Analytics
Every breakdown the experiment produced, over May 27 to July 20, 2026.
Read these charts carefully. They are computed on 4 executed trades, 2 of which were closed by human intervention. At that sample size a win rate, an hourly breakdown, or a per-asset comparison is noise, not signal. Nothing on this page supports a conclusion about whether the approach works.
Equity Curve
Tracks how the simulated $10,000 portfolio evolves over time, compared to simply holding Bitcoin. No real money is involved.
Sim. Monthly PnL
Month-over-month change in simulated portfolio value, as a percentage of starting capital.
Sim. Win Rate by Asset
Percentage of simulated trades that ended positive, grouped by cryptocurrency.
Sim. Confidence vs Outcome
Does higher AI confidence correlate with better simulated outcomes?
Sim. PnL Distribution
Distribution of simulated trade outcomes across PnL ranges (2% buckets).
Sim. Win Rate by Hour
Simulated win rate by the UTC hour when the observation was generated.
Sim. Avg PnL by Close Reason
Average simulated outcome depending on how the position was closed.
Performance by prompt version
The system prompt was revised during the experiment. HOLD rate is the meaningful column here: the trade columns rest on too few trades to compare versions.