BB Squeeze SHORT

Overview

BB Squeeze SHORT is a volatility expansion strategy that detects when Bollinger Band width contracts below its own moving average (the “squeeze”), then enters a short position when the bands expand. The core idea: after extreme compression, markets tend to make large directional moves.

Status: KILLED (2026-06-28). This preset was previously presented as verified on the strength of the 2026-04 backtest below. A fresh out-of-sample re-test (product engine) failed it on three adversarial axes in choppy / range-bound (CHOP) regimes: the headline PF was bear-beta plus regime luck, not a standalone short edge. The backtest numbers remain reproducible and are kept here for the record, but we no longer claim a live directional edge — this is now a documented failure, not a recommendation.

How It Works

1. Detect Squeeze — BB width contracts below 0.8x its own 10-period moving average
2. Wait for Expansion — BB width must increase >= 10% from squeeze state
3. Volume Confirmation — Volume must be >= 2.0x the 10-period average
4. Enter Short — Only during allowed hours (excludes 7 low-edge UTC hours)
5. Risk Management — SL 10%, TP 8%, max hold 48 hours

Key Parameters

Parameter	Value	Rationale
Stop Loss	10%	C5 OOS verified, 54% reduction in SL hit rate
Take Profit	8%	Optimal R:R per MFE analysis, +24.3% PnL improvement
Volume Filter	2.0x	Filters fake signals, verified across 2+ years
Time Filter	7 hours blocked	UTC [2,3,10,20,21,22,23] — statistically negative hours
Max Hold	48 hours	Beyond this, edge degrades to noise
BB Expansion	>= 10%	Confirms genuine volatility expansion

Backtest Results (2+ Years, 535 Coins)

Scope note (2026-08-11): 535 coins is the original Binance-era run (Dec 2023–Feb 2026). The frontmatter shows coins: 235, which reflects the current OKX universe. Both numbers are correct for their respective contexts; 535 is preserved for the historical record. The kill verdict holds in both universes.

• Total Trades: 2,898
• Win Rate: 68.6%
• Profit Factor: 2.22
• Average Win: +4.8%
• Average Loss: -5.2%
• SL Hit Rate: 7.5%
• TP Hit Rate: 42%
• TIMEOUT Rate: 50.5%

Why SHORT Only?

We originally tested both LONG and SHORT variants and the SHORT side looked far stronger in the 2026-04 sample:

• BB Squeeze SHORT: +$794, 68.6% win rate
• BB Squeeze LONG: -$26, 51.0% win rate
• Momentum LONG: Negative PnL, 37.5% win rate

At the time we read this as short-side mean reversion having a consistent edge in the 2024–2026 market structure. The 2026-06-28 re-test reframes it: the 2024–2026 window was a falling/choppy market, so the SHORT advantage is largely bear-beta — a short book wins because price fell, not because the squeeze filter found a durable edge. The LONG result simply confirms there was no upside edge either.

What We Learned

1. TP expansion beats SL reduction — Increasing TP from 6% to 8% improved PnL by 24.3%, while reducing SL made things worse
2. Time filtering matters — 7 blocked hours remove ~30% of losing trades
3. Volume is not optional — Without the 2.0x filter, false signals dominate
4. 48-hour timeout is optimal — Tested 30+ early exit variations, all underperformed

Simulation Assumptions

Scope note (2026-08-20): the costs below are the archived Binance-era run’s parameters (Dec 2023–Feb 2026), kept as a historical record — the same scope as the 535-coin note above. They are not what the live simulator charges today: the current engine applies 0.05% taker fee per side and slippage tiered by liquidity (0.05% Top-50 / 0.10% Top-200 / 0.20% for the rest), documented on /methodology/. The kill verdict holds under both cost models.

All results include realistic trading costs: 0.04% futures fee per side (0.08% round-trip) and 0.02% estimated slippage per trade. Position sizing: $60 per trade with 5x leverage. These costs are deducted from every simulated trade — no cherry-picking.

Status: KILLED (2026-06-28)

The 2026-04 backtest is reproducible across 535 coins and 2+ years of hourly data — that part still holds. But a fresh out-of-sample re-test killed the SHORT result on 3 adversarial axes in choppy markets: the edge was bear-beta and regime luck, not a standalone directional skill. We are keeping this page as a documented failure rather than deleting it. Default configuration shown ($60 per trade, 5x leverage, hourly scans) is what was tested; you can still run the simulation yourself, but do not treat this as a confirmed live edge.