29% Exposure – 1.66 Sharpe – 32% Max Drawdown (2018-2026)
Most crypto backtests lead with absurd equity curves.
This one doesn’t.
Let’s start with what actually matters:
- Net exposure: 29.4%
- Max drawdown: −32.3%
- Sharpe ratio: 1.66
- Calmar ratio: 2.10
- Ulcer index: 5.56%
- Trades: 71 over 8 years
CAGR is 67.7%, but that’s not the point.
The point is return per unit of exposure and drawdown control.
1. Strategy Design Philosophy
This is not a prediction model.
It is a regime participation system designed to:
- Enter volatility expansion phases
- Confirm structural trend alignment
- Exit aggressively when regime deteriorates
- Stay in cash most of the time
Average exposure over 8 years: 29%
It avoids the structural crypto bear phases instead of trying to survive them.
2. Architecture
Entry Layer
- Bollinger-based regime trigger (
bband2a) - Detects volatility compression → expansion
Exit Layer
Ensemble of:
ichimoku6aichimoku5aichimoku4a
Exit logic is deliberately redundant.
Risk Layer
- 32% trailing stop
- No leverage
- Position target: 40%
- Max position cap: 60%
- 7% annual interest on idle cash
Universe:
- 9 liquid USD crypto pairs
- Daily bars
- Long-only
3. Configuration
Strategy Configuration
| Category | Parameter | Value |
|---|---|---|
| Core Setup | Bar Size | 1 Day |
| Universe Size | 9 Crypto Pairs (USD) | |
| Leverage | 1.0 (No leverage) | |
| Initial Capital | $100,000 | |
| Interest on Cash | 7% Annual | |
| Signal Engine | Entry Logic | Bollinger Regime (bband2a) |
| Exit Logic | Ichimoku Ensemble (6a + 5a + 4a) | |
| Scoring Model | Weighted Composite (16 factors) | |
| Freeze Bars | 2 Bars | |
| Risk Management | Trailing Stop | 32% |
| Position Target Weight | 40% | |
| Max Position Weight | 60% | |
| Max Entries per Bar | 10 | |
| Exposure Controls | Exposure Penalty Trigger | 5% |
| Stability Penalty Trigger | 0.30 | |
| Backtest Period | Data Start | 20 Dec 2016 |
| Simulation Start | 02 Jan 2018 – 31 Jan 2026 |
4. Performance Summary (2018-2026)
Initial capital: $100,000
Final equity: $6,531,155
But again, focus on structure:
Performance Summary (2018-2026)
| Metric | Value |
|---|---|
| Final Equity | $6,531,156 |
| Total Return | +6,431% |
| CAGR | 67.74% |
| Max Drawdown | −32.26% |
| Ulcer Index | 5.56% |
| Sharpe Ratio | 1.66 |
| Sortino Ratio | 2.53 |
| Calmar Ratio | 2.10 |
| Stability | 0.935 |
| K-Ratio | 203.41 |
| Net Exposure | 29.40% |
| Trades | 71 |
| Win Rate | 60.56% |
| Risk / Reward | 7.76 |
| Kelly Fraction | 55.48% |
71 trades in 8 years.
This is a slow, selective trend system, not hyperactive trading.
5. Distribution Profile
Trade return distribution:
- Median: +3.0%
- 80th percentile: +54%
- 95th percentile: +192%
- Worst trade: −25.9%
This is clearly positively skewed.
Performance is driven by:
- Few large structural trend captures
- Limited downside per trade
- Strict exit discipline
It does not rely on high win rate.
It relies on asymmetric payoff.
6. Risk Characteristics
Drawdown Profile
Max DD: −32%
In crypto terms, that is materially lower than passive exposure.
More important:
Ulcer index = 5.56%
Meaning drawdowns are not only shallow, but relatively short-lived.
Stability
Stability score: 0.935
K-ratio: 203
The equity curve is statistically smooth relative to asset volatility.
That suggests:
- Regime filtering is doing real work
- Not pure beta harvesting
6. What This Is NOT
Let’s be explicit:
- Not slippage-stress-tested yet
- No liquidity impact modeling
- Universe limited to 9 assets
- Crypto structural bull market tailwind (2020-2021)
- Trailing stop execution on daily bars may overestimate fills
This is research-grade, not audited live performance.
7. Why This Is Interesting
The key metric here is:
67% CAGR at 29% exposure
If exposure were 100%, risk profile would be completely different.
This suggests the system is:
- Efficient at timing participation
- Avoiding large bear regimes
- Capturing convexity phases
That is structurally different from buy-and-hold.
8. What Needs to Be Proven Next
Before considering capital deployment:
- Walk-forward optimization
- Parameter perturbation test (±20% sensitivity)
- Slippage stress test (0.1-0.5%)
- Sub-universe validation
- Post-2024 strict out-of-sample monitoring
- Exposure cap reduction to 20% to test robustness
If performance collapses under minor perturbations, edge is fragile.
If not, we may have something durable.
9. Bottom Line
This is not about turning $100k into $6M.
It’s about:
- Controlled participation
- Positive skew
- Low exposure efficiency
- Stable equity compounding
The next step is robustness validation – not marketing.
Out-of-sample cutoff. This strategy’s rules and parameters were frozen on 3 July 2025. All performance shown after that date is genuine out-of-sample / forward-tracked data – it post-dates the freeze, so no hindsight or selection could have shaped the rules.
This is a historical backtest, published for informational and educational purposes only – not financial advice, not a recommendation, and not a trading signal. Past performance is not indicative of future results.
Backtest output and ongoing research:
→ Community channels
→ https://x.com/kreamedge
See the discussion on X/Twitter.
Frequently asked questions
What does this crypto backtest test?
A long only, daily bar systematic strategy on nine liquid USD crypto pairs from 2 January 2018 to 31 January 2026, starting from 100,000 USD with no leverage.
Why lead with exposure instead of CAGR?
Because average net exposure is 29.4%, so the book is in cash roughly seven days in ten. The CAGR of 67.7% was earned on that fraction of participation. Return per unit of exposure and drawdown control are the informative measures; the headline growth number on its own is not.
How deep did the drawdown go?
The maximum drawdown was 32.3%, with an Ulcer index of 5.56%. Sharpe was 1.66 and Calmar 2.10 over 71 trades in eight years.
How does the strategy handle crypto bear markets?
It exits rather than endures. Entry is a Bollinger based volatility expansion trigger, the exit layer is a deliberately redundant Ichimoku ensemble, and a 32% trailing stop caps the loss on any single position. The design avoids the structural bear phases instead of trying to sit through them.
Is 71 trades enough to draw conclusions from?
It is a small sample and the post treats it as one. Seventy one trades over eight years means the result rests on a handful of regime participations, so the uncertainty around any risk adjusted statistic here is wide. Sample size is part of the evidence, not a footnote to it.
0 Comments