversal page.
What win rate does Turtle Soup have?
For the intraday adaptation we ran — 15-minute candles, NY hours, lookback-10 swing, structure stop, 2R target, costs on — the win rate and profit factor are whatever the results table on this page shows; we publish the engine’s output unedited. The 1995 daily-bars original is a different animal — far fewer signals and multi-day holds — and we won’t quote numbers for it that we haven’t generated ourselves.
Does Turtle Soup work on crypto?
The raw pattern — a failed breakout of an obvious older level — occurs constantly on BTC and ETH. Whether the mechanical version carries positive expectancy after fees is exactly what the table above answers for the unfiltered reading; treat it as the baseline, then test your filters (session, trend bias, displacement) against it on Secuora.