Agent 01 / Trend
Follow persistence.
Compares moving averages. Tests three window pairs. Buys only after a signal, at the next observation.
Hydra / Trading agents
Test the idea.
Keep the evidence.
Two rules-based agents. One shared simulation engine. Train on the past, select on validation data, then challenge the winner on a reserved test segment.
RESEARCH ONLY · 10,000 paper points · no wallet accessProfitable in a test ≠ proven live
The research bench
Ready to testNo result without a run.
Six fixed parameter candidates compete on training data. One finalist per family moves to validation. The selected winner sees the test segment once.
60% train20% validate20% test
Test return · after costs—
10% buy & hold—
Test drawdown—
Closed test trades—
Selected agent10% buy & hold
Inspect training & validation
Read the limitations
Includes decisions, trades, costs and data hash.
This session’s experiment log
All completed runs stay listed until you close or reload the page. Export each report to keep it. Reusing a test segment to tune ideas makes it training data.
Agent 02 / Reversion
Test the pullback.
Looks for a price below its rolling average and exits toward the mean. Tests three fixed parameter choices.
Review / Sentinel
Challenge the result.
Checks net results, a matched-exposure benchmark, trade count and drawdown. Cannot grant live authority.