Swing Exhaustion Fade | Learned Odds + Liquidity Sweeps
Swing Exhaustion Fade (SEF) is a mean-reversion strategy that fades a swing once its extreme is probably in. Instead of fixed rules like “RSI above 70”, it learns from the chart you load how often swings of a given size and pullback depth had already peaked. It trades only when those learned odds are strong.
How it works
The odds table contains only swings that finished before the current bar, so the learning step does not peek at future data.
Features
Tips
Warnings
How it works
- Swing engine. A swing ends when price pulls back k × ATR from its extreme. The default is 2.5 ATR. Swings are measured in ATR, so the logic adapts to volatility and works across markets and timeframes.
- Learned odds. Each bar of a live swing is filed by direction, how far price has retraced, and the swing’s size. When a swing confirms, those bars are labelled “was the extreme already in?”. Each cell’s hit rate is shrunk toward its neighbours and then reduced to a Wilson lower bound, a conservative estimate that penalises small samples. The strategy only acts when that bound clears your threshold (60% by default).
- Liquidity sweeps. Estimated liquidation clusters are placed above swing highs and below swing lows at 1%, 2% and 4% distances. They are weighted by volume and decay with age. A fade needs a recent sweep through a cluster, such as a short-liquidation pool above a high, as a sign the move is exhausted.
- Day position (intraday only). Shorts must be near the day’s high and longs near the day’s low, measured in average daily ranges.
- Entry. The strategy fades the live swing once price has retraced 35–80% of k, the swing is at least 2 ATR, and all filters pass. It takes one trade per swing.
- Exits. The stop sits beyond the swing extreme plus a buffer. The target is a fixed R multiple, or the nearest opposite liquidity cluster if it is 1–4R away. A time stop closes the trade after 40 bars.
The odds table contains only swings that finished before the current bar, so the learning step does not peek at future data.
Features
- Self-learning reversal odds built from the loaded chart’s own history
- ATR-based swing detection that adapts to volatility
- Wilson lower-bound filter that penalises thin samples
- Estimated liquidity clusters with volume weighting, merging, decay and sweep detection
- Optional day-extreme filter measured in ADR
- Two sizing modes: % of equity (default 40%) or risk per trade, with a max notional cap
- Optional target at the nearest opposite liquidity cluster
- Warm-up gate: no trades until 20 swings per direction have been learned
- Aurora band around an EMA, with width following volatility and colour following price’s side of the EMA
- Zigzag legs with a dashed live swing
- Faint liquidity lines near price
- Red and green risk and reward zones while in a trade
- Dashboard with swing size, odds meter, win rate, profit factor, net profit and max drawdown
- Neon and Classic themes, and every visual can be switched off
Tips
- Start on intraday charts. The day filter works only there. On Daily and above it switches off automatically.
- Let the warm-up finish. The dashboard shows “Warming x/20” until each direction has 20 swings, so you need enough history loaded. On short histories, lower the warm-up count or use a lower timeframe.
- Tune swing size first. Lower k (1.5–2.0) gives smaller, more frequent swings. Higher k (3.0–4.0) gives fewer, larger ones. Re-check the other inputs after changing it.
- Raise the odds threshold for fewer, pickier trades. Lower it for more trades and more noise.
- Test each filter. Turn the sweep and day filters off one at a time to see what each adds on your market.
- Judge it on trade count. A strong result over 15 trades means little. Look for at least 100 trades on the loaded history.
- Size matters more than it looks. Equity-based sizing at 40% scales wins and losses equally. If max drawdown climbs past about 20%, reduce it or switch to “Risk per trade”.
- Set realistic costs. Change commission and slippage to match your market and broker.
- Winners vs win rate. Raising the R target or disabling the cluster target lets winners run further, usually at the cost of a lower win rate.
Warnings
- This is not financial advice, and past results do not predict future results. Backtests are hypothetical and can be overfit.
- The default sizing is aggressive. 40% of equity per trade, up to 2× notional, can produce large drawdowns. Reduce it before drawing any conclusion.
- The liquidation clusters are estimates. They come from swing structure and fixed percentage tiers, not from exchange liquidation or order-book data.
- The odds are in-sample. They are learned from the history loaded on your chart, so results change with how much history you load and which symbol and timeframe you use.
- Fills are approximate. The strategy evaluates on bar close, and stops and targets are modelled from OHLC data. Real fills, gaps, spreads and slippage will differ.
- Fading reversals is risky. A strong trend can run through the stop repeatedly, and a mean-reversion strategy can have long losing streaks.
- The live swing leg moves. The dashed leg updates as the current swing extends. Entries are evaluated at bar close, but treat the dashed leg as provisional.
- Test on a demo account or paper trading first before risking real capital.
免責聲明
這些資訊和出版物並非旨在提供，也不構成TradingView提供或認可的任何形式的財務、投資、交易或其他類型的建議或推薦。請閱讀使用條款以了解更多資訊。
免責聲明
這些資訊和出版物並非旨在提供，也不構成TradingView提供或認可的任何形式的財務、投資、交易或其他類型的建議或推薦。請閱讀使用條款以了解更多資訊。