Hi everyone, this is SuperTommi.
As I mentioned in my previous updates, ‘raising’ an AI agent is a continuous process of trial, error, and refinement. Today, I want to take a deep dive into the **pros, cons, and specific reasons** behind the major architectural upgrades of our Bitcoin trading algorithm, from Version 1.0 to our current **Version 3.0**.
Version 1.0: The Naive Initializer
When we started, the logic was simple: Buy when indicators show a bounce and keep a tight leash.
- Pros: Extremely active. It would trade multiple times an hour, providing plenty of data points.
- Cons: Vulnerable to the ‘Fee Trap.’ With a fixed -1.0% stop-loss, we were doomed from the start. Combined with the 1.2% round-trip fee on Swyftx, even a ‘breakeven’ trade resulted in a net loss.
- Why we updated: We realized that Bitcoin’s ‘market noise’ is often greater than 1.0%. Beta was getting stopped out by standard volatility, not real trend reversals.
Version 2.0: The Conservative Guardian
To fix v1.0, we introduced professional risk management and a ‘big brother’ filter.
- Pros: Significantly more stable. By using ATR (Average True Range), the stop-loss moved with the market’s breathing. The 1-hour trend filter ensured we only bought during macro uptrends.
- Cons: It was *too* quiet. During sideways markets—which make up 70% of crypto price action—Beta would sit on the sidelines for days, missing out on profitable 2-3% swings within a box range.
- Why we updated: Stability is good, but missing opportunities is a cost in itself. We needed a way for Beta to recognize different market ‘personalities.’
Version 3.0: The Versatile Predator
Our latest version is designed to be as versatile as the market itself. This is where we are today.
- The Solution – Mode Switching: Beta now measures the ‘Bollinger Band Width.’ If the bands are tight, it assumes a Sideways Market and switches to ‘Range Mode.’ If the bands expand, it switches to Trend Mode.
- The Goldilocks Timeframe: We moved from a 1-hour filter to a 15-minute primary filter. It’s fast enough to catch daily swings but slow enough to ignore the 1-minute chaos.
- RR (Risk-Reward) Gatekeeper: Before any trade, Beta now checks if there is at least 2.5% of ‘room to grow’ to the next resistance. This ensures we always have enough margin to cover fees and net a clean profit.
Summary: Why This Evolution Matters
Trading is not a static game. By migrating our infrastructure to Sydney for lower latency and upgrading the logic to v3.0, we’ve given Beta the tools to survive both a pumping trend and a boring range. We’ve moved from a ‘fixed script’ to an ‘adaptive system.’
I’m excited to see how Version 3.0 handles the 1,000 NZD seed. Follow along as we continue to push the boundaries of Agentic AI in Finance!
Best regards,
SuperTommi.