ATR Breakout SHORT

Overview

ATR Breakout SHORT fires when the Average True Range (14-period) spikes above its moving average threshold — signalling a volatility expansion — and price is moving downward. Unlike BB Squeeze which waits for compression first, this strategy catches the early momentum of a directional volatility burst.

Re-verified 2026-05-04 via full sweep (1,680 combinations × 50 coins, IS/OOS split). Prior shelved status was based on comparison with BB Squeeze (now killed) using ATR-multiple configs. Fixed SL/TP format reveals a clear short-side edge.

How It Works

1. ATR spike — 14-period ATR crosses above 1.5× its 20-period moving average
2. Directional filter — price below its 20-period MA confirms downward bias
3. Volume confirmation — volume ≥ 1.5× 20-period average
4. Entry — SHORT on bar close
5. Exit — TP 10% / SL 3% / 48-bar (2-day) timeout
6. Risk — asymmetric: 1:3.3 R:R (3% risk, 10% reward)

Why It Works (Thesis)

ATR spikes in crypto typically mean one of two things: a genuine directional move, or a wick-spike that reverts. On a 1H timeframe with the 3% SL, you get stopped out quickly when it’s a wick. When it’s a real downward expansion — an altcoin breaks support, a leverage cascade starts — the 10% TP captures the bulk of the move.

The SHORT edge (not LONG) is structural: in crypto, downward volatility expansions tend to be sharper and less frequently faded by new buyers in the first 2 hours. Upward expansions attract FOMO buyers that compress the follow-through.

Results (2-year backtest, IS/OOS split, measured 2026-05-04)

Metric	IS (May24–May25)	OOS (May25–May26)	Combined
Total trades	1,513	1,326	2,839
Win rate	~38%	~34%	35.9%
Profit factor	1.57	1.28	1.42
Coins profitable	48/50	36/50	—

Low win rate is by design — the 1:3.3 R:R means each win covers 3+ losses.

Caveats

• OOS PF (1.28) is lower than IS (1.57) — some IS outperformance did not carry forward. This is the weakest of the three sweep survivors.
• Tight 3% SL generates frequent stop-outs. Requires accepting 64% loss rate.
• Not live-tracked. Backtest only.
• All LONG configurations remain unprofitable on OOS data.