Feb 2026 Crash: What Our SHORT System Did

Market Context

February 2026 was brutal for crypto:

BTC: ATH $126K → $60K (-48% from highs)
Feb 7: $233M in short liquidations in one day. Fear & Greed Index hit 6 (near all-time low).
Feb 5: 311,000+ traders liquidated in 24 hours (CoinGlass)
Altcoins: Relative strength vs BTC (mean reversion from oversold)

For context: derivatives make up ~79% of all crypto trading volume (CoinMarketCap). Most of those liquidated traders were using leverage they didn’t fully understand.

What Our System Did

BB Squeeze SHORT runs automatically. It doesn’t know about news, doesn’t read Twitter, doesn’t panic. It follows rules.

Week 1 (Feb 7-12):

The short squeeze on Feb 7 hit hard. 75 positions were open.
16 stop-losses triggered. 4 take-profits.
Unrealized PnL: -$190
Portfolio dropped -6% in 48 hours
Daily loss limit reached 86% of the 7% cap

Week 2 (Feb 12-15):

48-hour timeout wave: 62+ positions hit max holding period
Positions dropped from 79 → 14 → 10
Average timeout PnL: -2.14%
Total drawdown reached -8.3%

Was This Normal?

We checked the 2-year backtest (762 trading days):

Metric	Feb 2026 (worst 3 days)	Backtest (worst 3 days)	Ratio
Loss	-6.4%	-33%	17% of worst case
Max consecutive loss days	3	10	Much better
MDD	4.5%	33%	Well within range
3-day periods worse than ours	—	84 out of 762 (11.1%)	1 in 9 days

Conclusion: The worst period was in the bottom 17% of what the backtest showed was possible. The system is performing within expected parameters.

What We Did NOT Do

Did not turn off the system — The temptation to “stop the bleeding” is real. But the backtest shows 10+ consecutive loss days happen, and the system recovers.
Did not add a BTC regime filter — 6 expert analysts recommended it. We backtested it. All 4 variations made things worse. (Read the full analysis)
Did not reduce position size — Risk analysis showed $60 is the Calmar-optimal size. Reducing to $30 would cut both risk and upside equally.
Did not add time filters for 16-18 UTC — The 7-day live data suggested these hours were bad. 2-year backtest showed this was noise. Adding the filter would have cost -$1,234 in PnL.

Key Takeaway

Markets crash. Systems get tested. The question isn’t “did it lose money during the crash?” — it’s “did it lose more than the backtest predicted?”

In our case: no. Not even close.

This doesn’t guarantee future performance. But it means the system is working as designed, within the risk parameters we set.

That’s what verification looks like in practice.

All data sourced from backtests run on 535 coins with 2+ years of hourly OHLCV data on Binance Futures. Full methodology documented in our changelog.

Not financial advice. Past performance does not guarantee future results.