Executive summary
This post documents a systematic backtest run on 9 crypto assets from the PAXOS exchange, including Bitcoin (BTC), using hourly data and a multi-signal technical framework combining Bollinger Bands, Ichimoku, and Chande Oscillator logic.
Between January 2022 and October 2025, the strategy compounded $100,000 into ~$485,000, delivering a 52.3% CAGR with a 15.5% maximum drawdown, under realistic constraints (leverage = 1, interest costs, position limits).
This article explains:
- The strategy logic
- The portfolio & execution constraints
- The risk metrics that actually matter
- What the results mean (and what they don’t)
⚠️ This is a research and engineering report, not financial advice.
Strategy overview
Signal architecture
The system is rule-based and fully systematic, combining entry and exit confirmation from multiple independent indicators.
Buy logic
bband2a(volatility & mean reversion context)ichimoku3a(trend regime & structure)
Sell logic
ichimoku3a(trend invalidation)cho3a(momentum exhaustion)
The goal is regime-aware trend capture, not scalping.
Timeframe & universe
| Parameter | Value |
|---|---|
| Exchange | PAXOS |
| Assets | 9 crypto pairs BTC ETH LTC BCH AAVE LINK UNI MATIC SOL |
| Bar size | 1 hour |
| Simulation start | 2022-01-01 |
| End date | 2025-10-03 |
| Base currency | USD |
Portfolio & risk constraints
This is not an unconstrained backtest.
Key constraints:
- Initial capital: $100,000
- Leverage: 1.0 (spot-like behavior)
- Interest rate: 7% annual
- Target position weight: 40%
- Max position weight: 60%
- Trailing stop: 35%
- Max entries per bar: 10
- Freeze logic: enabled (anti-overtrading)
The system is designed to survive real market conditions, not optimize paper metrics.
Performance summary
Core metrics
| Metric | Value |
|---|---|
| Final equity | $485,138 |
| Total return | +385% |
| CAGR | 52.3% |
| Max drawdown | -15.5% |
| Sharpe | 1.74 |
| Sortino | 2.57 |
| Calmar | 3.37 |
| Stability | 0.97 |
| Trades | 214 |
| Win rate | 56.1% |
Benchmark CAGR assumed: 5%
Why these numbers matter
- CAGR > 50% with DD < 16% → strong convexity
- Sharpe < 2 but Sortino > 2.5 → downside risk controlled
- Calmar > 3 → drawdown efficiency is the real alpha
- Stability ~0.97 → equity curve is not luck-driven noise
This is not a fragile, over-optimized curve.
Return distribution (important)
Key distribution stats per trade (%):
- Worst trade: -13.49%
- 1% quantile: -8.80%
- Median: +0.17%
- 95% quantile: +17.03%
- Best trade: +58.8%
➡️ Losses are bounded, winners are fat-tailed.
This is exactly what a trend-following crypto system should look like.
Exposure & execution behavior
- Net exposure: ~10x cumulative over time, not constant leverage
- Trades cluster during high-volatility regimes
- Flat or reduced exposure during chop → capital preservation
The strategy earns volatility, it does not fight it.
What this backtest does not prove
Let’s be clear:
❌ No guarantee of future performance
❌ No claim of optimality
❌ No curve-fitting denial (all models have bias)
This is one robust configuration, validated across:
- Multiple assets
- Multi-year period
- Different volatility regimes
Why this matters for KreamEdge
This backtest is part of the KreamEdge research pipeline:
- Modular signal design
- Strict risk constraints
- Reproducible backtests
- Live-ready execution logic
The goal is signal quality over signal quantity.
More research notes, live experiments, and tooling breakdowns will follow.
Final thoughts
If you’re serious about crypto trading:
- Focus on drawdown control
- Optimize for stability, not peak return
- Prefer simple signals + strong risk logic
- Always assume regime shifts
This system passes the first filter: it survives reality.
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.
Join our free community channels for backtest output and strategy commentary.
0 Comments