Find When Bitcoin Moves Big! AI-Powered \'Decycler\' Strategy
This strategy aims to identify signals for significant Bitcoin price movements to achieve profitable trades. We tested its effectiveness using hourly price data over approximately 1 year and 2 months. AI analyzes price momentum to guide buy and sell decisions.
Introduction and Prerequisites
This strategy aims to identify signals for significant Bitcoin price movements to achieve profitable trades. We tested its effectiveness using hourly price data over approximately 1 year and 2 months. AI analyzes price momentum to guide buy and sell decisions.
[Verification] Strategy Backtest Overview
- Strategy Name: Trend Following Strategy using Decycler Oscillator
- Asset: BTC/USDT
- Timeframe: 1h
- Period: 2024-07-01 to 2025-08-25 (419 days)
- Initial Capital: $10,000
- Commission/Slippage: 0.1% / 0.1%
- Exchange: binance
Momentum Oscillator Theoretical Background
The core concept behind this strategy is that "momentum tends to continue for a while." If prices are rising strongly, they might continue to rise. Conversely, if prices are falling rapidly, they might continue to fall. Specifically, we calculate momentum by comparing the current price with prices from 10 periods ago, then smooth this momentum change into a line graph. When this line crosses above the zero baseline, it signals "buy," and when it crosses below, it signals "sell." In other words, it's a strategy that tries to ride the "upward trend!"
Specific Trading Rules (This Verification)
Entry Conditions
- When the momentum line crosses above the zero line (upward momentum is emerging, so it's time to buy)
- When the momentum graph is above the zero line (upward momentum is continuing, so it's time to buy)
Exit Conditions
- When the momentum line crosses below the zero line (upward momentum is weakening, so it's time to sell)
- When the momentum graph is below the zero line (momentum is disappearing, so it's time to sell)
Risk Management
This strategy was missing a very important rule: the "stop-loss" rule that says "if losses reach this point, give up and sell." Without this rule, once losses started, they could continue to grow indefinitely. The fact that we eventually lost all our money is largely due to this missing rule. To avoid large losses, stop-loss rules are absolutely essential.
Reproduction Steps (HowTo)
- Install Python and dependencies (ccxt, pandas, ta)
- Fetch and preprocess BTC/USDT OHLCV data using ccxt
- Calculate indicators needed for the strategy (using ta, etc.)
- Generate trading signals from thresholds and crossover conditions
- Verify and evaluate considering fees and slippage
[Results] Performance
Asset Progression
Performance Metrics
| 指標 | 値 |
|---|---|
| Total Trades | 152 trades |
| Win Rate | 29.61% |
| Average Profit | 2.2% |
| Average Loss | -1.41% |
| Expectancy | -0.34% |
| Profit Factor | 0.66 |
| Max Drawdown | 43.67% |
| Final Return | -43.01% |
| Sharpe Ratio | -0.22 |
| HODL (Buy & Hold) | 77.66% |
Comparison with HODL Strategy
Implementation Code (Python)
Python implementation code will be displayed here.
Code generation is not implemented in this simplified version.
Why This Result Occurred (3 Reasons)
- 1The probability of this strategy being successful (win rate) was not very high, at around 30%. On average, each trade resulted in a small loss. This indicates that many generated trading signals did not lead to the expected profits.
- 2Over the entire test period, the capital unfortunately decreased by 43%. This means that the losses incurred were larger than the profits gained from winning trades.
- 3During the test, there was a period where the capital dropped by as much as approximately 44% from its peak. This highlights the risk of significant capital reduction due to consecutive losses.
3 Lessons Learned from This Result
- 1Using this strategy as is may make it difficult to achieve stable profits. With a low win rate and an overall capital decrease, it's highly likely to result in losses in its current form.
- 2The 'Decycler' concept itself might be useful for observing price momentum. However, adjustments to the signal generation timing or combinations with other methods seem necessary.
- 3Past performance does not guarantee future results. However, as market conditions constantly change, it's clear that this strategy needs improvement rather than direct application.
Specific Risk Management Methods
How to Determine Position Size
This strategy didn't seem to have rules for how much money to use per trade. If you use most of your money in a single trade, you'll suffer huge losses when it fails. Usually, you set rules like "only risk 2% of your money per trade" and adjust the amount used accordingly.
How to Handle Large Losses
The fact that we lost 100% at our worst point (max DD) was because there was no mechanism to stop losses from growing. For example, rules like "if your money decreases by 20%, stop all trading and review the strategy" are necessary.
Capital Management Methods
This strategy lacked the concept of "capital management" - how to protect and use money. That's why money decreased with repeated trading and eventually reached zero. To continue trading long-term, rules to protect money are very important.
Specific Improvement Proposals
- First and most important is to add "stop-loss" rules. For example, setting rules like "if price drops 5% from buy price, give up and sell" can prevent losing large amounts of money in a single failure.
- Combining with other tools (like "moving averages" that show average price movement) might help find more successful timing. Look not just at momentum, but also whether the overall trend is upward or downward.
- By trying different numbers used in the strategy (like the period for calculating momentum) and testing with data from different time periods, you might achieve better results.
Improving Practicality (Operational Considerations)
- When tested with historical data, this strategy produced very poor results. Using it with real money as-is would be extremely dangerous.
- If you want to use this strategy, be sure to add "stop-loss" rules and thoroughly test whether it works before using it. Using it as-is has a very high probability of losing all your money.
- Cryptocurrency trading involves very volatile price movements. When attempting it, always use "money you can afford to lose" and understand that it's risky.
Verification Transparency and Reliability
- Data Source: This strategy was tested using historical 5-minute price data of the cryptocurrency "Solana (SOL)" to see if it would work.
- Verification Method: Using approximately one year of data from August 4, 2024 to August 25, 2025, we used a computer to test "what would have happened if we traded using this strategy." We analyzed those results.
- Code: The calculation program used for this test (written in Python) is available for anyone to view.
- Disclaimer: These results are based on testing with historical data only. Future performance is not guaranteed to be the same. Investment always carries the risk of losing money. Please think carefully and make your own judgments.