Week 3: Lessons on Regimes and Record-Keeping
Lessons on market regimes, strategy overlap, and keeping a run record. A lessons post, not a results report.
Simulated paper trading, not live results; not financial advice. No real money was traded. This post makes no weekly performance claims. A later review of the ledger and decision logs we kept found that the simulated run (Feb 26 to Apr 8, 2026) ended net negative; the figures and their caveats are in the retired-project case study. The bot has since been retired; see why we retired it.
Lesson: Classify the Market Regime Before Choosing a Strategy
Mean reversion tends to suit ranging markets and momentum tends to suit trending ones, so a bot that cannot tell the two apart will use the wrong tool. Common regime inputs include trend strength, volatility bands, and volume. Any classifier can mislabel a regime on thin or noisy data, so it should be validated on held-out history before it is trusted. We are describing a design principle here, not reporting a specific regime change from our run.
Lesson: Strategy Count Is Not Diversification
Strategies that share inputs tend to fire together. A system with many strategies may behave like far fewer independent bets, so measure overlap between strategies before counting them as diversification.
Lesson: Make Performance Visible, and Keep Records
Tracking cumulative profit and loss, drawdown, and risk-adjusted measures makes circuit breakers easier to respect. It only works if the underlying run record is retained. Our ledger and logs had gaps (documented in the retired-project case study), so we do not publish weekly figures here.
Where the Experiment Stands
The bot was simulated paper trading only and has been retired. The run ended net negative on a simulated basis; the numbers are in the retired-project case study. This is not a live result and not financial advice. Read why we retired it.
Related Reading
The trading bot has been retired. These posts cover the design and what we concluded.